summaryrefslogtreecommitdiffstats
path: root/src/3rdparty/pcre2/src/pcre2_compile.c
blob: 068735ae8eed65c9e7bb6da26377ff5a67bb3a79 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
8882
8883
8884
8885
8886
8887
8888
8889
8890
8891
8892
8893
8894
8895
8896
8897
8898
8899
8900
8901
8902
8903
8904
8905
8906
8907
8908
8909
8910
8911
8912
8913
8914
8915
8916
8917
8918
8919
8920
8921
8922
8923
8924
8925
8926
8927
8928
8929
8930
8931
8932
8933
8934
8935
8936
8937
8938
8939
8940
8941
8942
8943
8944
8945
8946
8947
8948
8949
8950
8951
8952
8953
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
8985
8986
8987
8988
8989
8990
8991
8992
8993
8994
8995
8996
8997
8998
8999
9000
9001
9002
9003
9004
9005
9006
9007
9008
9009
9010
9011
9012
9013
9014
9015
9016
9017
9018
9019
9020
9021
9022
9023
9024
9025
9026
9027
9028
9029
9030
9031
9032
9033
9034
9035
9036
9037
9038
9039
9040
9041
9042
9043
9044
9045
9046
9047
9048
9049
9050
9051
9052
9053
9054
9055
9056
9057
9058
9059
9060
9061
9062
9063
9064
9065
9066
9067
9068
9069
9070
9071
9072
9073
9074
9075
9076
9077
9078
9079
9080
9081
9082
9083
9084
9085
9086
9087
9088
9089
9090
9091
9092
9093
9094
9095
9096
9097
9098
9099
9100
9101
9102
9103
9104
9105
9106
9107
9108
9109
9110
9111
9112
9113
9114
9115
9116
9117
9118
9119
9120
9121
9122
9123
9124
9125
9126
9127
9128
9129
9130
9131
9132
9133
9134
9135
9136
9137
9138
9139
9140
9141
9142
9143
9144
9145
9146
9147
9148
9149
9150
9151
9152
9153
9154
9155
9156
9157
9158
9159
9160
9161
9162
9163
9164
9165
9166
9167
9168
9169
9170
9171
9172
9173
9174
9175
9176
9177
9178
9179
9180
9181
9182
9183
9184
9185
9186
9187
9188
9189
9190
9191
9192
9193
9194
9195
9196
9197
9198
9199
9200
9201
9202
9203
9204
9205
9206
9207
9208
9209
9210
9211
9212
9213
9214
9215
9216
9217
9218
9219
9220
9221
9222
9223
9224
9225
9226
9227
9228
9229
9230
9231
9232
9233
9234
9235
9236
9237
9238
9239
9240
9241
9242
9243
9244
9245
9246
9247
9248
9249
9250
9251
9252
9253
9254
9255
9256
9257
9258
9259
9260
9261
9262
9263
9264
9265
9266
9267
9268
9269
9270
9271
9272
9273
9274
9275
9276
9277
9278
9279
9280
9281
9282
9283
9284
9285
9286
9287
9288
9289
9290
9291
9292
9293
9294
9295
9296
9297
9298
9299
9300
9301
9302
9303
9304
9305
9306
9307
9308
9309
9310
9311
9312
9313
9314
9315
9316
9317
9318
9319
9320
9321
9322
9323
9324
9325
9326
9327
9328
9329
9330
9331
9332
9333
9334
9335
9336
9337
9338
9339
9340
9341
9342
9343
9344
9345
9346
9347
9348
9349
9350
9351
9352
9353
9354
9355
9356
9357
9358
9359
9360
9361
9362
9363
9364
9365
9366
9367
9368
9369
9370
9371
9372
9373
9374
9375
9376
9377
9378
9379
9380
9381
9382
9383
9384
9385
9386
9387
9388
9389
9390
9391
9392
9393
9394
9395
9396
9397
9398
9399
9400
9401
9402
9403
9404
9405
9406
9407
9408
9409
9410
9411
9412
9413
9414
9415
9416
9417
9418
9419
9420
9421
9422
9423
9424
9425
9426
9427
9428
9429
9430
9431
9432
9433
9434
9435
9436
9437
9438
9439
9440
9441
9442
9443
9444
9445
9446
9447
9448
9449
9450
9451
9452
9453
9454
9455
9456
9457
9458
9459
9460
9461
9462
9463
9464
9465
9466
9467
9468
9469
9470
9471
9472
9473
9474
9475
9476
9477
9478
9479
9480
9481
9482
9483
9484
9485
9486
9487
9488
9489
9490
9491
9492
9493
9494
9495
9496
9497
9498
9499
9500
9501
9502
9503
9504
9505
9506
9507
9508
9509
9510
9511
9512
9513
9514
9515
9516
9517
9518
9519
9520
9521
9522
9523
9524
9525
9526
9527
9528
9529
9530
9531
9532
9533
9534
9535
9536
9537
9538
9539
9540
9541
9542
9543
9544
9545
9546
9547
9548
9549
9550
9551
9552
9553
9554
9555
9556
9557
9558
9559
9560
9561
9562
9563
9564
9565
9566
9567
9568
9569
9570
9571
9572
9573
9574
9575
9576
9577
9578
9579
9580
9581
9582
9583
9584
9585
9586
9587
9588
9589
9590
9591
9592
9593
9594
9595
9596
9597
9598
9599
9600
9601
9602
9603
9604
9605
9606
9607
9608
9609
9610
9611
9612
9613
9614
9615
9616
9617
9618
9619
9620
9621
9622
9623
9624
9625
9626
9627
9628
9629
9630
9631
9632
9633
9634
9635
9636
9637
9638
9639
9640
9641
9642
9643
9644
9645
9646
9647
9648
9649
9650
9651
9652
9653
9654
9655
9656
9657
9658
9659
9660
9661
9662
9663
9664
9665
9666
9667
9668
9669
9670
9671
9672
9673
9674
9675
9676
9677
9678
9679
9680
9681
9682
9683
9684
9685
9686
9687
9688
9689
9690
9691
9692
9693
9694
9695
9696
9697
9698
9699
9700
9701
9702
9703
9704
9705
9706
9707
9708
9709
9710
9711
9712
9713
9714
9715
9716
9717
9718
9719
9720
9721
9722
9723
9724
9725
9726
9727
9728
9729
9730
9731
9732
9733
9734
9735
9736
9737
9738
9739
9740
9741
9742
9743
9744
9745
9746
9747
9748
9749
9750
9751
9752
9753
9754
9755
9756
9757
9758
9759
9760
9761
9762
9763
9764
9765
9766
9767
9768
9769
9770
9771
9772
9773
9774
9775
9776
9777
9778
9779
9780
9781
9782
9783
9784
9785
9786
9787
9788
9789
9790
9791
9792
9793
9794
9795
9796
9797
9798
9799
9800
9801
9802
9803
9804
9805
9806
9807
9808
9809
9810
9811
9812
9813
9814
9815
9816
9817
9818
9819
9820
9821
9822
9823
9824
9825
9826
9827
9828
9829
9830
9831
9832
9833
9834
9835
9836
9837
9838
9839
9840
9841
9842
9843
9844
9845
9846
9847
9848
9849
9850
9851
9852
9853
9854
9855
9856
9857
9858
9859
9860
9861
9862
9863
9864
9865
9866
9867
9868
9869
9870
9871
9872
9873
9874
9875
9876
9877
9878
9879
9880
9881
9882
9883
9884
9885
9886
9887
9888
9889
9890
9891
9892
9893
9894
9895
9896
9897
9898
9899
9900
9901
9902
9903
9904
9905
9906
9907
9908
9909
9910
9911
9912
9913
9914
9915
9916
9917
9918
9919
9920
9921
9922
9923
9924
9925
9926
9927
9928
9929
9930
9931
9932
9933
9934
9935
9936
9937
9938
9939
9940
9941
9942
9943
9944
9945
9946
9947
9948
9949
9950
9951
9952
9953
9954
9955
9956
9957
9958
9959
9960
9961
9962
9963
9964
9965
9966
9967
9968
9969
9970
9971
9972
9973
9974
9975
9976
9977
9978
9979
9980
9981
9982
9983
9984
9985
9986
9987
9988
9989
9990
9991
9992
9993
9994
9995
9996
9997
9998
9999
10000
10001
10002
10003
10004
10005
10006
10007
10008
10009
10010
10011
10012
10013
10014
10015
10016
10017
10018
10019
10020
10021
10022
10023
10024
10025
10026
10027
10028
10029
10030
10031
10032
10033
10034
10035
10036
10037
10038
10039
10040
10041
10042
10043
10044
10045
10046
10047
10048
10049
10050
10051
10052
10053
10054
10055
10056
10057
10058
10059
10060
10061
10062
10063
10064
10065
10066
10067
10068
10069
10070
10071
10072
10073
10074
10075
10076
10077
10078
10079
10080
10081
10082
10083
10084
10085
10086
10087
10088
10089
10090
10091
10092
10093
10094
10095
10096
10097
10098
10099
10100
10101
10102
10103
10104
10105
10106
10107
10108
10109
10110
10111
10112
10113
10114
10115
10116
10117
10118
10119
10120
10121
10122
10123
10124
10125
10126
10127
10128
10129
10130
10131
10132
10133
10134
10135
10136
10137
10138
10139
10140
10141
10142
10143
10144
10145
10146
10147
10148
10149
10150
10151
10152
10153
10154
10155
10156
10157
10158
10159
10160
10161
10162
10163
10164
10165
10166
10167
10168
10169
10170
10171
10172
10173
10174
10175
10176
10177
10178
10179
10180
10181
10182
10183
10184
10185
10186
10187
10188
10189
10190
10191
10192
10193
10194
10195
10196
10197
10198
10199
10200
10201
10202
10203
10204
10205
10206
10207
10208
10209
10210
10211
10212
10213
10214
10215
10216
10217
10218
10219
10220
10221
10222
10223
10224
10225
10226
10227
10228
10229
10230
10231
10232
10233
/*************************************************
*      Perl-Compatible Regular Expressions       *
*************************************************/

/* PCRE is a library of functions to support regular expressions whose syntax
and semantics are as close as possible to those of the Perl 5 language.

                       Written by Philip Hazel
     Original API code Copyright (c) 1997-2012 University of Cambridge
          New API code Copyright (c) 2016-2019 University of Cambridge

-----------------------------------------------------------------------------
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

    * Redistributions of source code must retain the above copyright notice,
      this list of conditions and the following disclaimer.

    * Redistributions in binary form must reproduce the above copyright
      notice, this list of conditions and the following disclaimer in the
      documentation and/or other materials provided with the distribution.

    * Neither the name of the University of Cambridge nor the names of its
      contributors may be used to endorse or promote products derived from
      this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
-----------------------------------------------------------------------------
*/


#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#define NLBLOCK cb             /* Block containing newline information */
#define PSSTART start_pattern  /* Field containing processed string start */
#define PSEND   end_pattern    /* Field containing processed string end */

#include "pcre2_internal.h"

/* In rare error cases debugging might require calling pcre2_printint(). */

#if 0
#ifdef EBCDIC
#define PRINTABLE(c) ((c) >= 64 && (c) < 255)
#else
#define PRINTABLE(c) ((c) >= 32 && (c) < 127)
#endif
#include "pcre2_printint.c"
#define DEBUG_CALL_PRINTINT
#endif

/* Other debugging code can be enabled by these defines. */

/* #define DEBUG_SHOW_CAPTURES */
/* #define DEBUG_SHOW_PARSED */

/* There are a few things that vary with different code unit sizes. Handle them
by defining macros in order to minimize #if usage. */

#if PCRE2_CODE_UNIT_WIDTH == 8
#define STRING_UTFn_RIGHTPAR     STRING_UTF8_RIGHTPAR, 5
#define XDIGIT(c)                xdigitab[c]

#else  /* Either 16-bit or 32-bit */
#define XDIGIT(c)                (MAX_255(c)? xdigitab[c] : 0xff)

#if PCRE2_CODE_UNIT_WIDTH == 16
#define STRING_UTFn_RIGHTPAR     STRING_UTF16_RIGHTPAR, 6

#else  /* 32-bit */
#define STRING_UTFn_RIGHTPAR     STRING_UTF32_RIGHTPAR, 6
#endif
#endif

/* Macros to store and retrieve a PCRE2_SIZE value in the parsed pattern, which
consists of uint32_t elements. Assume that if uint32_t can't hold it, two of
them will be able to (i.e. assume a 64-bit world). */

#if PCRE2_SIZE_MAX <= UINT32_MAX
#define PUTOFFSET(s,p) *p++ = s
#define GETOFFSET(s,p) s = *p++
#define GETPLUSOFFSET(s,p) s = *(++p)
#define READPLUSOFFSET(s,p) s = p[1]
#define SKIPOFFSET(p) p++
#define SIZEOFFSET 1
#else
#define PUTOFFSET(s,p) \
  { *p++ = (uint32_t)(s >> 32); *p++ = (uint32_t)(s & 0xffffffff); }
#define GETOFFSET(s,p) \
  { s = ((PCRE2_SIZE)p[0] << 32) | (PCRE2_SIZE)p[1]; p += 2; }
#define GETPLUSOFFSET(s,p) \
  { s = ((PCRE2_SIZE)p[1] << 32) | (PCRE2_SIZE)p[2]; p += 2; }
#define READPLUSOFFSET(s,p) \
  { s = ((PCRE2_SIZE)p[1] << 32) | (PCRE2_SIZE)p[2]; }
#define SKIPOFFSET(p) p += 2
#define SIZEOFFSET 2
#endif

/* Macros for manipulating elements of the parsed pattern vector. */

#define META_CODE(x)   (x & 0xffff0000u)
#define META_DATA(x)   (x & 0x0000ffffu)
#define META_DIFF(x,y) ((x-y)>>16)

/* Function definitions to allow mutual recursion */

#ifdef SUPPORT_UNICODE
static unsigned int
  add_list_to_class_internal(uint8_t *, PCRE2_UCHAR **, uint32_t,
    compile_block *, const uint32_t *, unsigned int);
#endif

static int
  compile_regex(uint32_t, PCRE2_UCHAR **, uint32_t **, int *, uint32_t,
    uint32_t *, int32_t *, uint32_t *, int32_t *, branch_chain *,
    compile_block *, PCRE2_SIZE *);

static int
  get_branchlength(uint32_t **, int *, int *, parsed_recurse_check *,
    compile_block *);

static BOOL
  set_lookbehind_lengths(uint32_t **, int *, int *, parsed_recurse_check *,
    compile_block *);



/*************************************************
*      Code parameters and static tables         *
*************************************************/

#define MAX_GROUP_NUMBER   65535u
#define MAX_REPEAT_COUNT   65535u
#define REPEAT_UNLIMITED   (MAX_REPEAT_COUNT+1)

/* COMPILE_WORK_SIZE specifies the size of stack workspace, which is used in
different ways in the different pattern scans. The parsing and group-
identifying pre-scan uses it to handle nesting, and needs it to be 16-bit
aligned for this. Having defined the size in code units, we set up
C16_WORK_SIZE as the number of elements in the 16-bit vector.

During the first compiling phase, when determining how much memory is required,
the regex is partly compiled into this space, but the compiled parts are
discarded as soon as they can be, so that hopefully there will never be an
overrun. The code does, however, check for an overrun, which can occur for
pathological patterns. The size of the workspace depends on LINK_SIZE because
the length of compiled items varies with this.

In the real compile phase, this workspace is not currently used. */

#define COMPILE_WORK_SIZE (3000*LINK_SIZE)   /* Size in code units */

#define C16_WORK_SIZE \
  ((COMPILE_WORK_SIZE * sizeof(PCRE2_UCHAR))/sizeof(uint16_t))

/* A uint32_t vector is used for caching information about the size of
capturing groups, to improve performance. A default is created on the stack of
this size. */

#define GROUPINFO_DEFAULT_SIZE 256

/* The overrun tests check for a slightly smaller size so that they detect the
overrun before it actually does run off the end of the data block. */

#define WORK_SIZE_SAFETY_MARGIN (100)

/* This value determines the size of the initial vector that is used for
remembering named groups during the pre-compile. It is allocated on the stack,
but if it is too small, it is expanded, in a similar way to the workspace. The
value is the number of slots in the list. */

#define NAMED_GROUP_LIST_SIZE  20

/* The pre-compiling pass over the pattern creates a parsed pattern in a vector
of uint32_t. For short patterns this lives on the stack, with this size. Heap
memory is used for longer patterns. */

#define PARSED_PATTERN_DEFAULT_SIZE 1024

/* Maximum length value to check against when making sure that the variable
that holds the compiled pattern length does not overflow. We make it a bit less
than INT_MAX to allow for adding in group terminating code units, so that we
don't have to check them every time. */

#define OFLOW_MAX (INT_MAX - 20)

/* Code values for parsed patterns, which are stored in a vector of 32-bit
unsigned ints. Values less than META_END are literal data values. The coding
for identifying the item is in the top 16-bits, leaving 16 bits for the
additional data that some of them need. The META_CODE, META_DATA, and META_DIFF
macros are used to manipulate parsed pattern elements.

NOTE: When these definitions are changed, the table of extra lengths for each
code (meta_extra_lengths, just below) must be updated to remain in step. */

#define META_END              0x80000000u  /* End of pattern */

#define META_ALT              0x80010000u  /* alternation */
#define META_ATOMIC           0x80020000u  /* atomic group */
#define META_BACKREF          0x80030000u  /* Back ref */
#define META_BACKREF_BYNAME   0x80040000u  /* \k'name' */
#define META_BIGVALUE         0x80050000u  /* Next is a literal > META_END */
#define META_CALLOUT_NUMBER   0x80060000u  /* (?C with numerical argument */
#define META_CALLOUT_STRING   0x80070000u  /* (?C with string argument */
#define META_CAPTURE          0x80080000u  /* Capturing parenthesis */
#define META_CIRCUMFLEX       0x80090000u  /* ^ metacharacter */
#define META_CLASS            0x800a0000u  /* start non-empty class */
#define META_CLASS_EMPTY      0x800b0000u  /* empty class */
#define META_CLASS_EMPTY_NOT  0x800c0000u  /* negative empty class */
#define META_CLASS_END        0x800d0000u  /* end of non-empty class */
#define META_CLASS_NOT        0x800e0000u  /* start non-empty negative class */
#define META_COND_ASSERT      0x800f0000u  /* (?(?assertion)... */
#define META_COND_DEFINE      0x80100000u  /* (?(DEFINE)... */
#define META_COND_NAME        0x80110000u  /* (?(<name>)... */
#define META_COND_NUMBER      0x80120000u  /* (?(digits)... */
#define META_COND_RNAME       0x80130000u  /* (?(R&name)... */
#define META_COND_RNUMBER     0x80140000u  /* (?(Rdigits)... */
#define META_COND_VERSION     0x80150000u  /* (?(VERSION<op>x.y)... */
#define META_DOLLAR           0x80160000u  /* $ metacharacter */
#define META_DOT              0x80170000u  /* . metacharacter */
#define META_ESCAPE           0x80180000u  /* \d and friends */
#define META_KET              0x80190000u  /* closing parenthesis */
#define META_NOCAPTURE        0x801a0000u  /* no capture parens */
#define META_OPTIONS          0x801b0000u  /* (?i) and friends */
#define META_POSIX            0x801c0000u  /* POSIX class item */
#define META_POSIX_NEG        0x801d0000u  /* negative POSIX class item */
#define META_RANGE_ESCAPED    0x801e0000u  /* range with at least one escape */
#define META_RANGE_LITERAL    0x801f0000u  /* range defined literally */
#define META_RECURSE          0x80200000u  /* Recursion */
#define META_RECURSE_BYNAME   0x80210000u  /* (?&name) */
#define META_SCRIPT_RUN       0x80220000u  /* (*script_run:...) */

/* These must be kept together to make it easy to check that an assertion
is present where expected in a conditional group. */

#define META_LOOKAHEAD        0x80230000u  /* (?= */
#define META_LOOKAHEADNOT     0x80240000u  /* (?! */
#define META_LOOKBEHIND       0x80250000u  /* (?<= */
#define META_LOOKBEHINDNOT    0x80260000u  /* (?<! */

/* These must be kept in this order, with consecutive values, and the _ARG
versions of COMMIT, PRUNE, SKIP, and THEN immediately after their non-argument
versions. */

#define META_MARK             0x80270000u  /* (*MARK) */
#define META_ACCEPT           0x80280000u  /* (*ACCEPT) */
#define META_FAIL             0x80290000u  /* (*FAIL) */
#define META_COMMIT           0x802a0000u  /* These               */
#define META_COMMIT_ARG       0x802b0000u  /*   pairs             */
#define META_PRUNE            0x802c0000u  /*     must            */
#define META_PRUNE_ARG        0x802d0000u  /*       be            */
#define META_SKIP             0x802e0000u  /*         kept        */
#define META_SKIP_ARG         0x802f0000u  /*           in        */
#define META_THEN             0x80300000u  /*             this    */
#define META_THEN_ARG         0x80310000u  /*               order */

/* These must be kept in groups of adjacent 3 values, and all together. */

#define META_ASTERISK         0x80320000u  /* *  */
#define META_ASTERISK_PLUS    0x80330000u  /* *+ */
#define META_ASTERISK_QUERY   0x80340000u  /* *? */
#define META_PLUS             0x80350000u  /* +  */
#define META_PLUS_PLUS        0x80360000u  /* ++ */
#define META_PLUS_QUERY       0x80370000u  /* +? */
#define META_QUERY            0x80380000u  /* ?  */
#define META_QUERY_PLUS       0x80390000u  /* ?+ */
#define META_QUERY_QUERY      0x803a0000u  /* ?? */
#define META_MINMAX           0x803b0000u  /* {n,m}  repeat */
#define META_MINMAX_PLUS      0x803c0000u  /* {n,m}+ repeat */
#define META_MINMAX_QUERY     0x803d0000u  /* {n,m}? repeat */

#define META_FIRST_QUANTIFIER META_ASTERISK
#define META_LAST_QUANTIFIER  META_MINMAX_QUERY

/* This is a special "meta code" that is used only to distinguish (*asr: from
(*sr: in the table of aphabetic assertions. It is never stored in the parsed
pattern because (*asr: is turned into (*sr:(*atomic: at that stage. There is
therefore no need for it to have a length entry, so use a high value. */

#define META_ATOMIC_SCRIPT_RUN 0x8fff0000u

/* Table of extra lengths for each of the meta codes. Must be kept in step with
the definitions above. For some items these values are a basic length to which
a variable amount has to be added. */

static unsigned char meta_extra_lengths[] = {
  0,             /* META_END */
  0,             /* META_ALT */
  0,             /* META_ATOMIC */
  0,             /* META_BACKREF - more if group is >= 10 */
  1+SIZEOFFSET,  /* META_BACKREF_BYNAME */
  1,             /* META_BIGVALUE */
  3,             /* META_CALLOUT_NUMBER */
  3+SIZEOFFSET,  /* META_CALLOUT_STRING */
  0,             /* META_CAPTURE */
  0,             /* META_CIRCUMFLEX */
  0,             /* META_CLASS */
  0,             /* META_CLASS_EMPTY */
  0,             /* META_CLASS_EMPTY_NOT */
  0,             /* META_CLASS_END */
  0,             /* META_CLASS_NOT */
  0,             /* META_COND_ASSERT */
  SIZEOFFSET,    /* META_COND_DEFINE */
  1+SIZEOFFSET,  /* META_COND_NAME */
  1+SIZEOFFSET,  /* META_COND_NUMBER */
  1+SIZEOFFSET,  /* META_COND_RNAME */
  1+SIZEOFFSET,  /* META_COND_RNUMBER */
  3,             /* META_COND_VERSION */
  0,             /* META_DOLLAR */
  0,             /* META_DOT */
  0,             /* META_ESCAPE - more for ESC_P, ESC_p, ESC_g, ESC_k */
  0,             /* META_KET */
  0,             /* META_NOCAPTURE */
  1,             /* META_OPTIONS */
  1,             /* META_POSIX */
  1,             /* META_POSIX_NEG */
  0,             /* META_RANGE_ESCAPED */
  0,             /* META_RANGE_LITERAL */
  SIZEOFFSET,    /* META_RECURSE */
  1+SIZEOFFSET,  /* META_RECURSE_BYNAME */
  0,             /* META_SCRIPT_RUN */
  0,             /* META_LOOKAHEAD */
  0,             /* META_LOOKAHEADNOT */
  SIZEOFFSET,    /* META_LOOKBEHIND */
  SIZEOFFSET,    /* META_LOOKBEHINDNOT */
  1,             /* META_MARK - plus the string length */
  0,             /* META_ACCEPT */
  0,             /* META_FAIL */
  0,             /* META_COMMIT */
  1,             /* META_COMMIT_ARG - plus the string length */
  0,             /* META_PRUNE */
  1,             /* META_PRUNE_ARG - plus the string length */
  0,             /* META_SKIP */
  1,             /* META_SKIP_ARG - plus the string length */
  0,             /* META_THEN */
  1,             /* META_THEN_ARG - plus the string length */
  0,             /* META_ASTERISK */
  0,             /* META_ASTERISK_PLUS */
  0,             /* META_ASTERISK_QUERY */
  0,             /* META_PLUS */
  0,             /* META_PLUS_PLUS */
  0,             /* META_PLUS_QUERY */
  0,             /* META_QUERY */
  0,             /* META_QUERY_PLUS */
  0,             /* META_QUERY_QUERY */
  2,             /* META_MINMAX */
  2,             /* META_MINMAX_PLUS */
  2              /* META_MINMAX_QUERY */
};

/* Types for skipping parts of a parsed pattern. */

enum { PSKIP_ALT, PSKIP_CLASS, PSKIP_KET };

/* Macro for setting individual bits in class bitmaps. It took some
experimenting to figure out how to stop gcc 5.3.0 from warning with
-Wconversion. This version gets a warning:

  #define SETBIT(a,b) a[(b)/8] |= (uint8_t)(1u << ((b)&7))

Let's hope the apparently less efficient version isn't actually so bad if the
compiler is clever with identical subexpressions. */

#define SETBIT(a,b) a[(b)/8] = (uint8_t)(a[(b)/8] | (1u << ((b)&7)))

/* Private flags added to firstcu and reqcu. */

#define REQ_CASELESS    (1u << 0)       /* Indicates caselessness */
#define REQ_VARY        (1u << 1)       /* reqcu followed non-literal item */
/* Negative values for the firstcu and reqcu flags */
#define REQ_UNSET       (-2)            /* Not yet found anything */
#define REQ_NONE        (-1)            /* Found not fixed char */

/* These flags are used in the groupinfo vector. */

#define GI_SET_FIXED_LENGTH    0x80000000u
#define GI_NOT_FIXED_LENGTH    0x40000000u
#define GI_FIXED_LENGTH_MASK   0x0000ffffu

/* This simple test for a decimal digit works for both ASCII/Unicode and EBCDIC
and is fast (a good compiler can turn it into a subtraction and unsigned
comparison). */

#define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)

/* Table to identify hex digits. The tables in chartables are dependent on the
locale, and may mark arbitrary characters as digits. We want to recognize only
0-9, a-z, and A-Z as hex digits, which is why we have a private table here. It
costs 256 bytes, but it is a lot faster than doing character value tests (at
least in some simple cases I timed), and in some applications one wants PCRE2
to compile efficiently as well as match efficiently. The value in the table is
the binary hex digit value, or 0xff for non-hex digits. */

/* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
UTF-8 mode. */

#ifndef EBCDIC
static const uint8_t xdigitab[] =
  {
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*   0-  7 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*   8- 15 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  16- 23 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  24- 31 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*    - '  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  ( - /  */
  0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07, /*  0 - 7  */
  0x08,0x09,0xff,0xff,0xff,0xff,0xff,0xff, /*  8 - ?  */
  0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /*  @ - G  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  H - O  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  P - W  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  X - _  */
  0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /*  ` - g  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  h - o  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  p - w  */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  x -127 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 128-135 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 136-143 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 144-151 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 152-159 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 160-167 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 168-175 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 176-183 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 184-191 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 192-199 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 2ff-207 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 208-215 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 216-223 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 224-231 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 232-239 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 240-247 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff};/* 248-255 */

#else

/* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */

static const uint8_t xdigitab[] =
  {
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*   0-  7  0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*   8- 15    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  16- 23 10 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  24- 31    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  32- 39 20 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  40- 47    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  48- 55 30 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  56- 63    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*    - 71 40 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  72- |     */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  & - 87 50 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  88- 95    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  - -103 60 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 104- ?     */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 112-119 70 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 120- "     */
  0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /* 128- g  80 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  h -143    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 144- p  90 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  q -159    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 160- x  A0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  y -175    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  ^ -183 B0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 184-191    */
  0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /*  { - G  C0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  H -207    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  } - P  D0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  Q -223    */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  \ - X  E0 */
  0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /*  Y -239    */
  0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07, /*  0 - 7  F0 */
  0x08,0x09,0xff,0xff,0xff,0xff,0xff,0xff};/*  8 -255    */
#endif  /* EBCDIC */


/* Table for handling alphanumeric escaped characters. Positive returns are
simple data values; negative values are for special things like \d and so on.
Zero means further processing is needed (for things like \x), or the escape is
invalid. */

/* This is the "normal" table for ASCII systems or for EBCDIC systems running
in UTF-8 mode. It runs from '0' to 'z'. */

#ifndef EBCDIC
#define ESCAPES_FIRST       CHAR_0
#define ESCAPES_LAST        CHAR_z
#define UPPER_CASE(c)       (c-32)

static const short int escapes[] = {
     0,                       0,
     0,                       0,
     0,                       0,
     0,                       0,
     0,                       0,
     CHAR_COLON,              CHAR_SEMICOLON,
     CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
     CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
     CHAR_COMMERCIAL_AT,      -ESC_A,
     -ESC_B,                  -ESC_C,
     -ESC_D,                  -ESC_E,
     0,                       -ESC_G,
     -ESC_H,                  0,
     0,                       -ESC_K,
     0,                       0,
     -ESC_N,                  0,
     -ESC_P,                  -ESC_Q,
     -ESC_R,                  -ESC_S,
     0,                       0,
     -ESC_V,                  -ESC_W,
     -ESC_X,                  0,
     -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
     CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
     CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
     CHAR_GRAVE_ACCENT,       CHAR_BEL,
     -ESC_b,                  0,
     -ESC_d,                  CHAR_ESC,
     CHAR_FF,                 0,
     -ESC_h,                  0,
     0,                       -ESC_k,
     0,                       0,
     CHAR_LF,                 0,
     -ESC_p,                  0,
     CHAR_CR,                 -ESC_s,
     CHAR_HT,                 0,
     -ESC_v,                  -ESC_w,
     0,                       0,
     -ESC_z
};

#else

/* This is the "abnormal" table for EBCDIC systems without UTF-8 support.
It runs from 'a' to '9'. For some minimal testing of EBCDIC features, the code
is sometimes compiled on an ASCII system. In this case, we must not use CHAR_a
because it is defined as 'a', which of course picks up the ASCII value. */

#if 'a' == 0x81                    /* Check for a real EBCDIC environment */
#define ESCAPES_FIRST       CHAR_a
#define ESCAPES_LAST        CHAR_9
#define UPPER_CASE(c)       (c+64)
#else                              /* Testing in an ASCII environment */
#define ESCAPES_FIRST  ((unsigned char)'\x81')   /* EBCDIC 'a' */
#define ESCAPES_LAST   ((unsigned char)'\xf9')   /* EBCDIC '9' */
#define UPPER_CASE(c)  (c-32)
#endif

static const short int escapes[] = {
/*  80 */         CHAR_BEL, -ESC_b,       0, -ESC_d, CHAR_ESC, CHAR_FF,      0,
/*  88 */ -ESC_h,        0,      0,     '{',      0,        0,       0,      0,
/*  90 */      0,        0, -ESC_k,       0,      0,  CHAR_LF,       0, -ESC_p,
/*  98 */      0,  CHAR_CR,      0,     '}',      0,        0,       0,      0,
/*  A0 */      0,      '~', -ESC_s, CHAR_HT,      0,   -ESC_v,  -ESC_w,      0,
/*  A8 */      0,   -ESC_z,      0,       0,      0,      '[',       0,      0,
/*  B0 */      0,        0,      0,       0,      0,        0,       0,      0,
/*  B8 */      0,        0,      0,       0,      0,      ']',     '=',    '-',
/*  C0 */    '{',   -ESC_A, -ESC_B,  -ESC_C, -ESC_D,   -ESC_E,       0, -ESC_G,
/*  C8 */ -ESC_H,        0,      0,       0,      0,        0,       0,      0,
/*  D0 */    '}',        0, -ESC_K,       0,      0,   -ESC_N,       0, -ESC_P,
/*  D8 */ -ESC_Q,   -ESC_R,      0,       0,      0,        0,       0,      0,
/*  E0 */   '\\',        0, -ESC_S,       0,      0,   -ESC_V,  -ESC_W, -ESC_X,
/*  E8 */      0,   -ESC_Z,      0,       0,      0,        0,       0,      0,
/*  F0 */      0,        0,      0,       0,      0,        0,       0,      0,
/*  F8 */      0,        0
};

/* We also need a table of characters that may follow \c in an EBCDIC
environment for characters 0-31. */

static unsigned char ebcdic_escape_c[] = "@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_";

#endif   /* EBCDIC */


/* Table of special "verbs" like (*PRUNE). This is a short table, so it is
searched linearly. Put all the names into a single string, in order to reduce
the number of relocations when a shared library is dynamically linked. The
string is built from string macros so that it works in UTF-8 mode on EBCDIC
platforms. */

typedef struct verbitem {
  unsigned int len;          /* Length of verb name */
  uint32_t meta;             /* Base META_ code */
  int has_arg;               /* Argument requirement */
} verbitem;

static const char verbnames[] =
  "\0"                       /* Empty name is a shorthand for MARK */
  STRING_MARK0
  STRING_ACCEPT0
  STRING_F0
  STRING_FAIL0
  STRING_COMMIT0
  STRING_PRUNE0
  STRING_SKIP0
  STRING_THEN;

static const verbitem verbs[] = {
  { 0, META_MARK,   +1 },  /* > 0 => must have an argument */
  { 4, META_MARK,   +1 },
  { 6, META_ACCEPT, -1 },  /* < 0 => Optional argument, convert to pre-MARK */
  { 1, META_FAIL,   -1 },
  { 4, META_FAIL,   -1 },
  { 6, META_COMMIT,  0 },
  { 5, META_PRUNE,   0 },  /* Optional argument; bump META code if found */
  { 4, META_SKIP,    0 },
  { 4, META_THEN,    0 }
};

static const int verbcount = sizeof(verbs)/sizeof(verbitem);

/* Verb opcodes, indexed by their META code offset from META_MARK. */

static const uint32_t verbops[] = {
  OP_MARK, OP_ACCEPT, OP_FAIL, OP_COMMIT, OP_COMMIT_ARG, OP_PRUNE,
  OP_PRUNE_ARG, OP_SKIP, OP_SKIP_ARG, OP_THEN, OP_THEN_ARG };

/* Table of "alpha assertions" like (*pla:...), similar to the (*VERB) table. */

typedef struct alasitem {
  unsigned int len;          /* Length of name */
  uint32_t meta;             /* Base META_ code */
} alasitem;

static const char alasnames[] =
  STRING_pla0
  STRING_plb0
  STRING_nla0
  STRING_nlb0
  STRING_positive_lookahead0
  STRING_positive_lookbehind0
  STRING_negative_lookahead0
  STRING_negative_lookbehind0
  STRING_atomic0
  STRING_sr0
  STRING_asr0
  STRING_script_run0
  STRING_atomic_script_run;

static const alasitem alasmeta[] = {
  {  3, META_LOOKAHEAD         },
  {  3, META_LOOKBEHIND        },
  {  3, META_LOOKAHEADNOT      },
  {  3, META_LOOKBEHINDNOT     },
  { 18, META_LOOKAHEAD         },
  { 19, META_LOOKBEHIND        },
  { 18, META_LOOKAHEADNOT      },
  { 19, META_LOOKBEHINDNOT     },
  {  6, META_ATOMIC            },
  {  2, META_SCRIPT_RUN        }, /* sr = script run */
  {  3, META_ATOMIC_SCRIPT_RUN }, /* asr = atomic script run */
  { 10, META_SCRIPT_RUN        }, /* script run */
  { 17, META_ATOMIC_SCRIPT_RUN }  /* atomic script run */
};

static const int alascount = sizeof(alasmeta)/sizeof(alasitem);

/* Offsets from OP_STAR for case-independent and negative repeat opcodes. */

static uint32_t chartypeoffset[] = {
  OP_STAR - OP_STAR,    OP_STARI - OP_STAR,
  OP_NOTSTAR - OP_STAR, OP_NOTSTARI - OP_STAR };

/* Tables of names of POSIX character classes and their lengths. The names are
now all in a single string, to reduce the number of relocations when a shared
library is dynamically loaded. The list of lengths is terminated by a zero
length entry. The first three must be alpha, lower, upper, as this is assumed
for handling case independence. The indices for graph, print, and punct are
needed, so identify them. */

static const char posix_names[] =
  STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
  STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
  STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
  STRING_word0  STRING_xdigit;

static const uint8_t posix_name_lengths[] = {
  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };

#define PC_GRAPH  8
#define PC_PRINT  9
#define PC_PUNCT 10

/* Table of class bit maps for each POSIX class. Each class is formed from a
base map, with an optional addition or removal of another map. Then, for some
classes, there is some additional tweaking: for [:blank:] the vertical space
characters are removed, and for [:alpha:] and [:alnum:] the underscore
character is removed. The triples in the table consist of the base map offset,
second map offset or -1 if no second map, and a non-negative value for map
addition or a negative value for map subtraction (if there are two maps). The
absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
remove vertical space characters, 2 => remove underscore. */

static const int posix_class_maps[] = {
  cbit_word,  cbit_digit, -2,             /* alpha */
  cbit_lower, -1,          0,             /* lower */
  cbit_upper, -1,          0,             /* upper */
  cbit_word,  -1,          2,             /* alnum - word without underscore */
  cbit_print, cbit_cntrl,  0,             /* ascii */
  cbit_space, -1,          1,             /* blank - a GNU extension */
  cbit_cntrl, -1,          0,             /* cntrl */
  cbit_digit, -1,          0,             /* digit */
  cbit_graph, -1,          0,             /* graph */
  cbit_print, -1,          0,             /* print */
  cbit_punct, -1,          0,             /* punct */
  cbit_space, -1,          0,             /* space */
  cbit_word,  -1,          0,             /* word - a Perl extension */
  cbit_xdigit,-1,          0              /* xdigit */
};

#ifdef SUPPORT_UNICODE

/* The POSIX class Unicode property substitutes that are used in UCP mode must
be in the order of the POSIX class names, defined above. */

static int posix_substitutes[] = {
  PT_GC, ucp_L,     /* alpha */
  PT_PC, ucp_Ll,    /* lower */
  PT_PC, ucp_Lu,    /* upper */
  PT_ALNUM, 0,      /* alnum */
  -1, 0,            /* ascii, treat as non-UCP */
  -1, 1,            /* blank, treat as \h */
  PT_PC, ucp_Cc,    /* cntrl */
  PT_PC, ucp_Nd,    /* digit */
  PT_PXGRAPH, 0,    /* graph */
  PT_PXPRINT, 0,    /* print */
  PT_PXPUNCT, 0,    /* punct */
  PT_PXSPACE, 0,    /* space */   /* Xps is POSIX space, but from 8.34 */
  PT_WORD, 0,       /* word  */   /* Perl and POSIX space are the same */
  -1, 0             /* xdigit, treat as non-UCP */
};
#define POSIX_SUBSIZE (sizeof(posix_substitutes) / (2*sizeof(uint32_t)))
#endif  /* SUPPORT_UNICODE */

/* Masks for checking option settings. When PCRE2_LITERAL is set, only a subset
are allowed. */

#define PUBLIC_LITERAL_COMPILE_OPTIONS \
  (PCRE2_ANCHORED|PCRE2_AUTO_CALLOUT|PCRE2_CASELESS|PCRE2_ENDANCHORED| \
   PCRE2_FIRSTLINE|PCRE2_LITERAL|PCRE2_NO_START_OPTIMIZE| \
   PCRE2_NO_UTF_CHECK|PCRE2_USE_OFFSET_LIMIT|PCRE2_UTF)

#define PUBLIC_COMPILE_OPTIONS \
  (PUBLIC_LITERAL_COMPILE_OPTIONS| \
   PCRE2_ALLOW_EMPTY_CLASS|PCRE2_ALT_BSUX|PCRE2_ALT_CIRCUMFLEX| \
   PCRE2_ALT_VERBNAMES|PCRE2_DOLLAR_ENDONLY|PCRE2_DOTALL|PCRE2_DUPNAMES| \
   PCRE2_EXTENDED|PCRE2_EXTENDED_MORE|PCRE2_MATCH_UNSET_BACKREF| \
   PCRE2_MULTILINE|PCRE2_NEVER_BACKSLASH_C|PCRE2_NEVER_UCP| \
   PCRE2_NEVER_UTF|PCRE2_NO_AUTO_CAPTURE|PCRE2_NO_AUTO_POSSESS| \
   PCRE2_NO_DOTSTAR_ANCHOR|PCRE2_UCP|PCRE2_UNGREEDY)

#define PUBLIC_LITERAL_COMPILE_EXTRA_OPTIONS \
   (PCRE2_EXTRA_MATCH_LINE|PCRE2_EXTRA_MATCH_WORD)

#define PUBLIC_COMPILE_EXTRA_OPTIONS \
   (PUBLIC_LITERAL_COMPILE_EXTRA_OPTIONS| \
    PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES|PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL| \
    PCRE2_EXTRA_ESCAPED_CR_IS_LF|PCRE2_EXTRA_ALT_BSUX)

/* Compile time error code numbers. They are given names so that they can more
easily be tracked. When a new number is added, the tables called eint1 and
eint2 in pcre2posix.c may need to be updated, and a new error text must be
added to compile_error_texts in pcre2_error.c. */

enum { ERR0 = COMPILE_ERROR_BASE,
       ERR1,  ERR2,  ERR3,  ERR4,  ERR5,  ERR6,  ERR7,  ERR8,  ERR9,  ERR10,
       ERR11, ERR12, ERR13, ERR14, ERR15, ERR16, ERR17, ERR18, ERR19, ERR20,
       ERR21, ERR22, ERR23, ERR24, ERR25, ERR26, ERR27, ERR28, ERR29, ERR30,
       ERR31, ERR32, ERR33, ERR34, ERR35, ERR36, ERR37, ERR38, ERR39, ERR40,
       ERR41, ERR42, ERR43, ERR44, ERR45, ERR46, ERR47, ERR48, ERR49, ERR50,
       ERR51, ERR52, ERR53, ERR54, ERR55, ERR56, ERR57, ERR58, ERR59, ERR60,
       ERR61, ERR62, ERR63, ERR64, ERR65, ERR66, ERR67, ERR68, ERR69, ERR70,
       ERR71, ERR72, ERR73, ERR74, ERR75, ERR76, ERR77, ERR78, ERR79, ERR80,
       ERR81, ERR82, ERR83, ERR84, ERR85, ERR86, ERR87, ERR88, ERR89, ERR90,
       ERR91, ERR92, ERR93, ERR94, ERR95, ERR96 };

/* This is a table of start-of-pattern options such as (*UTF) and settings such
as (*LIMIT_MATCH=nnnn) and (*CRLF). For completeness and backward
compatibility, (*UTFn) is supported in the relevant libraries, but (*UTF) is
generic and always supported. */

enum { PSO_OPT,     /* Value is an option bit */
       PSO_FLG,     /* Value is a flag bit */
       PSO_NL,      /* Value is a newline type */
       PSO_BSR,     /* Value is a \R type */
       PSO_LIMH,    /* Read integer value for heap limit */
       PSO_LIMM,    /* Read integer value for match limit */
       PSO_LIMD };  /* Read integer value for depth limit */

typedef struct pso {
  const uint8_t *name;
  uint16_t length;
  uint16_t type;
  uint32_t value;
} pso;

/* NB: STRING_UTFn_RIGHTPAR contains the length as well */

static pso pso_list[] = {
  { (uint8_t *)STRING_UTFn_RIGHTPAR,                  PSO_OPT, PCRE2_UTF },
  { (uint8_t *)STRING_UTF_RIGHTPAR,                4, PSO_OPT, PCRE2_UTF },
  { (uint8_t *)STRING_UCP_RIGHTPAR,                4, PSO_OPT, PCRE2_UCP },
  { (uint8_t *)STRING_NOTEMPTY_RIGHTPAR,           9, PSO_FLG, PCRE2_NOTEMPTY_SET },
  { (uint8_t *)STRING_NOTEMPTY_ATSTART_RIGHTPAR,  17, PSO_FLG, PCRE2_NE_ATST_SET },
  { (uint8_t *)STRING_NO_AUTO_POSSESS_RIGHTPAR,   16, PSO_OPT, PCRE2_NO_AUTO_POSSESS },
  { (uint8_t *)STRING_NO_DOTSTAR_ANCHOR_RIGHTPAR, 18, PSO_OPT, PCRE2_NO_DOTSTAR_ANCHOR },
  { (uint8_t *)STRING_NO_JIT_RIGHTPAR,             7, PSO_FLG, PCRE2_NOJIT },
  { (uint8_t *)STRING_NO_START_OPT_RIGHTPAR,      13, PSO_OPT, PCRE2_NO_START_OPTIMIZE },
  { (uint8_t *)STRING_LIMIT_HEAP_EQ,              11, PSO_LIMH, 0 },
  { (uint8_t *)STRING_LIMIT_MATCH_EQ,             12, PSO_LIMM, 0 },
  { (uint8_t *)STRING_LIMIT_DEPTH_EQ,             12, PSO_LIMD, 0 },
  { (uint8_t *)STRING_LIMIT_RECURSION_EQ,         16, PSO_LIMD, 0 },
  { (uint8_t *)STRING_CR_RIGHTPAR,                 3, PSO_NL,  PCRE2_NEWLINE_CR },
  { (uint8_t *)STRING_LF_RIGHTPAR,                 3, PSO_NL,  PCRE2_NEWLINE_LF },
  { (uint8_t *)STRING_CRLF_RIGHTPAR,               5, PSO_NL,  PCRE2_NEWLINE_CRLF },
  { (uint8_t *)STRING_ANY_RIGHTPAR,                4, PSO_NL,  PCRE2_NEWLINE_ANY },
  { (uint8_t *)STRING_NUL_RIGHTPAR,                4, PSO_NL,  PCRE2_NEWLINE_NUL },
  { (uint8_t *)STRING_ANYCRLF_RIGHTPAR,            8, PSO_NL,  PCRE2_NEWLINE_ANYCRLF },
  { (uint8_t *)STRING_BSR_ANYCRLF_RIGHTPAR,       12, PSO_BSR, PCRE2_BSR_ANYCRLF },
  { (uint8_t *)STRING_BSR_UNICODE_RIGHTPAR,       12, PSO_BSR, PCRE2_BSR_UNICODE }
};

/* This table is used when converting repeating opcodes into possessified
versions as a result of an explicit possessive quantifier such as ++. A zero
value means there is no possessified version - in those cases the item in
question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
because all relevant opcodes are less than that. */

static const uint8_t opcode_possessify[] = {
  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */

  0,                       /* NOTI */
  OP_POSSTAR, 0,           /* STAR, MINSTAR */
  OP_POSPLUS, 0,           /* PLUS, MINPLUS */
  OP_POSQUERY, 0,          /* QUERY, MINQUERY */
  OP_POSUPTO, 0,           /* UPTO, MINUPTO */
  0,                       /* EXACT */
  0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */

  OP_POSSTARI, 0,          /* STARI, MINSTARI */
  OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
  OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
  OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
  0,                       /* EXACTI */
  0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */

  OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
  OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
  OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
  OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
  0,                       /* NOTEXACT */
  0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */

  OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
  OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
  OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
  OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
  0,                       /* NOTEXACTI */
  0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */

  OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
  OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
  OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
  OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
  0,                       /* TYPEEXACT */
  0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */

  OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
  OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
  OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
  OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
  0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */

  0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
  0, 0,                    /* REF, REFI */
  0, 0,                    /* DNREF, DNREFI */
  0, 0                     /* RECURSE, CALLOUT */
};


#ifdef DEBUG_SHOW_PARSED
/*************************************************
*     Show the parsed pattern for debugging      *
*************************************************/

/* For debugging the pre-scan, this code, which outputs the parsed data vector,
can be enabled. */

static void show_parsed(compile_block *cb)
{
uint32_t *pptr = cb->parsed_pattern;

for (;;)
  {
  int max, min;
  PCRE2_SIZE offset;
  uint32_t i;
  uint32_t length;
  uint32_t meta_arg = META_DATA(*pptr);

  fprintf(stderr, "+++ %02d %.8x ", (int)(pptr - cb->parsed_pattern), *pptr);

  if (*pptr < META_END)
    {
    if (*pptr > 32 && *pptr < 128) fprintf(stderr, "%c", *pptr);
    pptr++;
    }

  else switch (META_CODE(*pptr++))
    {
    default:
    fprintf(stderr, "**** OOPS - unknown META value - giving up ****\n");
    return;

    case META_END:
    fprintf(stderr, "META_END\n");
    return;

    case META_CAPTURE:
    fprintf(stderr, "META_CAPTURE %d", meta_arg);
    break;

    case META_RECURSE:
    GETOFFSET(offset, pptr);
    fprintf(stderr, "META_RECURSE %d %zd", meta_arg, offset);
    break;

    case META_BACKREF:
    if (meta_arg < 10)
      offset = cb->small_ref_offset[meta_arg];
    else
      GETOFFSET(offset, pptr);
    fprintf(stderr, "META_BACKREF %d %zd", meta_arg, offset);
    break;

    case META_ESCAPE:
    if (meta_arg == ESC_P || meta_arg == ESC_p)
      {
      uint32_t ptype = *pptr >> 16;
      uint32_t pvalue = *pptr++ & 0xffff;
      fprintf(stderr, "META \\%c %d %d", (meta_arg == ESC_P)? 'P':'p',
        ptype, pvalue);
      }
    else
      {
      uint32_t cc;
      /* There's just one escape we might have here that isn't negated in the
      escapes table. */
      if (meta_arg == ESC_g) cc = CHAR_g;
      else for (cc = ESCAPES_FIRST; cc <= ESCAPES_LAST; cc++)
        {
        if (meta_arg == (uint32_t)(-escapes[cc - ESCAPES_FIRST])) break;
        }
      if (cc > ESCAPES_LAST) cc = CHAR_QUESTION_MARK;
      fprintf(stderr, "META \\%c", cc);
      }
    break;

    case META_MINMAX:
    min = *pptr++;
    max = *pptr++;
    if (max != REPEAT_UNLIMITED)
      fprintf(stderr, "META {%d,%d}", min, max);
    else
      fprintf(stderr, "META {%d,}", min);
    break;

    case META_MINMAX_QUERY:
    min = *pptr++;
    max = *pptr++;
    if (max != REPEAT_UNLIMITED)
      fprintf(stderr, "META {%d,%d}?", min, max);
    else
      fprintf(stderr, "META {%d,}?", min);
    break;

    case META_MINMAX_PLUS:
    min = *pptr++;
    max = *pptr++;
    if (max != REPEAT_UNLIMITED)
      fprintf(stderr, "META {%d,%d}+", min, max);
    else
      fprintf(stderr, "META {%d,}+", min);
    break;

    case META_BIGVALUE: fprintf(stderr, "META_BIGVALUE %.8x", *pptr++); break;
    case META_CIRCUMFLEX: fprintf(stderr, "META_CIRCUMFLEX"); break;
    case META_COND_ASSERT: fprintf(stderr, "META_COND_ASSERT"); break;
    case META_DOLLAR: fprintf(stderr, "META_DOLLAR"); break;
    case META_DOT: fprintf(stderr, "META_DOT"); break;
    case META_ASTERISK: fprintf(stderr, "META *"); break;
    case META_ASTERISK_QUERY: fprintf(stderr, "META *?"); break;
    case META_ASTERISK_PLUS: fprintf(stderr, "META *+"); break;
    case META_PLUS: fprintf(stderr, "META +"); break;
    case META_PLUS_QUERY: fprintf(stderr, "META +?"); break;
    case META_PLUS_PLUS: fprintf(stderr, "META ++"); break;
    case META_QUERY: fprintf(stderr, "META ?"); break;
    case META_QUERY_QUERY: fprintf(stderr, "META ??"); break;
    case META_QUERY_PLUS: fprintf(stderr, "META ?+"); break;

    case META_ATOMIC: fprintf(stderr, "META (?>"); break;
    case META_NOCAPTURE: fprintf(stderr, "META (?:"); break;
    case META_LOOKAHEAD: fprintf(stderr, "META (?="); break;
    case META_LOOKAHEADNOT: fprintf(stderr, "META (?!"); break;
    case META_SCRIPT_RUN: fprintf(stderr, "META (*sr:"); break;
    case META_KET: fprintf(stderr, "META )"); break;
    case META_ALT: fprintf(stderr, "META | %d", meta_arg); break;

    case META_CLASS: fprintf(stderr, "META ["); break;
    case META_CLASS_NOT: fprintf(stderr, "META [^"); break;
    case META_CLASS_END: fprintf(stderr, "META ]"); break;
    case META_CLASS_EMPTY: fprintf(stderr, "META []"); break;
    case META_CLASS_EMPTY_NOT: fprintf(stderr, "META [^]"); break;

    case META_RANGE_LITERAL: fprintf(stderr, "META - (literal)"); break;
    case META_RANGE_ESCAPED: fprintf(stderr, "META - (escaped)"); break;

    case META_POSIX: fprintf(stderr, "META_POSIX %d", *pptr++); break;
    case META_POSIX_NEG: fprintf(stderr, "META_POSIX_NEG %d", *pptr++); break;

    case META_ACCEPT: fprintf(stderr, "META (*ACCEPT)"); break;
    case META_FAIL: fprintf(stderr, "META (*FAIL)"); break;
    case META_COMMIT: fprintf(stderr, "META (*COMMIT)"); break;
    case META_PRUNE: fprintf(stderr, "META (*PRUNE)"); break;
    case META_SKIP: fprintf(stderr, "META (*SKIP)"); break;
    case META_THEN: fprintf(stderr, "META (*THEN)"); break;

    case META_OPTIONS: fprintf(stderr, "META_OPTIONS 0x%02x", *pptr++); break;

    case META_LOOKBEHIND:
    fprintf(stderr, "META (?<= %d offset=", meta_arg);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_LOOKBEHINDNOT:
    fprintf(stderr, "META (?<! %d offset=", meta_arg);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_CALLOUT_NUMBER:
    fprintf(stderr, "META (?C%d) next=%d/%d", pptr[2], pptr[0],
       pptr[1]);
    pptr += 3;
    break;

    case META_CALLOUT_STRING:
      {
      uint32_t patoffset = *pptr++;    /* Offset of next pattern item */
      uint32_t patlength = *pptr++;    /* Length of next pattern item */
      fprintf(stderr, "META (?Cstring) length=%d offset=", *pptr++);
      GETOFFSET(offset, pptr);
      fprintf(stderr, "%zd next=%d/%d", offset, patoffset, patlength);
      }
    break;

    case META_RECURSE_BYNAME:
    fprintf(stderr, "META (?(&name) length=%d offset=", *pptr++);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_BACKREF_BYNAME:
    fprintf(stderr, "META_BACKREF_BYNAME length=%d offset=", *pptr++);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_COND_NUMBER:
    fprintf(stderr, "META_COND_NUMBER %d offset=", pptr[SIZEOFFSET]);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    pptr++;
    break;

    case META_COND_DEFINE:
    fprintf(stderr, "META (?(DEFINE) offset=");
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_COND_VERSION:
    fprintf(stderr, "META (?(VERSION%s", (*pptr++ == 0)? "=" : ">=");
    fprintf(stderr, "%d.", *pptr++);
    fprintf(stderr, "%d)", *pptr++);
    break;

    case META_COND_NAME:
    fprintf(stderr, "META (?(<name>) length=%d offset=", *pptr++);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_COND_RNAME:
    fprintf(stderr, "META (?(R&name) length=%d offset=", *pptr++);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    /* This is kept as a name, because it might be. */

    case META_COND_RNUMBER:
    fprintf(stderr, "META (?(Rnumber) length=%d offset=", *pptr++);
    GETOFFSET(offset, pptr);
    fprintf(stderr, "%zd", offset);
    break;

    case META_MARK:
    fprintf(stderr, "META (*MARK:");
    goto SHOWARG;

    case META_COMMIT_ARG:
    fprintf(stderr, "META (*COMMIT:");
    goto SHOWARG;

    case META_PRUNE_ARG:
    fprintf(stderr, "META (*PRUNE:");
    goto SHOWARG;

    case META_SKIP_ARG:
    fprintf(stderr, "META (*SKIP:");
    goto SHOWARG;

    case META_THEN_ARG:
    fprintf(stderr, "META (*THEN:");
    SHOWARG:
    length = *pptr++;
    for (i = 0; i < length; i++)
      {
      uint32_t cc = *pptr++;
      if (cc > 32 && cc < 128) fprintf(stderr, "%c", cc);
        else fprintf(stderr, "\\x{%x}", cc);
      }
    fprintf(stderr, ") length=%u", length);
    break;
    }
  fprintf(stderr, "\n");
  }
return;
}
#endif  /* DEBUG_SHOW_PARSED */



/*************************************************
*               Copy compiled code               *
*************************************************/

/* Compiled JIT code cannot be copied, so the new compiled block has no
associated JIT data. */

PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION
pcre2_code_copy(const pcre2_code *code)
{
PCRE2_SIZE* ref_count;
pcre2_code *newcode;

if (code == NULL) return NULL;
newcode = code->memctl.malloc(code->blocksize, code->memctl.memory_data);
if (newcode == NULL) return NULL;
memcpy(newcode, code, code->blocksize);
newcode->executable_jit = NULL;

/* If the code is one that has been deserialized, increment the reference count
in the decoded tables. */

if ((code->flags & PCRE2_DEREF_TABLES) != 0)
  {
  ref_count = (PCRE2_SIZE *)(code->tables + tables_length);
  (*ref_count)++;
  }

return newcode;
}



/*************************************************
*     Copy compiled code and character tables    *
*************************************************/

/* Compiled JIT code cannot be copied, so the new compiled block has no
associated JIT data. This version of code_copy also makes a separate copy of
the character tables. */

PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION
pcre2_code_copy_with_tables(const pcre2_code *code)
{
PCRE2_SIZE* ref_count;
pcre2_code *newcode;
uint8_t *newtables;

if (code == NULL) return NULL;
newcode = code->memctl.malloc(code->blocksize, code->memctl.memory_data);
if (newcode == NULL) return NULL;
memcpy(newcode, code, code->blocksize);
newcode->executable_jit = NULL;

newtables = code->memctl.malloc(tables_length + sizeof(PCRE2_SIZE),
  code->memctl.memory_data);
if (newtables == NULL)
  {
  code->memctl.free((void *)newcode, code->memctl.memory_data);
  return NULL;
  }
memcpy(newtables, code->tables, tables_length);
ref_count = (PCRE2_SIZE *)(newtables + tables_length);
*ref_count = 1;

newcode->tables = newtables;
newcode->flags |= PCRE2_DEREF_TABLES;
return newcode;
}



/*************************************************
*               Free compiled code               *
*************************************************/

PCRE2_EXP_DEFN void PCRE2_CALL_CONVENTION
pcre2_code_free(pcre2_code *code)
{
PCRE2_SIZE* ref_count;

if (code != NULL)
  {
  if (code->executable_jit != NULL)
    PRIV(jit_free)(code->executable_jit, &code->memctl);

  if ((code->flags & PCRE2_DEREF_TABLES) != 0)
    {
    /* Decoded tables belong to the codes after deserialization, and they must
    be freed when there are no more references to them. The *ref_count should
    always be > 0. */

    ref_count = (PCRE2_SIZE *)(code->tables + tables_length);
    if (*ref_count > 0)
      {
      (*ref_count)--;
      if (*ref_count == 0)
        code->memctl.free((void *)code->tables, code->memctl.memory_data);
      }
    }

  code->memctl.free(code, code->memctl.memory_data);
  }
}



/*************************************************
*         Read a number, possibly signed         *
*************************************************/

/* This function is used to read numbers in the pattern. The initial pointer
must be the sign or first digit of the number. When relative values (introduced
by + or -) are allowed, they are relative group numbers, and the result must be
greater than zero.

Arguments:
  ptrptr      points to the character pointer variable
  ptrend      points to the end of the input string
  allow_sign  if < 0, sign not allowed; if >= 0, sign is relative to this
  max_value   the largest number allowed
  max_error   the error to give for an over-large number
  intptr      where to put the result
  errcodeptr  where to put an error code

Returns:      TRUE  - a number was read
              FALSE - errorcode == 0 => no number was found
                      errorcode != 0 => an error occurred
*/

static BOOL
read_number(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, int32_t allow_sign,
  uint32_t max_value, uint32_t max_error, int *intptr, int *errorcodeptr)
{
int sign = 0;
uint32_t n = 0;
PCRE2_SPTR ptr = *ptrptr;
BOOL yield = FALSE;

*errorcodeptr = 0;

if (allow_sign >= 0 && ptr < ptrend)
  {
  if (*ptr == CHAR_PLUS)
    {
    sign = +1;
    max_value -= allow_sign;
    ptr++;
    }
  else if (*ptr == CHAR_MINUS)
    {
    sign = -1;
    ptr++;
    }
  }

if (ptr >= ptrend || !IS_DIGIT(*ptr)) return FALSE;
while (ptr < ptrend && IS_DIGIT(*ptr))
  {
  n = n * 10 + *ptr++ - CHAR_0;
  if (n > max_value)
    {
    *errorcodeptr = max_error;
    goto EXIT;
    }
  }

if (allow_sign >= 0 && sign != 0)
  {
  if (n == 0)
    {
    *errorcodeptr = ERR26;  /* +0 and -0 are not allowed */
    goto EXIT;
    }

  if (sign > 0) n += allow_sign;
  else if ((int)n > allow_sign)
    {
    *errorcodeptr = ERR15;  /* Non-existent subpattern */
    goto EXIT;
    }
  else n = allow_sign + 1 - n;
  }

yield = TRUE;

EXIT:
*intptr = n;
*ptrptr = ptr;
return yield;
}



/*************************************************
*         Read repeat counts                     *
*************************************************/

/* Read an item of the form {n,m} and return the values if non-NULL pointers
are supplied. Repeat counts must be less than 65536 (MAX_REPEAT_COUNT); a
larger value is used for "unlimited". We have to use signed arguments for
read_number() because it is capable of returning a signed value.

Arguments:
  ptrptr         points to pointer to character after'{'
  ptrend         pointer to end of input
  minp           if not NULL, pointer to int for min
  maxp           if not NULL, pointer to int for max (-1 if no max)
                 returned as -1 if no max
  errorcodeptr   points to error code variable

Returns:         FALSE if not a repeat quantifier, errorcode set zero
                 FALSE on error, with errorcode set non-zero
                 TRUE on success, with pointer updated to point after '}'
*/

static BOOL
read_repeat_counts(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, uint32_t *minp,
  uint32_t *maxp, int *errorcodeptr)
{
PCRE2_SPTR p = *ptrptr;
BOOL yield = FALSE;
int32_t min = 0;
int32_t max = REPEAT_UNLIMITED; /* This value is larger than MAX_REPEAT_COUNT */

/* NB read_number() initializes the error code to zero. The only error is for a
number that is too big. */

if (!read_number(&p, ptrend, -1, MAX_REPEAT_COUNT, ERR5, &min, errorcodeptr))
  goto EXIT;

if (p >= ptrend) goto EXIT;

if (*p == CHAR_RIGHT_CURLY_BRACKET)
  {
  p++;
  max = min;
  }

else
  {
  if (*p++ != CHAR_COMMA || p >= ptrend) goto EXIT;
  if (*p != CHAR_RIGHT_CURLY_BRACKET)
    {
    if (!read_number(&p, ptrend, -1, MAX_REPEAT_COUNT, ERR5, &max,
        errorcodeptr) || p >= ptrend ||  *p != CHAR_RIGHT_CURLY_BRACKET)
      goto EXIT;
    if (max < min)
      {
      *errorcodeptr = ERR4;
      goto EXIT;
      }
    }
  p++;
  }

yield = TRUE;
if (minp != NULL) *minp = (uint32_t)min;
if (maxp != NULL) *maxp = (uint32_t)max;

/* Update the pattern pointer on success, or after an error, but not when
the result is "not a repeat quantifier". */

EXIT:
if (yield || *errorcodeptr != 0) *ptrptr = p;
return yield;



}



/*************************************************
*            Handle escapes                      *
*************************************************/

/* This function is called when a \ has been encountered. It either returns a
positive value for a simple escape such as \d, or 0 for a data character, which
is placed in chptr. A backreference to group n is returned as negative n. On
entry, ptr is pointing at the character after \. On exit, it points after the
final code unit of the escape sequence.

This function is also called from pcre2_substitute() to handle escape sequences
in replacement strings. In this case, the cb argument is NULL, and in the case
of escapes that have further processing, only sequences that define a data
character are recognised. The isclass argument is not relevant; the options
argument is the final value of the compiled pattern's options.

Arguments:
  ptrptr         points to the input position pointer
  ptrend         points to the end of the input
  chptr          points to a returned data character
  errorcodeptr   points to the errorcode variable (containing zero)
  options        the current options bits
  isclass        TRUE if inside a character class
  cb             compile data block or NULL when called from pcre2_substitute()

Returns:         zero => a data character
                 positive => a special escape sequence
                 negative => a numerical back reference
                 on error, errorcodeptr is set non-zero
*/

int
PRIV(check_escape)(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, uint32_t *chptr,
  int *errorcodeptr, uint32_t options, uint32_t extra_options, BOOL isclass,
  compile_block *cb)
{
BOOL utf = (options & PCRE2_UTF) != 0;
PCRE2_SPTR ptr = *ptrptr;
uint32_t c, cc;
int escape = 0;
int i;

/* If backslash is at the end of the string, it's an error. */

if (ptr >= ptrend)
  {
  *errorcodeptr = ERR1;
  return 0;
  }

GETCHARINCTEST(c, ptr);         /* Get character value, increment pointer */
*errorcodeptr = 0;              /* Be optimistic */

/* Non-alphanumerics are literals, so we just leave the value in c. An initial
value test saves a memory lookup for code points outside the alphanumeric
range. */

if (c < ESCAPES_FIRST || c > ESCAPES_LAST) {}  /* Definitely literal */

/* Otherwise, do a table lookup. Non-zero values need little processing here. A
positive value is a literal value for something like \n. A negative value is
the negation of one of the ESC_ macros that is passed back for handling by the
calling function. Some extra checking is needed for \N because only \N{U+dddd}
is supported. If the value is zero, further processing is handled below. */

else if ((i = escapes[c - ESCAPES_FIRST]) != 0)
  {
  if (i > 0)
    {
    c = (uint32_t)i;
    if (c == CHAR_CR && (extra_options & PCRE2_EXTRA_ESCAPED_CR_IS_LF) != 0)
      c = CHAR_LF;
    }
  else  /* Negative table entry */
    {
    escape = -i;                    /* Else return a special escape */
    if (cb != NULL && (escape == ESC_P || escape == ESC_p || escape == ESC_X))
      cb->external_flags |= PCRE2_HASBKPORX;   /* Note \P, \p, or \X */

    /* Perl supports \N{name} for character names and \N{U+dddd} for numerical
    Unicode code points, as well as plain \N for "not newline". PCRE does not
    support \N{name}. However, it does support quantification such as \N{2,3},
    so if \N{ is not followed by U+dddd we check for a quantifier. */

    if (escape == ESC_N && ptr < ptrend && *ptr == CHAR_LEFT_CURLY_BRACKET)
      {
      PCRE2_SPTR p = ptr + 1;

      /* \N{U+ can be handled by the \x{ code. However, this construction is
      not valid in EBCDIC environments because it specifies a Unicode
      character, not a codepoint in the local code. For example \N{U+0041}
      must be "A" in all environments. Also, in Perl, \N{U+ forces Unicode
      casing semantics for the entire pattern, so allow it only in UTF (i.e.
      Unicode) mode. */

      if (ptrend - p > 1 && *p == CHAR_U && p[1] == CHAR_PLUS)
        {
#ifdef EBCDIC
        *errorcodeptr = ERR93;
#else
        if (utf)
          {
          ptr = p + 1;
          escape = 0;   /* Not a fancy escape after all */
          goto COME_FROM_NU;
          }
        else *errorcodeptr = ERR93;
#endif
        }

      /* Give an error if what follows is not a quantifier, but don't override
      an error set by the quantifier reader (e.g. number overflow). */

      else
        {
        if (!read_repeat_counts(&p, ptrend, NULL, NULL, errorcodeptr) &&
             *errorcodeptr == 0)
          *errorcodeptr = ERR37;
        }
      }
    }
  }

/* Escapes that need further processing, including those that are unknown, have
a zero entry in the lookup table. When called from pcre2_substitute(), only \c,
\o, and \x are recognized (\u and \U can never appear as they are used for case
forcing). */

else
  {
  int s;
  PCRE2_SPTR oldptr;
  BOOL overflow;
  BOOL alt_bsux =
    ((options & PCRE2_ALT_BSUX) | (extra_options & PCRE2_EXTRA_ALT_BSUX)) != 0;

  /* Filter calls from pcre2_substitute(). */

  if (cb == NULL)
    {
    if (c != CHAR_c && c != CHAR_o && c != CHAR_x)
      {
      *errorcodeptr = ERR3;
      return 0;
      }
    alt_bsux = FALSE;   /* Do not modify \x handling */
    }

  switch (c)
    {
    /* A number of Perl escapes are not handled by PCRE. We give an explicit
    error. */

    case CHAR_F:
    case CHAR_l:
    case CHAR_L:
    *errorcodeptr = ERR37;
    break;

    /* \u is unrecognized when neither PCRE2_ALT_BSUX nor PCRE2_EXTRA_ALT_BSUX
    is set. Otherwise, \u must be followed by exactly four hex digits or, if
    PCRE2_EXTRA_ALT_BSUX is set, by any number of hex digits in braces.
    Otherwise it is a lowercase u letter. This gives some compatibility with
    ECMAScript (aka JavaScript). */

    case CHAR_u:
    if (!alt_bsux) *errorcodeptr = ERR37; else
      {
      uint32_t xc;

      if (ptr >= ptrend) break;
      if (*ptr == CHAR_LEFT_CURLY_BRACKET &&
          (extra_options & PCRE2_EXTRA_ALT_BSUX) != 0)
        {
        PCRE2_SPTR hptr = ptr + 1;
        cc = 0;

        while (hptr < ptrend && (xc = XDIGIT(*hptr)) != 0xff)
          {
          if ((cc & 0xf0000000) != 0)  /* Test for 32-bit overflow */
            {
            *errorcodeptr = ERR77;
            ptr = hptr;   /* Show where */
            break;        /* *hptr != } will cause another break below */
            }
          cc = (cc << 4) | xc;
          hptr++;
          }

        if (hptr == ptr + 1 ||   /* No hex digits */
            hptr >= ptrend ||    /* Hit end of input */
            *hptr != CHAR_RIGHT_CURLY_BRACKET)  /* No } terminator */
          break;         /* Hex escape not recognized */

        c = cc;          /* Accept the code point */
        ptr = hptr + 1;
        }

      else  /* Must be exactly 4 hex digits */
        {
        if (ptrend - ptr < 4) break;               /* Less than 4 chars */
        if ((cc = XDIGIT(ptr[0])) == 0xff) break;  /* Not a hex digit */
        if ((xc = XDIGIT(ptr[1])) == 0xff) break;  /* Not a hex digit */
        cc = (cc << 4) | xc;
        if ((xc = XDIGIT(ptr[2])) == 0xff) break;  /* Not a hex digit */
        cc = (cc << 4) | xc;
        if ((xc = XDIGIT(ptr[3])) == 0xff) break;  /* Not a hex digit */
        c = (cc << 4) | xc;
        ptr += 4;
        }

      if (utf)
        {
        if (c > 0x10ffffU) *errorcodeptr = ERR77;
        else
          if (c >= 0xd800 && c <= 0xdfff &&
              (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0)
                *errorcodeptr = ERR73;
        }
      else if (c > MAX_NON_UTF_CHAR) *errorcodeptr = ERR77;
      }
    break;

    /* \U is unrecognized unless PCRE2_ALT_BSUX or PCRE2_EXTRA_ALT_BSUX is set,
    in which case it is an upper case letter. */

    case CHAR_U:
    if (!alt_bsux) *errorcodeptr = ERR37;
    break;

    /* In a character class, \g is just a literal "g". Outside a character
    class, \g must be followed by one of a number of specific things:

    (1) A number, either plain or braced. If positive, it is an absolute
    backreference. If negative, it is a relative backreference. This is a Perl
    5.10 feature.

    (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
    is part of Perl's movement towards a unified syntax for back references. As
    this is synonymous with \k{name}, we fudge it up by pretending it really
    was \k{name}.

    (3) For Oniguruma compatibility we also support \g followed by a name or a
    number either in angle brackets or in single quotes. However, these are
    (possibly recursive) subroutine calls, _not_ backreferences. We return
    the ESC_g code.

    Summary: Return a negative number for a numerical back reference, ESC_k for
    a named back reference, and ESC_g for a named or numbered subroutine call.
    */

    case CHAR_g:
    if (isclass) break;

    if (ptr >= ptrend)
      {
      *errorcodeptr = ERR57;
      break;
      }

    if (*ptr == CHAR_LESS_THAN_SIGN || *ptr == CHAR_APOSTROPHE)
      {
      escape = ESC_g;
      break;
      }

    /* If there is a brace delimiter, try to read a numerical reference. If
    there isn't one, assume we have a name and treat it as \k. */

    if (*ptr == CHAR_LEFT_CURLY_BRACKET)
      {
      PCRE2_SPTR p = ptr + 1;
      if (!read_number(&p, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &s,
          errorcodeptr))
        {
        if (*errorcodeptr == 0) escape = ESC_k;  /* No number found */
        break;
        }
      if (p >= ptrend || *p != CHAR_RIGHT_CURLY_BRACKET)
        {
        *errorcodeptr = ERR57;
        break;
        }
      ptr = p + 1;
      }

    /* Read an undelimited number */

    else
      {
      if (!read_number(&ptr, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &s,
          errorcodeptr))
        {
        if (*errorcodeptr == 0) *errorcodeptr = ERR57;  /* No number found */
        break;
        }
      }

    if (s <= 0)
      {
      *errorcodeptr = ERR15;
      break;
      }

    escape = -s;
    break;

    /* The handling of escape sequences consisting of a string of digits
    starting with one that is not zero is not straightforward. Perl has changed
    over the years. Nowadays \g{} for backreferences and \o{} for octal are
    recommended to avoid the ambiguities in the old syntax.

    Outside a character class, the digits are read as a decimal number. If the
    number is less than 10, or if there are that many previous extracting left
    brackets, it is a back reference. Otherwise, up to three octal digits are
    read to form an escaped character code. Thus \123 is likely to be octal 123
    (cf \0123, which is octal 012 followed by the literal 3).

    Inside a character class, \ followed by a digit is always either a literal
    8 or 9 or an octal number. */

    case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
    case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:

    if (!isclass)
      {
      oldptr = ptr;
      ptr--;   /* Back to the digit */
      if (!read_number(&ptr, ptrend, -1, INT_MAX/10 - 1, ERR61, &s,
          errorcodeptr))
        break;

      /* \1 to \9 are always back references. \8x and \9x are too; \1x to \7x
      are octal escapes if there are not that many previous captures. */

      if (s < 10 || oldptr[-1] >= CHAR_8 || s <= (int)cb->bracount)
        {
        if (s > (int)MAX_GROUP_NUMBER) *errorcodeptr = ERR61;
          else escape = -s;     /* Indicates a back reference */
        break;
        }
      ptr = oldptr;      /* Put the pointer back and fall through */
      }

    /* Handle a digit following \ when the number is not a back reference, or
    we are within a character class. If the first digit is 8 or 9, Perl used to
    generate a binary zero and then treat the digit as a following literal. At
    least by Perl 5.18 this changed so as not to insert the binary zero. */

    if (c >= CHAR_8) break;

    /* Fall through */

    /* \0 always starts an octal number, but we may drop through to here with a
    larger first octal digit. The original code used just to take the least
    significant 8 bits of octal numbers (I think this is what early Perls used
    to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
    but no more than 3 octal digits. */

    case CHAR_0:
    c -= CHAR_0;
    while(i++ < 2 && ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7)
        c = c * 8 + *ptr++ - CHAR_0;
#if PCRE2_CODE_UNIT_WIDTH == 8
    if (!utf && c > 0xff) *errorcodeptr = ERR51;
#endif
    break;

    /* \o is a relatively new Perl feature, supporting a more general way of
    specifying character codes in octal. The only supported form is \o{ddd}. */

    case CHAR_o:
    if (ptr >= ptrend || *ptr++ != CHAR_LEFT_CURLY_BRACKET)
      {
      ptr--;
      *errorcodeptr = ERR55;
      }
    else if (ptr >= ptrend || *ptr == CHAR_RIGHT_CURLY_BRACKET)
      *errorcodeptr = ERR78;
    else
      {
      c = 0;
      overflow = FALSE;
      while (ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7)
        {
        cc = *ptr++;
        if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
#if PCRE2_CODE_UNIT_WIDTH == 32
        if (c >= 0x20000000l) { overflow = TRUE; break; }
#endif
        c = (c << 3) + (cc - CHAR_0);
#if PCRE2_CODE_UNIT_WIDTH == 8
        if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
#elif PCRE2_CODE_UNIT_WIDTH == 16
        if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
#elif PCRE2_CODE_UNIT_WIDTH == 32
        if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
#endif
        }
      if (overflow)
        {
        while (ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
        *errorcodeptr = ERR34;
        }
      else if (ptr < ptrend && *ptr++ == CHAR_RIGHT_CURLY_BRACKET)
        {
        if (utf && c >= 0xd800 && c <= 0xdfff &&
            (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0)
          {
          ptr--;
          *errorcodeptr = ERR73;
          }
        }
      else
        {
        ptr--;
        *errorcodeptr = ERR64;
        }
      }
    break;

    /* When PCRE2_ALT_BSUX or PCRE2_EXTRA_ALT_BSUX is set, \x must be followed
    by two hexadecimal digits. Otherwise it is a lowercase x letter. */

    case CHAR_x:
    if (alt_bsux)
      {
      uint32_t xc;
      if (ptrend - ptr < 2) break;               /* Less than 2 characters */
      if ((cc = XDIGIT(ptr[0])) == 0xff) break;  /* Not a hex digit */
      if ((xc = XDIGIT(ptr[1])) == 0xff) break;  /* Not a hex digit */
      c = (cc << 4) | xc;
      ptr += 2;
      }

    /* Handle \x in Perl's style. \x{ddd} is a character code which can be
    greater than 0xff in UTF-8 or non-8bit mode, but only if the ddd are hex
    digits. If not, { used to be treated as a data character. However, Perl
    seems to read hex digits up to the first non-such, and ignore the rest, so
    that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
    now gives an error. */

    else
      {
      if (ptr < ptrend && *ptr == CHAR_LEFT_CURLY_BRACKET)
        {
#ifndef EBCDIC
        COME_FROM_NU:
#endif
        if (++ptr >= ptrend || *ptr == CHAR_RIGHT_CURLY_BRACKET)
          {
          *errorcodeptr = ERR78;
          break;
          }
        c = 0;
        overflow = FALSE;

        while (ptr < ptrend && (cc = XDIGIT(*ptr)) != 0xff)
          {
          ptr++;
          if (c == 0 && cc == 0) continue;   /* Leading zeroes */
#if PCRE2_CODE_UNIT_WIDTH == 32
          if (c >= 0x10000000l) { overflow = TRUE; break; }
#endif
          c = (c << 4) | cc;
          if ((utf && c > 0x10ffffU) || (!utf && c > MAX_NON_UTF_CHAR))
            {
            overflow = TRUE;
            break;
            }
          }

        if (overflow)
          {
          while (ptr < ptrend && XDIGIT(*ptr) != 0xff) ptr++;
          *errorcodeptr = ERR34;
          }
        else if (ptr < ptrend && *ptr++ == CHAR_RIGHT_CURLY_BRACKET)
          {
          if (utf && c >= 0xd800 && c <= 0xdfff &&
              (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0)
            {
            ptr--;
            *errorcodeptr = ERR73;
            }
          }

        /* If the sequence of hex digits does not end with '}', give an error.
        We used just to recognize this construct and fall through to the normal
        \x handling, but nowadays Perl gives an error, which seems much more
        sensible, so we do too. */

        else
          {
          ptr--;
          *errorcodeptr = ERR67;
          }
        }   /* End of \x{} processing */

      /* Read a up to two hex digits after \x */

      else
        {
        c = 0;
        if (ptr >= ptrend || (cc = XDIGIT(*ptr)) == 0xff) break;  /* Not a hex digit */
        ptr++;
        c = cc;
        if (ptr >= ptrend || (cc = XDIGIT(*ptr)) == 0xff) break;  /* Not a hex digit */
        ptr++;
        c = (c << 4) | cc;
        }     /* End of \xdd handling */
      }       /* End of Perl-style \x handling */
    break;

    /* The handling of \c is different in ASCII and EBCDIC environments. In an
    ASCII (or Unicode) environment, an error is given if the character
    following \c is not a printable ASCII character. Otherwise, the following
    character is upper-cased if it is a letter, and after that the 0x40 bit is
    flipped. The result is the value of the escape.

    In an EBCDIC environment the handling of \c is compatible with the
    specification in the perlebcdic document. The following character must be
    a letter or one of small number of special characters. These provide a
    means of defining the character values 0-31.

    For testing the EBCDIC handling of \c in an ASCII environment, recognize
    the EBCDIC value of 'c' explicitly. */

#if defined EBCDIC && 'a' != 0x81
    case 0x83:
#else
    case CHAR_c:
#endif
    if (ptr >= ptrend)
      {
      *errorcodeptr = ERR2;
      break;
      }
    c = *ptr;
    if (c >= CHAR_a && c <= CHAR_z) c = UPPER_CASE(c);

    /* Handle \c in an ASCII/Unicode environment. */

#ifndef EBCDIC    /* ASCII/UTF-8 coding */
    if (c < 32 || c > 126)  /* Excludes all non-printable ASCII */
      {
      *errorcodeptr = ERR68;
      break;
      }
    c ^= 0x40;

    /* Handle \c in an EBCDIC environment. The special case \c? is converted to
    255 (0xff) or 95 (0x5f) if other characters suggest we are using the
    POSIX-BC encoding. (This is the way Perl indicates that it handles \c?.)
    The other valid sequences correspond to a list of specific characters. */

#else
    if (c == CHAR_QUESTION_MARK)
      c = ('\\' == 188 && '`' == 74)? 0x5f : 0xff;
    else
      {
      for (i = 0; i < 32; i++)
        {
        if (c == ebcdic_escape_c[i]) break;
        }
      if (i < 32) c = i; else *errorcodeptr = ERR68;
      }
#endif  /* EBCDIC */

    ptr++;
    break;

    /* Any other alphanumeric following \ is an error. Perl gives an error only
    if in warning mode, but PCRE doesn't have a warning mode. */

    default:
    *errorcodeptr = ERR3;
    *ptrptr = ptr - 1;     /* Point to the character at fault */
    return 0;
    }
  }

/* Set the pointer to the next character before returning. */

*ptrptr = ptr;
*chptr = c;
return escape;
}



#ifdef SUPPORT_UNICODE
/*************************************************
*               Handle \P and \p                 *
*************************************************/

/* This function is called after \P or \p has been encountered, provided that
PCRE2 is compiled with support for UTF and Unicode properties. On entry, the
contents of ptrptr are pointing after the P or p. On exit, it is left pointing
after the final code unit of the escape sequence.

Arguments:
  ptrptr         the pattern position pointer
  negptr         a boolean that is set TRUE for negation else FALSE
  ptypeptr       an unsigned int that is set to the type value
  pdataptr       an unsigned int that is set to the detailed property value
  errorcodeptr   the error code variable
  cb             the compile data

Returns:         TRUE if the type value was found, or FALSE for an invalid type
*/

static BOOL
get_ucp(PCRE2_SPTR *ptrptr, BOOL *negptr, uint16_t *ptypeptr,
  uint16_t *pdataptr, int *errorcodeptr, compile_block *cb)
{
PCRE2_UCHAR c;
PCRE2_SIZE i, bot, top;
PCRE2_SPTR ptr = *ptrptr;
PCRE2_UCHAR name[32];

if (ptr >= cb->end_pattern) goto ERROR_RETURN;
c = *ptr++;
*negptr = FALSE;

/* \P or \p can be followed by a name in {}, optionally preceded by ^ for
negation. */

if (c == CHAR_LEFT_CURLY_BRACKET)
  {
  if (ptr >= cb->end_pattern) goto ERROR_RETURN;
  if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
    {
    *negptr = TRUE;
    ptr++;
    }
  for (i = 0; i < (int)(sizeof(name) / sizeof(PCRE2_UCHAR)) - 1; i++)
    {
    if (ptr >= cb->end_pattern) goto ERROR_RETURN;
    c = *ptr++;
    if (c == CHAR_NUL) goto ERROR_RETURN;
    if (c == CHAR_RIGHT_CURLY_BRACKET) break;
    name[i] = c;
    }
  if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
  name[i] = 0;
  }

/* Otherwise there is just one following character, which must be an ASCII
letter. */

else if (MAX_255(c) && (cb->ctypes[c] & ctype_letter) != 0)
  {
  name[0] = c;
  name[1] = 0;
  }
else goto ERROR_RETURN;

*ptrptr = ptr;

/* Search for a recognized property name using binary chop. */

bot = 0;
top = PRIV(utt_size);

while (bot < top)
  {
  int r;
  i = (bot + top) >> 1;
  r = PRIV(strcmp_c8)(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
  if (r == 0)
    {
    *ptypeptr = PRIV(utt)[i].type;
    *pdataptr = PRIV(utt)[i].value;
    return TRUE;
    }
  if (r > 0) bot = i + 1; else top = i;
  }
*errorcodeptr = ERR47;   /* Unrecognized name */
return FALSE;

ERROR_RETURN:            /* Malformed \P or \p */
*errorcodeptr = ERR46;
*ptrptr = ptr;
return FALSE;
}
#endif



/*************************************************
*           Check for POSIX class syntax         *
*************************************************/

/* This function is called when the sequence "[:" or "[." or "[=" is
encountered in a character class. It checks whether this is followed by a
sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
reach an unescaped ']' without the special preceding character, return FALSE.

Originally, this function only recognized a sequence of letters between the
terminators, but it seems that Perl recognizes any sequence of characters,
though of course unknown POSIX names are subsequently rejected. Perl gives an
"Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
didn't consider this to be a POSIX class. Likewise for [:1234:].

The problem in trying to be exactly like Perl is in the handling of escapes. We
have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
below handles the special cases \\ and \], but does not try to do any other
escape processing. This makes it different from Perl for cases such as
[:l\ower:] where Perl recognizes it as the POSIX class "lower" but PCRE does
not recognize "l\ower". This is a lesser evil than not diagnosing bad classes
when Perl does, I think.

A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
It seems that the appearance of a nested POSIX class supersedes an apparent
external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
a digit. This is handled by returning FALSE if the start of a new group with
the same terminator is encountered, since the next closing sequence must close
the nested group, not the outer one.

In Perl, unescaped square brackets may also appear as part of class names. For
example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
[:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
seem right at all. PCRE does not allow closing square brackets in POSIX class
names.

Arguments:
  ptr      pointer to the character after the initial [ (colon, dot, equals)
  ptrend   pointer to the end of the pattern
  endptr   where to return a pointer to the terminating ':', '.', or '='

Returns:   TRUE or FALSE
*/

static BOOL
check_posix_syntax(PCRE2_SPTR ptr, PCRE2_SPTR ptrend, PCRE2_SPTR *endptr)
{
PCRE2_UCHAR terminator;  /* Don't combine these lines; the Solaris cc */
terminator = *ptr++;     /* compiler warns about "non-constant" initializer. */

for (; ptrend - ptr >= 2; ptr++)
  {
  if (*ptr == CHAR_BACKSLASH &&
      (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET || ptr[1] == CHAR_BACKSLASH))
    ptr++;

  else if ((*ptr == CHAR_LEFT_SQUARE_BRACKET && ptr[1] == terminator) ||
            *ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;

  else if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
    {
    *endptr = ptr;
    return TRUE;
    }
  }

return FALSE;
}



/*************************************************
*          Check POSIX class name                *
*************************************************/

/* This function is called to check the name given in a POSIX-style class entry
such as [:alnum:].

Arguments:
  ptr        points to the first letter
  len        the length of the name

Returns:     a value representing the name, or -1 if unknown
*/

static int
check_posix_name(PCRE2_SPTR ptr, int len)
{
const char *pn = posix_names;
int yield = 0;
while (posix_name_lengths[yield] != 0)
  {
  if (len == posix_name_lengths[yield] &&
    PRIV(strncmp_c8)(ptr, pn, (unsigned int)len) == 0) return yield;
  pn += posix_name_lengths[yield] + 1;
  yield++;
  }
return -1;
}



/*************************************************
*       Read a subpattern or VERB name           *
*************************************************/

/* This function is called from parse_regex() below whenever it needs to read
the name of a subpattern or a (*VERB) or an (*alpha_assertion). The initial
pointer must be to the character before the name. If that character is '*' we
are reading a verb or alpha assertion name. The pointer is updated to point
after the name, for a VERB or alpha assertion name, or after tha name's
terminator for a subpattern name. Returning both the offset and the name
pointer is redundant information, but some callers use one and some the other,
so it is simplest just to return both.

Arguments:
  ptrptr      points to the character pointer variable
  ptrend      points to the end of the input string
  utf         true if the input is UTF-encoded
  terminator  the terminator of a subpattern name must be this
  offsetptr   where to put the offset from the start of the pattern
  nameptr     where to put a pointer to the name in the input
  namelenptr  where to put the length of the name
  errcodeptr  where to put an error code
  cb          pointer to the compile data block

Returns:    TRUE if a name was read
            FALSE otherwise, with error code set
*/

static BOOL
read_name(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, BOOL utf, uint32_t terminator,
  PCRE2_SIZE *offsetptr, PCRE2_SPTR *nameptr, uint32_t *namelenptr,
  int *errorcodeptr, compile_block *cb)
{
PCRE2_SPTR ptr = *ptrptr;
BOOL is_group = (*ptr != CHAR_ASTERISK);

if (++ptr >= ptrend)               /* No characters in name */
  {
  *errorcodeptr = is_group? ERR62: /* Subpattern name expected */
                            ERR60; /* Verb not recognized or malformed */
  goto FAILED;
  }

*nameptr = ptr;
*offsetptr = (PCRE2_SIZE)(ptr - cb->start_pattern);

/* In UTF mode, a group name may contain letters and decimal digits as defined
by Unicode properties, and underscores, but must not start with a digit. */

#ifdef SUPPORT_UNICODE
if (utf && is_group)
  {
  uint32_t c, type;

  GETCHAR(c, ptr);
  type = UCD_CHARTYPE(c);

  if (type == ucp_Nd)
    {
    *errorcodeptr = ERR44;
    goto FAILED;
    }

  for(;;)
    {
    if (type != ucp_Nd && PRIV(ucp_gentype)[type] != ucp_L &&
        c != CHAR_UNDERSCORE) break;
    ptr++;
    FORWARDCHARTEST(ptr, ptrend);
    if (ptr >= ptrend) break;
    GETCHAR(c, ptr);
    type = UCD_CHARTYPE(c);
    }
  }
else
#else
(void)utf;  /* Avoid compiler warning */
#endif      /* SUPPORT_UNICODE */

/* Handle non-group names and group names in non-UTF modes. A group name must
not start with a digit. If either of the others start with a digit it just
won't be recognized. */

  {
  if (is_group && IS_DIGIT(*ptr))
    {
    *errorcodeptr = ERR44;
    goto FAILED;
    }

  while (ptr < ptrend && MAX_255(*ptr) && (cb->ctypes[*ptr] & ctype_word) != 0)
    {
    ptr++;
    }
  }

/* Check name length */

if (ptr > *nameptr + MAX_NAME_SIZE)
  {
  *errorcodeptr = ERR48;
  goto FAILED;
  }
*namelenptr = ptr - *nameptr;

/* Subpattern names must not be empty, and their terminator is checked here.
(What follows a verb or alpha assertion name is checked separately.) */

if (is_group)
  {
  if (ptr == *nameptr)
    {
    *errorcodeptr = ERR62;   /* Subpattern name expected */
    goto FAILED;
    }
  if (ptr >= ptrend || *ptr != (PCRE2_UCHAR)terminator)
    {
    *errorcodeptr = ERR42;
    goto FAILED;
    }
  ptr++;
  }

*ptrptr = ptr;
return TRUE;

FAILED:
*ptrptr = ptr;
return FALSE;
}



/*************************************************
*          Manage callouts at start of cycle     *
*************************************************/

/* At the start of a new item in parse_regex() we are able to record the
details of the previous item in a prior callout, and also to set up an
automatic callout if enabled. Avoid having two adjacent automatic callouts,
which would otherwise happen for items such as \Q that contribute nothing to
the parsed pattern.

Arguments:
  ptr              current pattern pointer
  pcalloutptr      points to a pointer to previous callout, or NULL
  auto_callout     TRUE if auto_callouts are enabled
  parsed_pattern   the parsed pattern pointer
  cb               compile block

Returns: possibly updated parsed_pattern pointer.
*/

static uint32_t *
manage_callouts(PCRE2_SPTR ptr, uint32_t **pcalloutptr, BOOL auto_callout,
  uint32_t *parsed_pattern, compile_block *cb)
{
uint32_t *previous_callout = *pcalloutptr;

if (previous_callout != NULL) previous_callout[2] = (uint32_t)(ptr -
  cb->start_pattern - (PCRE2_SIZE)previous_callout[1]);

if (!auto_callout) previous_callout = NULL; else
  {
  if (previous_callout == NULL ||
      previous_callout != parsed_pattern - 4 ||
      previous_callout[3] != 255)
    {
    previous_callout = parsed_pattern;  /* Set up new automatic callout */
    parsed_pattern += 4;
    previous_callout[0] = META_CALLOUT_NUMBER;
    previous_callout[2] = 0;
    previous_callout[3] = 255;
    }
  previous_callout[1] = (uint32_t)(ptr - cb->start_pattern);
  }

*pcalloutptr = previous_callout;
return parsed_pattern;
}



/*************************************************
*      Parse regex and identify named groups     *
*************************************************/

/* This function is called first of all. It scans the pattern and does two
things: (1) It identifies capturing groups and makes a table of named capturing
groups so that information about them is fully available to both the compiling
scans. (2) It writes a parsed version of the pattern with comments omitted and
escapes processed into the parsed_pattern vector.

Arguments:
  ptr             points to the start of the pattern
  options         compiling dynamic options (may change during the scan)
  has_lookbehind  points to a boolean, set TRUE if a lookbehind is found
  cb              pointer to the compile data block

Returns:   zero on success or a non-zero error code, with the
             error offset placed in the cb field
*/

/* A structure and some flags for dealing with nested groups. */

typedef struct nest_save {
  uint16_t  nest_depth;
  uint16_t  reset_group;
  uint16_t  max_group;
  uint16_t  flags;
  uint32_t  options;
} nest_save;

#define NSF_RESET          0x0001u
#define NSF_CONDASSERT     0x0002u
#define NSF_ATOMICSR       0x0004u

/* Options that are changeable within the pattern must be tracked during
parsing. Some (e.g. PCRE2_EXTENDED) are implemented entirely during parsing,
but all must be tracked so that META_OPTIONS items set the correct values for
the main compiling phase. */

#define PARSE_TRACKED_OPTIONS (PCRE2_CASELESS|PCRE2_DOTALL|PCRE2_DUPNAMES| \
  PCRE2_EXTENDED|PCRE2_EXTENDED_MORE|PCRE2_MULTILINE|PCRE2_NO_AUTO_CAPTURE| \
  PCRE2_UNGREEDY)

/* States used for analyzing ranges in character classes. The two OK values
must be last. */

enum { RANGE_NO, RANGE_STARTED, RANGE_OK_ESCAPED, RANGE_OK_LITERAL };

/* Only in 32-bit mode can there be literals > META_END. A macros encapsulates
the storing of literal values in the parsed pattern. */

#if PCRE2_CODE_UNIT_WIDTH == 32
#define PARSED_LITERAL(c, p) \
  { \
  if (c >= META_END) *p++ = META_BIGVALUE; \
  *p++ = c; \
  okquantifier = TRUE; \
  }
#else
#define PARSED_LITERAL(c, p) *p++ = c; okquantifier = TRUE;
#endif

/* Here's the actual function. */

static int parse_regex(PCRE2_SPTR ptr, uint32_t options, BOOL *has_lookbehind,
  compile_block *cb)
{
uint32_t c;
uint32_t delimiter;
uint32_t namelen;
uint32_t class_range_state;
uint32_t *verblengthptr = NULL;     /* Value avoids compiler warning */
uint32_t *previous_callout = NULL;
uint32_t *parsed_pattern = cb->parsed_pattern;
uint32_t *parsed_pattern_end = cb->parsed_pattern_end;
uint32_t meta_quantifier = 0;
uint32_t add_after_mark = 0;
uint32_t extra_options = cb->cx->extra_options;
uint16_t nest_depth = 0;
int after_manual_callout = 0;
int expect_cond_assert = 0;
int errorcode = 0;
int escape;
int i;
BOOL inescq = FALSE;
BOOL inverbname = FALSE;
BOOL utf = (options & PCRE2_UTF) != 0;
BOOL auto_callout = (options & PCRE2_AUTO_CALLOUT) != 0;
BOOL isdupname;
BOOL negate_class;
BOOL okquantifier = FALSE;
PCRE2_SPTR thisptr;
PCRE2_SPTR name;
PCRE2_SPTR ptrend = cb->end_pattern;
PCRE2_SPTR verbnamestart = NULL;    /* Value avoids compiler warning */
named_group *ng;
nest_save *top_nest, *end_nests;

/* Insert leading items for word and line matching (features provided for the
benefit of pcre2grep). */

if ((extra_options & PCRE2_EXTRA_MATCH_LINE) != 0)
  {
  *parsed_pattern++ = META_CIRCUMFLEX;
  *parsed_pattern++ = META_NOCAPTURE;
  }
else if ((extra_options & PCRE2_EXTRA_MATCH_WORD) != 0)
  {
  *parsed_pattern++ = META_ESCAPE + ESC_b;
  *parsed_pattern++ = META_NOCAPTURE;
  }

/* If the pattern is actually a literal string, process it separately to avoid
cluttering up the main loop. */

if ((options & PCRE2_LITERAL) != 0)
  {
  while (ptr < ptrend)
    {
    if (parsed_pattern >= parsed_pattern_end)
      {
      errorcode = ERR63;  /* Internal error (parsed pattern overflow) */
      goto FAILED;
      }
    thisptr = ptr;
    GETCHARINCTEST(c, ptr);
    if (auto_callout)
      parsed_pattern = manage_callouts(thisptr, &previous_callout,
        auto_callout, parsed_pattern, cb);
    PARSED_LITERAL(c, parsed_pattern);
    }
  goto PARSED_END;
  }

/* Process a real regex which may contain meta-characters. */

top_nest = NULL;
end_nests = (nest_save *)(cb->start_workspace + cb->workspace_size);

/* The size of the nest_save structure might not be a factor of the size of the
workspace. Therefore we must round down end_nests so as to correctly avoid
creating a nest_save that spans the end of the workspace. */

end_nests = (nest_save *)((char *)end_nests -
  ((cb->workspace_size * sizeof(PCRE2_UCHAR)) % sizeof(nest_save)));

/* PCRE2_EXTENDED_MORE implies PCRE2_EXTENDED */

if ((options & PCRE2_EXTENDED_MORE) != 0) options |= PCRE2_EXTENDED;

/* Now scan the pattern */

while (ptr < ptrend)
  {
  int prev_expect_cond_assert;
  uint32_t min_repeat, max_repeat;
  uint32_t set, unset, *optset;
  uint32_t terminator;
  uint32_t prev_meta_quantifier;
  BOOL prev_okquantifier;
  PCRE2_SPTR tempptr;
  PCRE2_SIZE offset;

  if (parsed_pattern >= parsed_pattern_end)
    {
    errorcode = ERR63;  /* Internal error (parsed pattern overflow) */
    goto FAILED;
    }

  if (nest_depth > cb->cx->parens_nest_limit)
    {
    errorcode = ERR19;
    goto FAILED;        /* Parentheses too deeply nested */
    }

  /* Get next input character, save its position for callout handling. */

  thisptr = ptr;
  GETCHARINCTEST(c, ptr);

  /* Copy quoted literals until \E, allowing for the possibility of automatic
  callouts, except when processing a (*VERB) "name".  */

  if (inescq)
    {
    if (c == CHAR_BACKSLASH && ptr < ptrend && *ptr == CHAR_E)
      {
      inescq = FALSE;
      ptr++;   /* Skip E */
      }
    else
      {
      if (expect_cond_assert > 0)   /* A literal is not allowed if we are */
        {                           /* expecting a conditional assertion, */
        ptr--;                      /* but an empty \Q\E sequence is OK.  */
        errorcode = ERR28;
        goto FAILED;
        }
      if (!inverbname && after_manual_callout-- <= 0)
        parsed_pattern = manage_callouts(thisptr, &previous_callout,
          auto_callout, parsed_pattern, cb);
      PARSED_LITERAL(c, parsed_pattern);
      meta_quantifier = 0;
      }
    continue;  /* Next character */
    }

  /* If we are processing the "name" part of a (*VERB:NAME) item, all
  characters up to the closing parenthesis are literals except when
  PCRE2_ALT_VERBNAMES is set. That causes backslash interpretation, but only \Q
  and \E and escaped characters are allowed (no character types such as \d). If
  PCRE2_EXTENDED is also set, we must ignore white space and # comments. Do
  this by not entering the special (*VERB:NAME) processing - they are then
  picked up below. Note that c is a character, not a code unit, so we must not
  use MAX_255 to test its size because MAX_255 tests code units and is assumed
  TRUE in 8-bit mode. */

  if (inverbname &&
       (
        /* EITHER: not both options set */
        ((options & (PCRE2_EXTENDED | PCRE2_ALT_VERBNAMES)) !=
                    (PCRE2_EXTENDED | PCRE2_ALT_VERBNAMES)) ||
#ifdef SUPPORT_UNICODE
        /* OR: character > 255 AND not Unicode Pattern White Space */
        (c > 255 && (c|1) != 0x200f && (c|1) != 0x2029) ||
#endif
        /* OR: not a # comment or isspace() white space */
        (c < 256 && c != CHAR_NUMBER_SIGN && (cb->ctypes[c] & ctype_space) == 0
#ifdef SUPPORT_UNICODE
        /* and not CHAR_NEL when Unicode is supported */
          && c != CHAR_NEL
#endif
       )))
    {
    PCRE2_SIZE verbnamelength;

    switch(c)
      {
      default:
      PARSED_LITERAL(c, parsed_pattern);
      break;

      case CHAR_RIGHT_PARENTHESIS:
      inverbname = FALSE;
      okquantifier = FALSE;   /* Was probably set by literals */
      /* This is the length in characters */
      verbnamelength = (PCRE2_SIZE)(parsed_pattern - verblengthptr - 1);
      /* But the limit on the length is in code units */
      if (ptr - verbnamestart - 1 > (int)MAX_MARK)
        {
        ptr--;
        errorcode = ERR76;
        goto FAILED;
        }
      *verblengthptr = (uint32_t)verbnamelength;

      /* If this name was on a verb such as (*ACCEPT) which does not continue,
      a (*MARK) was generated for the name. We now add the original verb as the
      next item. */

      if (add_after_mark != 0)
        {
        *parsed_pattern++ = add_after_mark;
        add_after_mark = 0;
        }
      break;

      case CHAR_BACKSLASH:
      if ((options & PCRE2_ALT_VERBNAMES) != 0)
        {
        escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options,
          cb->cx->extra_options, FALSE, cb);
        if (errorcode != 0) goto FAILED;
        }
      else escape = 0;   /* Treat all as literal */

      switch(escape)
        {
        case 0:
        PARSED_LITERAL(c, parsed_pattern);
        break;

        case ESC_Q:
        inescq = TRUE;
        break;

        case ESC_E:           /* Ignore */
        break;

        default:
        errorcode = ERR40;    /* Invalid in verb name */
        goto FAILED;
        }
      }
    continue;   /* Next character in pattern */
    }

  /* Not a verb name character. At this point we must process everything that
  must not change the quantification state. This is mainly comments, but we
  handle \Q and \E here as well, so that an item such as A\Q\E+ is treated as
  A+, as in Perl. An isolated \E is ignored. */

  if (c == CHAR_BACKSLASH && ptr < ptrend)
    {
    if (*ptr == CHAR_Q || *ptr == CHAR_E)
      {
      inescq = *ptr == CHAR_Q;
      ptr++;
      continue;
      }
    }

  /* Skip over whitespace and # comments in extended mode. Note that c is a
  character, not a code unit, so we must not use MAX_255 to test its size
  because MAX_255 tests code units and is assumed TRUE in 8-bit mode. The
  whitespace characters are those designated as "Pattern White Space" by
  Unicode, which are the isspace() characters plus CHAR_NEL (newline), which is
  U+0085 in Unicode, plus U+200E, U+200F, U+2028, and U+2029. These are a
  subset of space characters that match \h and \v. */

  if ((options & PCRE2_EXTENDED) != 0)
    {
    if (c < 256 && (cb->ctypes[c] & ctype_space) != 0) continue;
#ifdef SUPPORT_UNICODE
    if (c == CHAR_NEL || (c|1) == 0x200f || (c|1) == 0x2029) continue;
#endif
    if (c == CHAR_NUMBER_SIGN)
      {
      while (ptr < ptrend)
        {
        if (IS_NEWLINE(ptr))      /* For non-fixed-length newline cases, */
          {                       /* IS_NEWLINE sets cb->nllen. */
          ptr += cb->nllen;
          break;
          }
        ptr++;
#ifdef SUPPORT_UNICODE
        if (utf) FORWARDCHARTEST(ptr, ptrend);
#endif
        }
      continue;  /* Next character in pattern */
      }
    }

  /* Skip over bracketed comments */

  if (c == CHAR_LEFT_PARENTHESIS && ptrend - ptr >= 2 &&
      ptr[0] == CHAR_QUESTION_MARK && ptr[1] == CHAR_NUMBER_SIGN)
    {
    while (++ptr < ptrend && *ptr != CHAR_RIGHT_PARENTHESIS);
    if (ptr >= ptrend)
      {
      errorcode = ERR18;  /* A special error for missing ) in a comment */
      goto FAILED;        /* to make it easier to debug. */
      }
    ptr++;
    continue;  /* Next character in pattern */
    }

  /* If the next item is not a quantifier, fill in length of any previous
  callout and create an auto callout if required. */

  if (c != CHAR_ASTERISK && c != CHAR_PLUS && c != CHAR_QUESTION_MARK &&
       (c != CHAR_LEFT_CURLY_BRACKET ||
         (tempptr = ptr,
         !read_repeat_counts(&tempptr, ptrend, NULL, NULL, &errorcode))))
    {
    if (after_manual_callout-- <= 0)
      parsed_pattern = manage_callouts(thisptr, &previous_callout, auto_callout,
        parsed_pattern, cb);
    }

  /* If expect_cond_assert is 2, we have just passed (?( and are expecting an
  assertion, possibly preceded by a callout. If the value is 1, we have just
  had the callout and expect an assertion. There must be at least 3 more
  characters in all cases. When expect_cond_assert is 2, we know that the
  current character is an opening parenthesis, as otherwise we wouldn't be
  here. However, when it is 1, we need to check, and it's easiest just to check
  always. Note that expect_cond_assert may be negative, since all callouts just
  decrement it. */

  if (expect_cond_assert > 0)
    {
    BOOL ok = c == CHAR_LEFT_PARENTHESIS && ptrend - ptr >= 3 &&
              (ptr[0] == CHAR_QUESTION_MARK || ptr[0] == CHAR_ASTERISK);
    if (ok)
      {
      if (ptr[0] == CHAR_ASTERISK)  /* New alpha assertion format, possibly */
        {
        ok = MAX_255(ptr[1]) && (cb->ctypes[ptr[1]] & ctype_lcletter) != 0;
        }
      else switch(ptr[1])  /* Traditional symbolic format */
        {
        case CHAR_C:
        ok = expect_cond_assert == 2;
        break;

        case CHAR_EQUALS_SIGN:
        case CHAR_EXCLAMATION_MARK:
        break;

        case CHAR_LESS_THAN_SIGN:
        ok = ptr[2] == CHAR_EQUALS_SIGN || ptr[2] == CHAR_EXCLAMATION_MARK;
        break;

        default:
        ok = FALSE;
        }
      }

    if (!ok)
      {
      ptr--;   /* Adjust error offset */
      errorcode = ERR28;
      goto FAILED;
      }
    }

  /* Remember whether we are expecting a conditional assertion, and set the
  default for this item. */

  prev_expect_cond_assert = expect_cond_assert;
  expect_cond_assert = 0;

  /* Remember quantification status for the previous significant item, then set
  default for this item. */

  prev_okquantifier = okquantifier;
  prev_meta_quantifier = meta_quantifier;
  okquantifier = FALSE;
  meta_quantifier = 0;

  /* If the previous significant item was a quantifier, adjust the parsed code
  if there is a following modifier. The base meta value is always followed by
  the PLUS and QUERY values, in that order. We do this here rather than after
  reading a quantifier so that intervening comments and /x whitespace can be
  ignored without having to replicate code. */

  if (prev_meta_quantifier != 0 && (c == CHAR_QUESTION_MARK || c == CHAR_PLUS))
    {
    parsed_pattern[(prev_meta_quantifier == META_MINMAX)? -3 : -1] =
      prev_meta_quantifier + ((c == CHAR_QUESTION_MARK)?
        0x00020000u : 0x00010000u);
    continue;  /* Next character in pattern */
    }


  /* Process the next item in the main part of a pattern. */

  switch(c)
    {
    default:              /* Non-special character */
    PARSED_LITERAL(c, parsed_pattern);
    break;


    /* ---- Escape sequence ---- */

    case CHAR_BACKSLASH:
    tempptr = ptr;
    escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options,
      cb->cx->extra_options, FALSE, cb);
    if (errorcode != 0)
      {
      ESCAPE_FAILED:
      if ((extra_options & PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL) == 0)
        goto FAILED;
      ptr = tempptr;
      if (ptr >= ptrend) c = CHAR_BACKSLASH; else
        {
        GETCHARINCTEST(c, ptr);   /* Get character value, increment pointer */
        }
      escape = 0;                 /* Treat as literal character */
      }

    /* The escape was a data escape or literal character. */

    if (escape == 0)
      {
      PARSED_LITERAL(c, parsed_pattern);
      }

    /* The escape was a back (or forward) reference. We keep the offset in
    order to give a more useful diagnostic for a bad forward reference. For
    references to groups numbered less than 10 we can't use more than two items
    in parsed_pattern because they may be just two characters in the input (and
    in a 64-bit world an offset may need two elements). So for them, the offset
    of the first occurrent is held in a special vector. */

    else if (escape < 0)
      {
      offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 1);
      escape = -escape;
      *parsed_pattern++ = META_BACKREF | (uint32_t)escape;
      if (escape < 10)
        {
        if (cb->small_ref_offset[escape] == PCRE2_UNSET)
          cb->small_ref_offset[escape] = offset;
        }
      else
        {
        PUTOFFSET(offset, parsed_pattern);
        }
      okquantifier = TRUE;
      }

    /* The escape was a character class such as \d etc. or other special
    escape indicator such as \A or \X. Most of them generate just a single
    parsed item, but \P and \p are followed by a 16-bit type and a 16-bit
    value. They are supported only when Unicode is available. The type and
    value are packed into a single 32-bit value so that the whole sequences
    uses only two elements in the parsed_vector. This is because the same
    coding is used if \d (for example) is turned into \p{Nd} when PCRE2_UCP is
    set.

    There are also some cases where the escape sequence is followed by a name:
    \k{name}, \k<name>, and \k'name' are backreferences by name, and \g<name>
    and \g'name' are subroutine calls by name; \g{name} is a synonym for
    \k{name}. Note that \g<number> and \g'number' are handled by check_escape()
    and returned as a negative value (handled above). A name is coded as an
    offset into the pattern and a length. */

    else switch (escape)
      {
      case ESC_C:
#ifdef NEVER_BACKSLASH_C
      errorcode = ERR85;
      goto ESCAPE_FAILED;
#else
      if ((options & PCRE2_NEVER_BACKSLASH_C) != 0)
        {
        errorcode = ERR83;
        goto ESCAPE_FAILED;
        }
#endif
      okquantifier = TRUE;
      *parsed_pattern++ = META_ESCAPE + escape;
      break;

      case ESC_X:
#ifndef SUPPORT_UNICODE
      errorcode = ERR45;   /* Supported only with Unicode support */
      goto ESCAPE_FAILED;
#endif
      case ESC_H:
      case ESC_h:
      case ESC_N:
      case ESC_R:
      case ESC_V:
      case ESC_v:
      okquantifier = TRUE;
      *parsed_pattern++ = META_ESCAPE + escape;
      break;

      default:  /* \A, \B, \b, \G, \K, \Z, \z cannot be quantified. */
      *parsed_pattern++ = META_ESCAPE + escape;
      break;

      /* Escapes that change in UCP mode. Note that PCRE2_UCP will never be set
      without Unicode support because it is checked when pcre2_compile() is
      called. */

      case ESC_d:
      case ESC_D:
      case ESC_s:
      case ESC_S:
      case ESC_w:
      case ESC_W:
      okquantifier = TRUE;
      if ((options & PCRE2_UCP) == 0)
        {
        *parsed_pattern++ = META_ESCAPE + escape;
        }
      else
        {
        *parsed_pattern++ = META_ESCAPE +
          ((escape == ESC_d || escape == ESC_s || escape == ESC_w)?
            ESC_p : ESC_P);
        switch(escape)
          {
          case ESC_d:
          case ESC_D:
          *parsed_pattern++ = (PT_PC << 16) | ucp_Nd;
          break;

          case ESC_s:
          case ESC_S:
          *parsed_pattern++ = PT_SPACE << 16;
          break;

          case ESC_w:
          case ESC_W:
          *parsed_pattern++ = PT_WORD << 16;
          break;
          }
        }
      break;

      /* Unicode property matching */

      case ESC_P:
      case ESC_p:
#ifdef SUPPORT_UNICODE
        {
        BOOL negated;
        uint16_t ptype = 0, pdata = 0;
        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcode, cb))
          goto ESCAPE_FAILED;
        if (negated) escape = (escape == ESC_P)? ESC_p : ESC_P;
        *parsed_pattern++ = META_ESCAPE + escape;
        *parsed_pattern++ = (ptype << 16) | pdata;
        okquantifier = TRUE;
        }
#else
      errorcode = ERR45;
      goto ESCAPE_FAILED;
#endif
      break;  /* End \P and \p */

      /* When \g is used with quotes or angle brackets as delimiters, it is a
      numerical or named subroutine call, and control comes here. When used
      with brace delimiters it is a numberical back reference and does not come
      here because check_escape() returns it directly as a reference. \k is
      always a named back reference. */

      case ESC_g:
      case ESC_k:
      if (ptr >= ptrend || (*ptr != CHAR_LEFT_CURLY_BRACKET &&
          *ptr != CHAR_LESS_THAN_SIGN && *ptr != CHAR_APOSTROPHE))
        {
        errorcode = (escape == ESC_g)? ERR57 : ERR69;
        goto ESCAPE_FAILED;
        }
      terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
        CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
        CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;

      /* For a non-braced \g, check for a numerical recursion. */

      if (escape == ESC_g && terminator != CHAR_RIGHT_CURLY_BRACKET)
        {
        PCRE2_SPTR p = ptr + 1;

        if (read_number(&p, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &i,
            &errorcode))
          {
          if (p >= ptrend || *p != terminator)
            {
            errorcode = ERR57;
            goto ESCAPE_FAILED;
            }
          ptr = p;
          goto SET_RECURSION;
          }
        if (errorcode != 0) goto ESCAPE_FAILED;
        }

      /* Not a numerical recursion */

      if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen,
          &errorcode, cb)) goto ESCAPE_FAILED;

      /* \k and \g when used with braces are back references, whereas \g used
      with quotes or angle brackets is a recursion */

      *parsed_pattern++ =
        (escape == ESC_k || terminator == CHAR_RIGHT_CURLY_BRACKET)?
          META_BACKREF_BYNAME : META_RECURSE_BYNAME;
      *parsed_pattern++ = namelen;

      PUTOFFSET(offset, parsed_pattern);
      okquantifier = TRUE;
      break;  /* End special escape processing */
      }
    break;    /* End escape sequence processing */


    /* ---- Single-character special items ---- */

    case CHAR_CIRCUMFLEX_ACCENT:
    *parsed_pattern++ = META_CIRCUMFLEX;
    break;

    case CHAR_DOLLAR_SIGN:
    *parsed_pattern++ = META_DOLLAR;
    break;

    case CHAR_DOT:
    *parsed_pattern++ = META_DOT;
    okquantifier = TRUE;
    break;


    /* ---- Single-character quantifiers ---- */

    case CHAR_ASTERISK:
    meta_quantifier = META_ASTERISK;
    goto CHECK_QUANTIFIER;

    case CHAR_PLUS:
    meta_quantifier = META_PLUS;
    goto CHECK_QUANTIFIER;

    case CHAR_QUESTION_MARK:
    meta_quantifier = META_QUERY;
    goto CHECK_QUANTIFIER;


    /* ---- Potential {n,m} quantifier ---- */

    case CHAR_LEFT_CURLY_BRACKET:
    if (!read_repeat_counts(&ptr, ptrend, &min_repeat, &max_repeat,
        &errorcode))
      {
      if (errorcode != 0) goto FAILED;     /* Error in quantifier. */
      PARSED_LITERAL(c, parsed_pattern);   /* Not a quantifier */
      break;                               /* No more quantifier processing */
      }
    meta_quantifier = META_MINMAX;
    /* Fall through */


    /* ---- Quantifier post-processing ---- */

    /* Check that a quantifier is allowed after the previous item. */

    CHECK_QUANTIFIER:
    if (!prev_okquantifier)
      {
      errorcode = ERR9;
      goto FAILED_BACK;
      }

    /* Now we can put the quantifier into the parsed pattern vector. At this
    stage, we have only the basic quantifier. The check for a following + or ?
    modifier happens at the top of the loop, after any intervening comments
    have been removed. */

    *parsed_pattern++ = meta_quantifier;
    if (c == CHAR_LEFT_CURLY_BRACKET)
      {
      *parsed_pattern++ = min_repeat;
      *parsed_pattern++ = max_repeat;
      }
    break;


    /* ---- Character class ---- */

    case CHAR_LEFT_SQUARE_BRACKET:
    okquantifier = TRUE;

    /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
    used for "start of word" and "end of word". As these are otherwise illegal
    sequences, we don't break anything by recognizing them. They are replaced
    by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
    erroneous and are handled by the normal code below. */

    if (ptrend - ptr >= 6 &&
         (PRIV(strncmp_c8)(ptr, STRING_WEIRD_STARTWORD, 6) == 0 ||
          PRIV(strncmp_c8)(ptr, STRING_WEIRD_ENDWORD, 6) == 0))
      {
      *parsed_pattern++ = META_ESCAPE + ESC_b;

      if (ptr[2] == CHAR_LESS_THAN_SIGN)
        {
        *parsed_pattern++ = META_LOOKAHEAD;
        }
      else
        {
        *parsed_pattern++ = META_LOOKBEHIND;
        *has_lookbehind = TRUE;

        /* The offset is used only for the "non-fixed length" error; this won't
        occur here, so just store zero. */

        PUTOFFSET((PCRE2_SIZE)0, parsed_pattern);
        }

      if ((options & PCRE2_UCP) == 0)
        *parsed_pattern++ = META_ESCAPE + ESC_w;
      else
        {
        *parsed_pattern++ = META_ESCAPE + ESC_p;
        *parsed_pattern++ = PT_WORD << 16;
        }
      *parsed_pattern++ = META_KET;
      ptr += 6;
      break;
      }

    /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
    they are encountered at the top level, so we'll do that too. */

    if (ptr < ptrend && (*ptr == CHAR_COLON || *ptr == CHAR_DOT ||
         *ptr == CHAR_EQUALS_SIGN) &&
        check_posix_syntax(ptr, ptrend, &tempptr))
      {
      errorcode = (*ptr-- == CHAR_COLON)? ERR12 : ERR13;
      goto FAILED;
      }

    /* Process a regular character class. If the first character is '^', set
    the negation flag. If the first few characters (either before or after ^)
    are \Q\E or \E or space or tab in extended-more mode, we skip them too.
    This makes for compatibility with Perl. */

    negate_class = FALSE;
    while (ptr < ptrend)
      {
      GETCHARINCTEST(c, ptr);
      if (c == CHAR_BACKSLASH)
        {
        if (ptr < ptrend && *ptr == CHAR_E) ptr++;
        else if (ptrend - ptr >= 3 &&
             PRIV(strncmp_c8)(ptr, STR_Q STR_BACKSLASH STR_E, 3) == 0)
          ptr += 3;
        else
          break;
        }
      else if ((options & PCRE2_EXTENDED_MORE) != 0 &&
               (c == CHAR_SPACE || c == CHAR_HT))  /* Note: just these two */
        continue;
      else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
        negate_class = TRUE;
      else break;
      }

    /* Now the real contents of the class; c has the first "real" character.
    Empty classes are permitted only if the option is set. */

    if (c == CHAR_RIGHT_SQUARE_BRACKET &&
        (cb->external_options & PCRE2_ALLOW_EMPTY_CLASS) != 0)
      {
      *parsed_pattern++ = negate_class? META_CLASS_EMPTY_NOT : META_CLASS_EMPTY;
      break;  /* End of class processing */
      }

    /* Process a non-empty class. */

    *parsed_pattern++ = negate_class? META_CLASS_NOT : META_CLASS;
    class_range_state = RANGE_NO;

    /* In an EBCDIC environment, Perl treats alphabetic ranges specially
    because there are holes in the encoding, and simply using the range A-Z
    (for example) would include the characters in the holes. This applies only
    to ranges where both values are literal; [\xC1-\xE9] is different to [A-Z]
    in this respect. In order to accommodate this, we keep track of whether
    character values are literal or not, and a state variable for handling
    ranges. */

    /* Loop for the contents of the class */

    for (;;)
      {
      BOOL char_is_literal = TRUE;

      /* Inside \Q...\E everything is literal except \E */

      if (inescq)
        {
        if (c == CHAR_BACKSLASH && ptr < ptrend && *ptr == CHAR_E)
          {
          inescq = FALSE;                   /* Reset literal state */
          ptr++;                            /* Skip the 'E' */
          goto CLASS_CONTINUE;
          }
        goto CLASS_LITERAL;
        }

      /* Skip over space and tab (only) in extended-more mode. */

      if ((options & PCRE2_EXTENDED_MORE) != 0 &&
          (c == CHAR_SPACE || c == CHAR_HT))
        goto CLASS_CONTINUE;

      /* Handle POSIX class names. Perl allows a negation extension of the
      form [:^name:]. A square bracket that doesn't match the syntax is
      treated as a literal. We also recognize the POSIX constructions
      [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
      5.6 and 5.8 do. */

      if (c == CHAR_LEFT_SQUARE_BRACKET &&
          ptrend - ptr >= 3 &&
          (*ptr == CHAR_COLON || *ptr == CHAR_DOT ||
           *ptr == CHAR_EQUALS_SIGN) &&
          check_posix_syntax(ptr, ptrend, &tempptr))
        {
        BOOL posix_negate = FALSE;
        int posix_class;

        /* Perl treats a hyphen before a POSIX class as a literal, not the
        start of a range. However, it gives a warning in its warning mode. PCRE
        does not have a warning mode, so we give an error, because this is
        likely an error on the user's part. */

        if (class_range_state == RANGE_STARTED)
          {
          errorcode = ERR50;
          goto FAILED;
          }

        if (*ptr != CHAR_COLON)
          {
          errorcode = ERR13;
          goto FAILED_BACK;
          }

        if (*(++ptr) == CHAR_CIRCUMFLEX_ACCENT)
          {
          posix_negate = TRUE;
          ptr++;
          }

        posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
        if (posix_class < 0)
          {
          errorcode = ERR30;
          goto FAILED;
          }
        ptr = tempptr + 2;

        /* Perl treats a hyphen after a POSIX class as a literal, not the
        start of a range. However, it gives a warning in its warning mode
        unless the hyphen is the last character in the class. PCRE does not
        have a warning mode, so we give an error, because this is likely an
        error on the user's part. */

        if (ptr < ptrend - 1 && *ptr == CHAR_MINUS &&
            ptr[1] != CHAR_RIGHT_SQUARE_BRACKET)
          {
          errorcode = ERR50;
          goto FAILED;
          }

        /* Set "a hyphen is not the start of a range" for the -] case, and also
        in case the POSIX class is followed by \E or \Q\E (possibly repeated -
        fuzzers do that kind of thing) and *then* a hyphen. This causes that
        hyphen to be treated as a literal. I don't think it's worth setting up
        special apparatus to do otherwise. */

        class_range_state = RANGE_NO;

        /* When PCRE2_UCP is set, some of the POSIX classes are converted to
        use Unicode properties \p or \P or, in one case, \h or \H. The
        substitutes table has two values per class, containing the type and
        value of a \p or \P item. The special cases are specified with a
        negative type: a non-zero value causes \h or \H to be used, and a zero
        value falls through to behave like a non-UCP POSIX class. */

#ifdef SUPPORT_UNICODE
        if ((options & PCRE2_UCP) != 0)
          {
          int ptype = posix_substitutes[2*posix_class];
          int pvalue = posix_substitutes[2*posix_class + 1];
          if (ptype >= 0)
            {
            *parsed_pattern++ = META_ESCAPE + (posix_negate? ESC_P : ESC_p);
            *parsed_pattern++ = (ptype << 16) | pvalue;
            goto CLASS_CONTINUE;
            }

          if (pvalue != 0)
            {
            *parsed_pattern++ = META_ESCAPE + (posix_negate? ESC_H : ESC_h);
            goto CLASS_CONTINUE;
            }

          /* Fall through */
          }
#endif  /* SUPPORT_UNICODE */

        /* Non-UCP POSIX class */

        *parsed_pattern++ = posix_negate? META_POSIX_NEG : META_POSIX;
        *parsed_pattern++ = posix_class;
        }

      /* Handle potential start of range */

      else if (c == CHAR_MINUS && class_range_state >= RANGE_OK_ESCAPED)
        {
        *parsed_pattern++ = (class_range_state == RANGE_OK_LITERAL)?
          META_RANGE_LITERAL : META_RANGE_ESCAPED;
        class_range_state = RANGE_STARTED;
        }

      /* Handle a literal character */

      else if (c != CHAR_BACKSLASH)
        {
        CLASS_LITERAL:
        if (class_range_state == RANGE_STARTED)
          {
          if (c == parsed_pattern[-2])       /* Optimize one-char range */
            parsed_pattern--;
          else if (parsed_pattern[-2] > c)   /* Check range is in order */
            {
            errorcode = ERR8;
            goto FAILED_BACK;
            }
          else
            {
            if (!char_is_literal && parsed_pattern[-1] == META_RANGE_LITERAL)
              parsed_pattern[-1] = META_RANGE_ESCAPED;
            PARSED_LITERAL(c, parsed_pattern);
            }
          class_range_state = RANGE_NO;
          }
        else  /* Potential start of range */
          {
          class_range_state = char_is_literal?
            RANGE_OK_LITERAL : RANGE_OK_ESCAPED;
          PARSED_LITERAL(c, parsed_pattern);
          }
        }

      /* Handle escapes in a class */

      else
        {
        tempptr = ptr;
        escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options,
          cb->cx->extra_options, TRUE, cb);

        if (errorcode != 0)
          {
          if ((extra_options & PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL) == 0)
            goto FAILED;
          ptr = tempptr;
          if (ptr >= ptrend) c = CHAR_BACKSLASH; else
            {
            GETCHARINCTEST(c, ptr);   /* Get character value, increment pointer */
            }
          escape = 0;                 /* Treat as literal character */
          }

        switch(escape)
          {
          case 0:  /* Escaped character code point is in c */
          char_is_literal = FALSE;
          goto CLASS_LITERAL;

          case ESC_b:
          c = CHAR_BS;    /* \b is backspace in a class */
          char_is_literal = FALSE;
          goto CLASS_LITERAL;

          case ESC_Q:
          inescq = TRUE;  /* Enter literal mode */
          goto CLASS_CONTINUE;

          case ESC_E:     /* Ignore orphan \E */
          goto CLASS_CONTINUE;

          case ESC_B:     /* Always an error in a class */
          case ESC_R:
          case ESC_X:
          errorcode = ERR7;
          ptr--;
          goto FAILED;
          }

        /* The second part of a range can be a single-character escape
        sequence (detected above), but not any of the other escapes. Perl
        treats a hyphen as a literal in such circumstances. However, in Perl's
        warning mode, a warning is given, so PCRE now faults it, as it is
        almost certainly a mistake on the user's part. */

        if (class_range_state == RANGE_STARTED)
          {
          errorcode = ERR50;
          goto FAILED;  /* Not CLASS_ESCAPE_FAILED; always an error */
          }

        /* Of the remaining escapes, only those that define characters are
        allowed in a class. None may start a range. */

        class_range_state = RANGE_NO;
        switch(escape)
          {
          case ESC_N:
          errorcode = ERR71;
          goto FAILED;

          case ESC_H:
          case ESC_h:
          case ESC_V:
          case ESC_v:
          *parsed_pattern++ = META_ESCAPE + escape;
          break;

          /* These escapes are converted to Unicode property tests when
          PCRE2_UCP is set. */

          case ESC_d:
          case ESC_D:
          case ESC_s:
          case ESC_S:
          case ESC_w:
          case ESC_W:
          if ((options & PCRE2_UCP) == 0)
            {
            *parsed_pattern++ = META_ESCAPE + escape;
            }
          else
            {
            *parsed_pattern++ = META_ESCAPE +
              ((escape == ESC_d || escape == ESC_s || escape == ESC_w)?
                ESC_p : ESC_P);
            switch(escape)
              {
              case ESC_d:
              case ESC_D:
              *parsed_pattern++ = (PT_PC << 16) | ucp_Nd;
              break;

              case ESC_s:
              case ESC_S:
              *parsed_pattern++ = PT_SPACE << 16;
              break;

              case ESC_w:
              case ESC_W:
              *parsed_pattern++ = PT_WORD << 16;
              break;
              }
            }
          break;

          /* Explicit Unicode property matching */

          case ESC_P:
          case ESC_p:
#ifdef SUPPORT_UNICODE
            {
            BOOL negated;
            uint16_t ptype = 0, pdata = 0;
            if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcode, cb))
              goto FAILED;
            if (negated) escape = (escape == ESC_P)? ESC_p : ESC_P;
            *parsed_pattern++ = META_ESCAPE + escape;
            *parsed_pattern++ = (ptype << 16) | pdata;
            }
#else
          errorcode = ERR45;
          goto FAILED;
#endif
          break;  /* End \P and \p */

          default:    /* All others are not allowed in a class */
          errorcode = ERR7;
          ptr--;
          goto FAILED;
          }

        /* Perl gives a warning unless a following hyphen is the last character
        in the class. PCRE throws an error. */

        if (ptr < ptrend - 1 && *ptr == CHAR_MINUS &&
            ptr[1] != CHAR_RIGHT_SQUARE_BRACKET)
          {
          errorcode = ERR50;
          goto FAILED;
          }
        }

      /* Proceed to next thing in the class. */

      CLASS_CONTINUE:
      if (ptr >= ptrend)
        {
        errorcode = ERR6;  /* Missing terminating ']' */
        goto FAILED;
        }
      GETCHARINCTEST(c, ptr);
      if (c == CHAR_RIGHT_SQUARE_BRACKET && !inescq) break;
      }     /* End of class-processing loop */

    if (class_range_state == RANGE_STARTED)
      {
      parsed_pattern[-1] = CHAR_MINUS;
      class_range_state = RANGE_NO;
      }

    *parsed_pattern++ = META_CLASS_END;
    break;  /* End of character class */


    /* ---- Opening parenthesis ---- */

    case CHAR_LEFT_PARENTHESIS:
    if (ptr >= ptrend) goto UNCLOSED_PARENTHESIS;

    /* If ( is not followed by ? it is either a capture or a special verb or an
    alpha assertion. */

    if (*ptr != CHAR_QUESTION_MARK)
      {
      const char *vn;

      /* Handle capturing brackets (or non-capturing if auto-capture is turned
      off). */

      if (*ptr != CHAR_ASTERISK)
        {
        nest_depth++;
        if ((options & PCRE2_NO_AUTO_CAPTURE) == 0)
          {
          cb->bracount++;
          *parsed_pattern++ = META_CAPTURE | cb->bracount;
          }
        else *parsed_pattern++ = META_NOCAPTURE;
        }

      /* Do nothing for (* followed by end of pattern or ) so it gives a "bad
      quantifier" error rather than "(*MARK) must have an argument". */

      else if (ptrend - ptr <= 1 || (c = ptr[1]) == CHAR_RIGHT_PARENTHESIS)
        break;

      /* Handle "alpha assertions" such as (*pla:...). Most of these are
      synonyms for the historical symbolic assertions, but the script run ones
      are new. They are distinguished by starting with a lower case letter.
      Checking both ends of the alphabet makes this work in all character
      codes. */

      else if (CHMAX_255(c) && (cb->ctypes[c] & ctype_lcletter) != 0)
        {
        uint32_t meta;

        vn = alasnames;
        if (!read_name(&ptr, ptrend, utf, 0, &offset, &name, &namelen,
          &errorcode, cb)) goto FAILED;
        if (ptr >= ptrend || *ptr != CHAR_COLON)
          {
          errorcode = ERR95;  /* Malformed */
          goto FAILED;
          }

        /* Scan the table of alpha assertion names */

        for (i = 0; i < alascount; i++)
          {
          if (namelen == alasmeta[i].len &&
              PRIV(strncmp_c8)(name, vn, namelen) == 0)
            break;
          vn += alasmeta[i].len + 1;
          }

        if (i >= alascount)
          {
          errorcode = ERR95;  /* Alpha assertion not recognized */
          goto FAILED;
          }

        /* Check for expecting an assertion condition. If so, only lookaround
        assertions are valid. */

        meta = alasmeta[i].meta;
        if (prev_expect_cond_assert > 0 &&
            (meta < META_LOOKAHEAD || meta > META_LOOKBEHINDNOT))
          {
          errorcode = ERR28;  /* Assertion expected */
          goto FAILED;
          }

        /* The lookaround alphabetic synonyms can be almost entirely handled by
        jumping to the code that handles the traditional symbolic forms. */

        switch(meta)
          {
          default:
          errorcode = ERR89;  /* Unknown code; should never occur because */
          goto FAILED;        /* the meta values come from a table above. */

          case META_ATOMIC:
          goto ATOMIC_GROUP;

          case META_LOOKAHEAD:
          goto POSITIVE_LOOK_AHEAD;

          case META_LOOKAHEADNOT:
          goto NEGATIVE_LOOK_AHEAD;

          case META_LOOKBEHIND:
          case META_LOOKBEHINDNOT:
          *parsed_pattern++ = meta;
          ptr--;
          goto POST_LOOKBEHIND;

          /* The script run facilities are handled here. Unicode support is
          required (give an error if not, as this is a security issue). Always
          record a META_SCRIPT_RUN item. Then, for the atomic version, insert
          META_ATOMIC and remember that we need two META_KETs at the end. */

          case META_SCRIPT_RUN:
          case META_ATOMIC_SCRIPT_RUN:
#ifdef SUPPORT_UNICODE
          *parsed_pattern++ = META_SCRIPT_RUN;
          nest_depth++;
          ptr++;
          if (meta == META_ATOMIC_SCRIPT_RUN)
            {
            *parsed_pattern++ = META_ATOMIC;
            if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace);
            else if (++top_nest >= end_nests)
              {
              errorcode = ERR84;
              goto FAILED;
              }
            top_nest->nest_depth = nest_depth;
            top_nest->flags = NSF_ATOMICSR;
            top_nest->options = options & PARSE_TRACKED_OPTIONS;
            }
          break;
#else  /* SUPPORT_UNICODE */
          errorcode = ERR96;
          goto FAILED;
#endif
          }
        }


      /* ---- Handle (*VERB) and (*VERB:NAME) ---- */

      else
        {
        vn = verbnames;
        if (!read_name(&ptr, ptrend, utf, 0, &offset, &name, &namelen,
          &errorcode, cb)) goto FAILED;
        if (ptr >= ptrend || (*ptr != CHAR_COLON &&
                              *ptr != CHAR_RIGHT_PARENTHESIS))
          {
          errorcode = ERR60;  /* Malformed */
          goto FAILED;
          }

        /* Scan the table of verb names */

        for (i = 0; i < verbcount; i++)
          {
          if (namelen == verbs[i].len &&
              PRIV(strncmp_c8)(name, vn, namelen) == 0)
            break;
          vn += verbs[i].len + 1;
          }

        if (i >= verbcount)
          {
          errorcode = ERR60;  /* Verb not recognized */
          goto FAILED;
          }

        /* An empty argument is treated as no argument. */

        if (*ptr == CHAR_COLON && ptr + 1 < ptrend &&
             ptr[1] == CHAR_RIGHT_PARENTHESIS)
          ptr++;    /* Advance to the closing parens */

        /* Check for mandatory non-empty argument; this is (*MARK) */

        if (verbs[i].has_arg > 0 && *ptr != CHAR_COLON)
          {
          errorcode = ERR66;
          goto FAILED;
          }

        /* It appears that Perl allows any characters whatsoever, other than a
        closing parenthesis, to appear in arguments ("names"), so we no longer
        insist on letters, digits, and underscores. Perl does not, however, do
        any interpretation within arguments, and has no means of including a
        closing parenthesis. PCRE supports escape processing but only when it
        is requested by an option. We set inverbname TRUE here, and let the
        main loop take care of this so that escape and \x processing is done by
        the main code above. */

        if (*ptr++ == CHAR_COLON)   /* Skip past : or ) */
          {
          /* Some optional arguments can be treated as a preceding (*MARK) */

          if (verbs[i].has_arg < 0)
            {
            add_after_mark = verbs[i].meta;
            *parsed_pattern++ = META_MARK;
            }

          /* The remaining verbs with arguments (except *MARK) need a different
          opcode. */

          else
            {
            *parsed_pattern++ = verbs[i].meta +
              ((verbs[i].meta != META_MARK)? 0x00010000u:0);
            }

          /* Set up for reading the name in the main loop. */

          verblengthptr = parsed_pattern++;
          verbnamestart = ptr;
          inverbname = TRUE;
          }
        else  /* No verb "name" argument */
          {
          *parsed_pattern++ = verbs[i].meta;
          }
        }     /* End of (*VERB) handling */
      break;  /* Done with this parenthesis */
      }       /* End of groups that don't start with (? */


    /* ---- Items starting (? ---- */

    /* The type of item is determined by what follows (?. Handle (?| and option
    changes under "default" because both need a new block on the nest stack.
    Comments starting with (?# are handled above. Note that there is some
    ambiguity about the sequence (?- because if a digit follows it's a relative
    recursion or subroutine call whereas otherwise it's an option unsetting. */

    if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS;

    switch(*ptr)
      {
      default:
      if (*ptr == CHAR_MINUS && ptrend - ptr > 1 && IS_DIGIT(ptr[1]))
        goto RECURSION_BYNUMBER;  /* The + case is handled by CHAR_PLUS */

      /* We now have either (?| or a (possibly empty) option setting,
      optionally followed by a non-capturing group. */

      nest_depth++;
      if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace);
      else if (++top_nest >= end_nests)
        {
        errorcode = ERR84;
        goto FAILED;
        }
      top_nest->nest_depth = nest_depth;
      top_nest->flags = 0;
      top_nest->options = options & PARSE_TRACKED_OPTIONS;

      /* Start of non-capturing group that resets the capture count for each
      branch. */

      if (*ptr == CHAR_VERTICAL_LINE)
        {
        top_nest->reset_group = (uint16_t)cb->bracount;
        top_nest->max_group = (uint16_t)cb->bracount;
        top_nest->flags |= NSF_RESET;
        cb->external_flags |= PCRE2_DUPCAPUSED;
        *parsed_pattern++ = META_NOCAPTURE;
        ptr++;
        }

      /* Scan for options imnsxJU to be set or unset. */

      else
        {
        BOOL hyphenok = TRUE;
        uint32_t oldoptions = options;

        top_nest->reset_group = 0;
        top_nest->max_group = 0;
        set = unset = 0;
        optset = &set;

        /* ^ at the start unsets imnsx and disables the subsequent use of - */

        if (ptr < ptrend && *ptr == CHAR_CIRCUMFLEX_ACCENT)
          {
          options &= ~(PCRE2_CASELESS|PCRE2_MULTILINE|PCRE2_NO_AUTO_CAPTURE|
                       PCRE2_DOTALL|PCRE2_EXTENDED|PCRE2_EXTENDED_MORE);
          hyphenok = FALSE;
          ptr++;
          }

        while (ptr < ptrend && *ptr != CHAR_RIGHT_PARENTHESIS &&
                               *ptr != CHAR_COLON)
          {
          switch (*ptr++)
            {
            case CHAR_MINUS:
            if (!hyphenok)
              {
              errorcode = ERR94;
              ptr--;  /* Correct the offset */
              goto FAILED;
              }
            optset = &unset;
            hyphenok = FALSE;
            break;

            case CHAR_J:  /* Record that it changed in the external options */
            *optset |= PCRE2_DUPNAMES;
            cb->external_flags |= PCRE2_JCHANGED;
            break;

            case CHAR_i: *optset |= PCRE2_CASELESS; break;
            case CHAR_m: *optset |= PCRE2_MULTILINE; break;
            case CHAR_n: *optset |= PCRE2_NO_AUTO_CAPTURE; break;
            case CHAR_s: *optset |= PCRE2_DOTALL; break;
            case CHAR_U: *optset |= PCRE2_UNGREEDY; break;

            /* If x appears twice it sets the extended extended option. */

            case CHAR_x:
            *optset |= PCRE2_EXTENDED;
            if (ptr < ptrend && *ptr == CHAR_x)
              {
              *optset |= PCRE2_EXTENDED_MORE;
              ptr++;
              }
            break;

            default:
            errorcode = ERR11;
            ptr--;    /* Correct the offset */
            goto FAILED;
            }
          }

        /* If we are setting extended without extended-more, ensure that any
        existing extended-more gets unset. Also, unsetting extended must also
        unset extended-more. */

        if ((set & (PCRE2_EXTENDED|PCRE2_EXTENDED_MORE)) == PCRE2_EXTENDED ||
            (unset & PCRE2_EXTENDED) != 0)
          unset |= PCRE2_EXTENDED_MORE;

        options = (options | set) & (~unset);

        /* If the options ended with ')' this is not the start of a nested
        group with option changes, so the options change at this level.
        In this case, if the previous level set up a nest block, discard the
        one we have just created. Otherwise adjust it for the previous level.
        If the options ended with ':' we are starting a non-capturing group,
        possibly with an options setting. */

        if (ptr >= ptrend) goto UNCLOSED_PARENTHESIS;
        if (*ptr++ == CHAR_RIGHT_PARENTHESIS)
          {
          nest_depth--;  /* This is not a nested group after all. */
          if (top_nest > (nest_save *)(cb->start_workspace) &&
              (top_nest-1)->nest_depth == nest_depth) top_nest--;
          else top_nest->nest_depth = nest_depth;
          }
        else *parsed_pattern++ = META_NOCAPTURE;

        /* If nothing changed, no need to record. */

        if (options != oldoptions)
          {
          *parsed_pattern++ = META_OPTIONS;
          *parsed_pattern++ = options;
          }
        }     /* End options processing */
      break;  /* End default case after (? */


      /* ---- Python syntax support ---- */

      case CHAR_P:
      if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS;

      /* (?P<name> is the same as (?<name>, which defines a named group. */

      if (*ptr == CHAR_LESS_THAN_SIGN)
        {
        terminator = CHAR_GREATER_THAN_SIGN;
        goto DEFINE_NAME;
        }

      /* (?P>name) is the same as (?&name), which is a recursion or subroutine
      call. */

      if (*ptr == CHAR_GREATER_THAN_SIGN) goto RECURSE_BY_NAME;

      /* (?P=name) is the same as \k<name>, a back reference by name. Anything
      else after (?P is an error. */

      if (*ptr != CHAR_EQUALS_SIGN)
        {
        errorcode = ERR41;
        goto FAILED;
        }
      if (!read_name(&ptr, ptrend, utf, CHAR_RIGHT_PARENTHESIS, &offset, &name,
          &namelen, &errorcode, cb)) goto FAILED;
      *parsed_pattern++ = META_BACKREF_BYNAME;
      *parsed_pattern++ = namelen;
      PUTOFFSET(offset, parsed_pattern);
      okquantifier = TRUE;
      break;   /* End of (?P processing */


      /* ---- Recursion/subroutine calls by number ---- */

      case CHAR_R:
      i = 0;         /* (?R) == (?R0) */
      ptr++;
      if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS)
        {
        errorcode = ERR58;
        goto FAILED;
        }
      goto SET_RECURSION;

      /* An item starting (?- followed by a digit comes here via the "default"
      case because (?- followed by a non-digit is an options setting. */

      case CHAR_PLUS:
      if (ptrend - ptr < 2 || !IS_DIGIT(ptr[1]))
        {
        errorcode = ERR29;   /* Missing number */
        goto FAILED;
        }
      /* Fall through */

      case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
      case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
      RECURSION_BYNUMBER:
      if (!read_number(&ptr, ptrend,
          (IS_DIGIT(*ptr))? -1:(int)(cb->bracount), /* + and - are relative */
          MAX_GROUP_NUMBER, ERR61,
          &i, &errorcode)) goto FAILED;
      if (i < 0)  /* NB (?0) is permitted */
        {
        errorcode = ERR15;   /* Unknown group */
        goto FAILED_BACK;
        }
      if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS)
        goto UNCLOSED_PARENTHESIS;

      SET_RECURSION:
      *parsed_pattern++ = META_RECURSE | (uint32_t)i;
      offset = (PCRE2_SIZE)(ptr - cb->start_pattern);
      ptr++;
      PUTOFFSET(offset, parsed_pattern);
      okquantifier = TRUE;
      break;  /* End of recursive call by number handling */


      /* ---- Recursion/subroutine calls by name ---- */

      case CHAR_AMPERSAND:
      RECURSE_BY_NAME:
      if (!read_name(&ptr, ptrend, utf, CHAR_RIGHT_PARENTHESIS, &offset, &name,
          &namelen, &errorcode, cb)) goto FAILED;
      *parsed_pattern++ = META_RECURSE_BYNAME;
      *parsed_pattern++ = namelen;
      PUTOFFSET(offset, parsed_pattern);
      okquantifier = TRUE;
      break;

      /* ---- Callout with numerical or string argument ---- */

      case CHAR_C:
      if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS;

      /* If the previous item was a condition starting (?(? an assertion,
      optionally preceded by a callout, is expected. This is checked later on,
      during actual compilation. However we need to identify this kind of
      assertion in this pass because it must not be qualified. The value of
      expect_cond_assert is set to 2 after (?(? is processed. We decrement it
      for a callout - still leaving a positive value that identifies the
      assertion. Multiple callouts or any other items will make it zero or
      less, which doesn't matter because they will cause an error later. */

      expect_cond_assert = prev_expect_cond_assert - 1;

      /* If previous_callout is not NULL, it means this follows a previous
      callout. If it was a manual callout, do nothing; this means its "length
      of next pattern item" field will remain zero. If it was an automatic
      callout, abolish it. */

      if (previous_callout != NULL && (options & PCRE2_AUTO_CALLOUT) != 0 &&
          previous_callout == parsed_pattern - 4 &&
          parsed_pattern[-1] == 255)
        parsed_pattern = previous_callout;

      /* Save for updating next pattern item length, and skip one item before
      completing. */

      previous_callout = parsed_pattern;
      after_manual_callout = 1;

      /* Handle a string argument; specific delimiter is required. */

      if (*ptr != CHAR_RIGHT_PARENTHESIS && !IS_DIGIT(*ptr))
        {
        PCRE2_SIZE calloutlength;
        PCRE2_SPTR startptr = ptr;

        delimiter = 0;
        for (i = 0; PRIV(callout_start_delims)[i] != 0; i++)
          {
          if (*ptr == PRIV(callout_start_delims)[i])
            {
            delimiter = PRIV(callout_end_delims)[i];
            break;
            }
          }
        if (delimiter == 0)
          {
          errorcode = ERR82;
          goto FAILED;
          }

        *parsed_pattern = META_CALLOUT_STRING;
        parsed_pattern += 3;   /* Skip pattern info */

        for (;;)
          {
          if (++ptr >= ptrend)
            {
            errorcode = ERR81;
            ptr = startptr;   /* To give a more useful message */
            goto FAILED;
            }
          if (*ptr == delimiter && (++ptr >= ptrend || *ptr != delimiter))
            break;
          }

        calloutlength = (PCRE2_SIZE)(ptr - startptr);
        if (calloutlength > UINT32_MAX)
          {
          errorcode = ERR72;
          goto FAILED;
          }
        *parsed_pattern++ = (uint32_t)calloutlength;
        offset = (PCRE2_SIZE)(startptr - cb->start_pattern);
        PUTOFFSET(offset, parsed_pattern);
        }

      /* Handle a callout with an optional numerical argument, which must be
      less than or equal to 255. A missing argument gives 0. */

      else
        {
        int n = 0;
        *parsed_pattern = META_CALLOUT_NUMBER;     /* Numerical callout */
        parsed_pattern += 3;                       /* Skip pattern info */
        while (ptr < ptrend && IS_DIGIT(*ptr))
          {
          n = n * 10 + *ptr++ - CHAR_0;
          if (n > 255)
            {
            errorcode = ERR38;
            goto FAILED;
            }
          }
        *parsed_pattern++ = n;
        }

      /* Both formats must have a closing parenthesis */

      if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS)
        {
        errorcode = ERR39;
        goto FAILED;
        }
      ptr++;

      /* Remember the offset to the next item in the pattern, and set a default
      length. This should get updated after the next item is read. */

      previous_callout[1] = (uint32_t)(ptr - cb->start_pattern);
      previous_callout[2] = 0;
      break;                  /* End callout */


      /* ---- Conditional group ---- */

      /* A condition can be an assertion, a number (referring to a numbered
      group's having been set), a name (referring to a named group), or 'R',
      referring to overall recursion. R<digits> and R&name are also permitted
      for recursion state tests. Numbers may be preceded by + or - to specify a
      relative group number.

      There are several syntaxes for testing a named group: (?(name)) is used
      by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).

      There are two unfortunate ambiguities. 'R' can be the recursive thing or
      the name 'R' (and similarly for 'R' followed by digits). 'DEFINE' can be
      the Perl DEFINE feature or the Python named test. We look for a name
      first; if not found, we try the other case.

      For compatibility with auto-callouts, we allow a callout to be specified
      before a condition that is an assertion. */

      case CHAR_LEFT_PARENTHESIS:
      if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS;
      nest_depth++;

      /* If the next character is ? or * there must be an assertion next
      (optionally preceded by a callout). We do not check this here, but
      instead we set expect_cond_assert to 2. If this is still greater than
      zero (callouts decrement it) when the next assertion is read, it will be
      marked as a condition that must not be repeated. A value greater than
      zero also causes checking that an assertion (possibly with callout)
      follows. */

      if (*ptr == CHAR_QUESTION_MARK || *ptr == CHAR_ASTERISK)
        {
        *parsed_pattern++ = META_COND_ASSERT;
        ptr--;   /* Pull pointer back to the opening parenthesis. */
        expect_cond_assert = 2;
        break;  /* End of conditional */
        }

      /* Handle (?([+-]number)... */

      if (read_number(&ptr, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &i,
          &errorcode))
        {
        if (i <= 0)
          {
          errorcode = ERR15;
          goto FAILED;
          }
        *parsed_pattern++ = META_COND_NUMBER;
        offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 2);
        PUTOFFSET(offset, parsed_pattern);
        *parsed_pattern++ = i;
        }
      else if (errorcode != 0) goto FAILED;   /* Number too big */

      /* No number found. Handle the special case (?(VERSION[>]=n.m)... */

      else if (ptrend - ptr >= 10 &&
               PRIV(strncmp_c8)(ptr, STRING_VERSION, 7) == 0 &&
               ptr[7] != CHAR_RIGHT_PARENTHESIS)
        {
        uint32_t ge = 0;
        int major = 0;
        int minor = 0;

        ptr += 7;
        if (*ptr == CHAR_GREATER_THAN_SIGN)
          {
          ge = 1;
          ptr++;
          }

        /* NOTE: cannot write IS_DIGIT(*(++ptr)) here because IS_DIGIT
        references its argument twice. */

        if (*ptr != CHAR_EQUALS_SIGN || (ptr++, !IS_DIGIT(*ptr)))
          goto BAD_VERSION_CONDITION;

        if (!read_number(&ptr, ptrend, -1, 1000, ERR79, &major, &errorcode))
          goto FAILED;

        if (ptr >= ptrend) goto BAD_VERSION_CONDITION;
        if (*ptr == CHAR_DOT)
          {
          if (++ptr >= ptrend || !IS_DIGIT(*ptr)) goto BAD_VERSION_CONDITION;
          minor = (*ptr++ - CHAR_0) * 10;
          if (IS_DIGIT(*ptr)) minor += *ptr++ - CHAR_0;
          if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS)
            goto BAD_VERSION_CONDITION;
          }

        *parsed_pattern++ = META_COND_VERSION;
        *parsed_pattern++ = ge;
        *parsed_pattern++ = major;
        *parsed_pattern++ = minor;
        }

      /* All the remaining cases now require us to read a name. We cannot at
      this stage distinguish ambiguous cases such as (?(R12) which might be a
      recursion test by number or a name, because the named groups have not yet
      all been identified. Those cases are treated as names, but given a
      different META code. */

      else
        {
        BOOL was_r_ampersand = FALSE;

        if (*ptr == CHAR_R && ptrend - ptr > 1 && ptr[1] == CHAR_AMPERSAND)
          {
          terminator = CHAR_RIGHT_PARENTHESIS;
          was_r_ampersand = TRUE;
          ptr++;
          }
        else if (*ptr == CHAR_LESS_THAN_SIGN)
          terminator = CHAR_GREATER_THAN_SIGN;
        else if (*ptr == CHAR_APOSTROPHE)
          terminator = CHAR_APOSTROPHE;
        else
          {
          terminator = CHAR_RIGHT_PARENTHESIS;
          ptr--;   /* Point to char before name */
          }
        if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen,
            &errorcode, cb)) goto FAILED;

        /* Handle (?(R&name) */

        if (was_r_ampersand)
          {
          *parsed_pattern = META_COND_RNAME;
          ptr--;   /* Back to closing parens */
          }

        /* Handle (?(name). If the name is "DEFINE" we identify it with a
        special code. Likewise if the name consists of R followed only by
        digits. Otherwise, handle it like a quoted name. */

        else if (terminator == CHAR_RIGHT_PARENTHESIS)
          {
          if (namelen == 6 && PRIV(strncmp_c8)(name, STRING_DEFINE, 6) == 0)
            *parsed_pattern = META_COND_DEFINE;
          else
            {
            for (i = 1; i < (int)namelen; i++)
              if (!IS_DIGIT(name[i])) break;
            *parsed_pattern = (*name == CHAR_R && i >= (int)namelen)?
              META_COND_RNUMBER : META_COND_NAME;
            }
          ptr--;   /* Back to closing parens */
          }

        /* Handle (?('name') or (?(<name>) */

        else *parsed_pattern = META_COND_NAME;

        /* All these cases except DEFINE end with the name length and offset;
        DEFINE just has an offset (for the "too many branches" error). */

        if (*parsed_pattern++ != META_COND_DEFINE) *parsed_pattern++ = namelen;
        PUTOFFSET(offset, parsed_pattern);
        }  /* End cases that read a name */

      /* Check the closing parenthesis of the condition */

      if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS)
        {
        errorcode = ERR24;
        goto FAILED;
        }
      ptr++;
      break;  /* End of condition processing */


      /* ---- Atomic group ---- */

      case CHAR_GREATER_THAN_SIGN:
      ATOMIC_GROUP:                          /* Come from (*atomic: */
      *parsed_pattern++ = META_ATOMIC;
      nest_depth++;
      ptr++;
      break;


      /* ---- Lookahead assertions ---- */

      case CHAR_EQUALS_SIGN:
      POSITIVE_LOOK_AHEAD:                   /* Come from (*pla: */
      *parsed_pattern++ = META_LOOKAHEAD;
      ptr++;
      goto POST_ASSERTION;

      case CHAR_EXCLAMATION_MARK:
      NEGATIVE_LOOK_AHEAD:                   /* Come from (*nla: */
      *parsed_pattern++ = META_LOOKAHEADNOT;
      ptr++;
      goto POST_ASSERTION;


      /* ---- Lookbehind assertions ---- */

      /* (?< followed by = or ! is a lookbehind assertion. Otherwise (?< is the
      start of the name of a capturing group. */

      case CHAR_LESS_THAN_SIGN:
      if (ptrend - ptr <= 1 ||
         (ptr[1] != CHAR_EQUALS_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK))
        {
        terminator = CHAR_GREATER_THAN_SIGN;
        goto DEFINE_NAME;
        }
      *parsed_pattern++ = (ptr[1] == CHAR_EQUALS_SIGN)?
        META_LOOKBEHIND : META_LOOKBEHINDNOT;

      POST_LOOKBEHIND:              /* Come from (*plb: and (*nlb: */
      *has_lookbehind = TRUE;
      offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 2);
      PUTOFFSET(offset, parsed_pattern);
      ptr += 2;
      /* Fall through */

      /* If the previous item was a condition starting (?(? an assertion,
      optionally preceded by a callout, is expected. This is checked later on,
      during actual compilation. However we need to identify this kind of
      assertion in this pass because it must not be qualified. The value of
      expect_cond_assert is set to 2 after (?(? is processed. We decrement it
      for a callout - still leaving a positive value that identifies the
      assertion. Multiple callouts or any other items will make it zero or
      less, which doesn't matter because they will cause an error later. */

      POST_ASSERTION:
      nest_depth++;
      if (prev_expect_cond_assert > 0)
        {
        if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace);
        else if (++top_nest >= end_nests)
          {
          errorcode = ERR84;
          goto FAILED;
          }
        top_nest->nest_depth = nest_depth;
        top_nest->flags = NSF_CONDASSERT;
        top_nest->options = options & PARSE_TRACKED_OPTIONS;
        }
      break;


      /* ---- Define a named group ---- */

      /* A named group may be defined as (?'name') or (?<name>). In the latter
      case we jump to DEFINE_NAME from the disambiguation of (?< above with the
      terminator set to '>'. */

      case CHAR_APOSTROPHE:
      terminator = CHAR_APOSTROPHE;    /* Terminator */

      DEFINE_NAME:
      if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen,
          &errorcode, cb)) goto FAILED;

      /* We have a name for this capturing group. It is also assigned a number,
      which is its primary means of identification. */

      cb->bracount++;
      *parsed_pattern++ = META_CAPTURE | cb->bracount;
      nest_depth++;

      /* Check not too many names */

      if (cb->names_found >= MAX_NAME_COUNT)
        {
        errorcode = ERR49;
        goto FAILED;
        }

      /* Adjust the entry size to accommodate the longest name found. */

      if (namelen + IMM2_SIZE + 1 > cb->name_entry_size)
        cb->name_entry_size = (uint16_t)(namelen + IMM2_SIZE + 1);

      /* Scan the list to check for duplicates. For duplicate names, if the
      number is the same, break the loop, which causes the name to be
      discarded; otherwise, if DUPNAMES is not set, give an error.
      If it is set, allow the name with a different number, but continue
      scanning in case this is a duplicate with the same number. For
      non-duplicate names, give an error if the number is duplicated. */

      isdupname = FALSE;
      ng = cb->named_groups;
      for (i = 0; i < cb->names_found; i++, ng++)
        {
        if (namelen == ng->length &&
            PRIV(strncmp)(name, ng->name, (PCRE2_SIZE)namelen) == 0)
          {
          if (ng->number == cb->bracount) break;
          if ((options & PCRE2_DUPNAMES) == 0)
            {
            errorcode = ERR43;
            goto FAILED;
            }
          isdupname = ng->isdup = TRUE;     /* Mark as a duplicate */
          cb->dupnames = TRUE;              /* Duplicate names exist */
          }
        else if (ng->number == cb->bracount)
          {
          errorcode = ERR65;
          goto FAILED;
          }
        }

      if (i < cb->names_found) break;   /* Ignore duplicate with same number */

      /* Increase the list size if necessary */

      if (cb->names_found >= cb->named_group_list_size)
        {
        uint32_t newsize = cb->named_group_list_size * 2;
        named_group *newspace =
          cb->cx->memctl.malloc(newsize * sizeof(named_group),
          cb->cx->memctl.memory_data);
        if (newspace == NULL)
          {
          errorcode = ERR21;
          goto FAILED;
          }

        memcpy(newspace, cb->named_groups,
          cb->named_group_list_size * sizeof(named_group));
        if (cb->named_group_list_size > NAMED_GROUP_LIST_SIZE)
          cb->cx->memctl.free((void *)cb->named_groups,
          cb->cx->memctl.memory_data);
        cb->named_groups = newspace;
        cb->named_group_list_size = newsize;
        }

      /* Add this name to the list */

      cb->named_groups[cb->names_found].name = name;
      cb->named_groups[cb->names_found].length = (uint16_t)namelen;
      cb->named_groups[cb->names_found].number = cb->bracount;
      cb->named_groups[cb->names_found].isdup = (uint16_t)isdupname;
      cb->names_found++;
      break;
      }        /* End of (? switch */
    break;     /* End of ( handling */


    /* ---- Branch terminators ---- */

    /* Alternation: reset the capture count if we are in a (?| group. */

    case CHAR_VERTICAL_LINE:
    if (top_nest != NULL && top_nest->nest_depth == nest_depth &&
        (top_nest->flags & NSF_RESET) != 0)
      {
      if (cb->bracount > top_nest->max_group)
        top_nest->max_group = (uint16_t)cb->bracount;
      cb->bracount = top_nest->reset_group;
      }
    *parsed_pattern++ = META_ALT;
    break;

    /* End of group; reset the capture count to the maximum if we are in a (?|
    group and/or reset the options that are tracked during parsing. Disallow
    quantifier for a condition that is an assertion. */

    case CHAR_RIGHT_PARENTHESIS:
    okquantifier = TRUE;
    if (top_nest != NULL && top_nest->nest_depth == nest_depth)
      {
      options = (options & ~PARSE_TRACKED_OPTIONS) | top_nest->options;
      if ((top_nest->flags & NSF_RESET) != 0 &&
          top_nest->max_group > cb->bracount)
        cb->bracount = top_nest->max_group;
      if ((top_nest->flags & NSF_CONDASSERT) != 0)
        okquantifier = FALSE;

      if ((top_nest->flags & NSF_ATOMICSR) != 0)
        {
        *parsed_pattern++ = META_KET;
        }



      if (top_nest == (nest_save *)(cb->start_workspace)) top_nest = NULL;
        else top_nest--;
      }
    if (nest_depth == 0)    /* Unmatched closing parenthesis */
      {
      errorcode = ERR22;
      goto FAILED_BACK;
      }
    nest_depth--;
    *parsed_pattern++ = META_KET;
    break;
    }  /* End of switch on pattern character */
  }    /* End of main character scan loop */

/* End of pattern reached. Check for missing ) at the end of a verb name. */

if (inverbname && ptr >= ptrend)
  {
  errorcode = ERR60;
  goto FAILED;
  }

/* Manage callout for the final item */

PARSED_END:
parsed_pattern = manage_callouts(ptr, &previous_callout, auto_callout,
  parsed_pattern, cb);

/* Insert trailing items for word and line matching (features provided for the
benefit of pcre2grep). */

if ((extra_options & PCRE2_EXTRA_MATCH_LINE) != 0)
  {
  *parsed_pattern++ = META_KET;
  *parsed_pattern++ = META_DOLLAR;
  }
else if ((extra_options & PCRE2_EXTRA_MATCH_WORD) != 0)
  {
  *parsed_pattern++ = META_KET;
  *parsed_pattern++ = META_ESCAPE + ESC_b;
  }

/* Terminate the parsed pattern, then return success if all groups are closed.
Otherwise we have unclosed parentheses. */

if (parsed_pattern >= parsed_pattern_end)
  {
  errorcode = ERR63;  /* Internal error (parsed pattern overflow) */
  goto FAILED;
  }

*parsed_pattern = META_END;
if (nest_depth == 0) return 0;

UNCLOSED_PARENTHESIS:
errorcode = ERR14;

/* Come here for all failures. */

FAILED:
cb->erroroffset = (PCRE2_SIZE)(ptr - cb->start_pattern);
return errorcode;

/* Some errors need to indicate the previous character. */

FAILED_BACK:
ptr--;
goto FAILED;

/* This failure happens several times. */

BAD_VERSION_CONDITION:
errorcode = ERR79;
goto FAILED;
}



/*************************************************
*       Find first significant opcode            *
*************************************************/

/* This is called by several functions that scan a compiled expression looking
for a fixed first character, or an anchoring opcode etc. It skips over things
that do not influence this. For some calls, it makes sense to skip negative
forward and all backward assertions, and also the \b assertion; for others it
does not.

Arguments:
  code         pointer to the start of the group
  skipassert   TRUE if certain assertions are to be skipped

Returns:       pointer to the first significant opcode
*/

static const PCRE2_UCHAR*
first_significant_code(PCRE2_SPTR code, BOOL skipassert)
{
for (;;)
  {
  switch ((int)*code)
    {
    case OP_ASSERT_NOT:
    case OP_ASSERTBACK:
    case OP_ASSERTBACK_NOT:
    if (!skipassert) return code;
    do code += GET(code, 1); while (*code == OP_ALT);
    code += PRIV(OP_lengths)[*code];
    break;

    case OP_WORD_BOUNDARY:
    case OP_NOT_WORD_BOUNDARY:
    if (!skipassert) return code;
    /* Fall through */

    case OP_CALLOUT:
    case OP_CREF:
    case OP_DNCREF:
    case OP_RREF:
    case OP_DNRREF:
    case OP_FALSE:
    case OP_TRUE:
    code += PRIV(OP_lengths)[*code];
    break;

    case OP_CALLOUT_STR:
    code += GET(code, 1 + 2*LINK_SIZE);
    break;

    case OP_SKIPZERO:
    code += 2 + GET(code, 2) + LINK_SIZE;
    break;

    case OP_COND:
    case OP_SCOND:
    if (code[1+LINK_SIZE] != OP_FALSE ||   /* Not DEFINE */
        code[GET(code, 1)] != OP_KET)      /* More than one branch */
      return code;
    code += GET(code, 1) + 1 + LINK_SIZE;
    break;

    case OP_MARK:
    case OP_COMMIT_ARG:
    case OP_PRUNE_ARG:
    case OP_SKIP_ARG:
    case OP_THEN_ARG:
    code += code[1] + PRIV(OP_lengths)[*code];
    break;

    default:
    return code;
    }
  }
/* Control never reaches here */
}



#ifdef SUPPORT_UNICODE
/*************************************************
*           Get othercase range                  *
*************************************************/

/* This function is passed the start and end of a class range in UCP mode. It
searches up the characters, looking for ranges of characters in the "other"
case. Each call returns the next one, updating the start address. A character
with multiple other cases is returned on its own with a special return value.

Arguments:
  cptr        points to starting character value; updated
  d           end value
  ocptr       where to put start of othercase range
  odptr       where to put end of othercase range

Yield:        -1 when no more
               0 when a range is returned
              >0 the CASESET offset for char with multiple other cases
                in this case, ocptr contains the original
*/

static int
get_othercase_range(uint32_t *cptr, uint32_t d, uint32_t *ocptr,
  uint32_t *odptr)
{
uint32_t c, othercase, next;
unsigned int co;

/* Find the first character that has an other case. If it has multiple other
cases, return its case offset value. */

for (c = *cptr; c <= d; c++)
  {
  if ((co = UCD_CASESET(c)) != 0)
    {
    *ocptr = c++;   /* Character that has the set */
    *cptr = c;      /* Rest of input range */
    return (int)co;
    }
  if ((othercase = UCD_OTHERCASE(c)) != c) break;
  }

if (c > d) return -1;  /* Reached end of range */

/* Found a character that has a single other case. Search for the end of the
range, which is either the end of the input range, or a character that has zero
or more than one other cases. */

*ocptr = othercase;
next = othercase + 1;

for (++c; c <= d; c++)
  {
  if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
  next++;
  }

*odptr = next - 1;     /* End of othercase range */
*cptr = c;             /* Rest of input range */
return 0;
}
#endif  /* SUPPORT_UNICODE */



/*************************************************
* Add a character or range to a class (internal) *
*************************************************/

/* This function packages up the logic of adding a character or range of
characters to a class. The character values in the arguments will be within the
valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
called only from within the "add to class" group of functions, some of which
are recursive and mutually recursive. The external entry point is
add_to_class().

Arguments:
  classbits     the bit map for characters < 256
  uchardptr     points to the pointer for extra data
  options       the options word
  cb            compile data
  start         start of range character
  end           end of range character

Returns:        the number of < 256 characters added
                the pointer to extra data is updated
*/

static unsigned int
add_to_class_internal(uint8_t *classbits, PCRE2_UCHAR **uchardptr,
  uint32_t options, compile_block *cb, uint32_t start, uint32_t end)
{
uint32_t c;
uint32_t classbits_end = (end <= 0xff ? end : 0xff);
unsigned int n8 = 0;

/* If caseless matching is required, scan the range and process alternate
cases. In Unicode, there are 8-bit characters that have alternate cases that
are greater than 255 and vice-versa. Sometimes we can just extend the original
range. */

if ((options & PCRE2_CASELESS) != 0)
  {
#ifdef SUPPORT_UNICODE
  if ((options & PCRE2_UTF) != 0)
    {
    int rc;
    uint32_t oc, od;

    options &= ~PCRE2_CASELESS;   /* Remove for recursive calls */
    c = start;

    while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
      {
      /* Handle a single character that has more than one other case. */

      if (rc > 0) n8 += add_list_to_class_internal(classbits, uchardptr, options, cb,
        PRIV(ucd_caseless_sets) + rc, oc);

      /* Do nothing if the other case range is within the original range. */

      else if (oc >= cb->class_range_start && od <= cb->class_range_end) continue;

      /* Extend the original range if there is overlap, noting that if oc < c, we
      can't have od > end because a subrange is always shorter than the basic
      range. Otherwise, use a recursive call to add the additional range. */

      else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
      else if (od > end && oc <= end + 1)
        {
        end = od;       /* Extend upwards */
        if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
        }
      else n8 += add_to_class_internal(classbits, uchardptr, options, cb, oc, od);
      }
    }
  else
#endif  /* SUPPORT_UNICODE */

  /* Not UTF mode */

  for (c = start; c <= classbits_end; c++)
    {
    SETBIT(classbits, cb->fcc[c]);
    n8++;
    }
  }

/* Now handle the originally supplied range. Adjust the final value according
to the bit length - this means that the same lists of (e.g.) horizontal spaces
can be used in all cases. */

if ((options & PCRE2_UTF) == 0 && end > MAX_NON_UTF_CHAR)
  end = MAX_NON_UTF_CHAR;

if (start > cb->class_range_start && end < cb->class_range_end) return n8;

/* Use the bitmap for characters < 256. Otherwise use extra data.*/

for (c = start; c <= classbits_end; c++)
  {
  /* Regardless of start, c will always be <= 255. */
  SETBIT(classbits, c);
  n8++;
  }

#ifdef SUPPORT_WIDE_CHARS
if (start <= 0xff) start = 0xff + 1;

if (end >= start)
  {
  PCRE2_UCHAR *uchardata = *uchardptr;

#ifdef SUPPORT_UNICODE
  if ((options & PCRE2_UTF) != 0)
    {
    if (start < end)
      {
      *uchardata++ = XCL_RANGE;
      uchardata += PRIV(ord2utf)(start, uchardata);
      uchardata += PRIV(ord2utf)(end, uchardata);
      }
    else if (start == end)
      {
      *uchardata++ = XCL_SINGLE;
      uchardata += PRIV(ord2utf)(start, uchardata);
      }
    }
  else
#endif  /* SUPPORT_UNICODE */

  /* Without UTF support, character values are constrained by the bit length,
  and can only be > 256 for 16-bit and 32-bit libraries. */

#if PCRE2_CODE_UNIT_WIDTH == 8
    {}
#else
  if (start < end)
    {
    *uchardata++ = XCL_RANGE;
    *uchardata++ = start;
    *uchardata++ = end;
    }
  else if (start == end)
    {
    *uchardata++ = XCL_SINGLE;
    *uchardata++ = start;
    }
#endif  /* PCRE2_CODE_UNIT_WIDTH == 8 */
  *uchardptr = uchardata;   /* Updata extra data pointer */
  }
#else  /* SUPPORT_WIDE_CHARS */
  (void)uchardptr;          /* Avoid compiler warning */
#endif /* SUPPORT_WIDE_CHARS */

return n8;    /* Number of 8-bit characters */
}



#ifdef SUPPORT_UNICODE
/*************************************************
* Add a list of characters to a class (internal) *
*************************************************/

/* This function is used for adding a list of case-equivalent characters to a
class when in UTF mode. This function is called only from within
add_to_class_internal(), with which it is mutually recursive.

Arguments:
  classbits     the bit map for characters < 256
  uchardptr     points to the pointer for extra data
  options       the options word
  cb            contains pointers to tables etc.
  p             points to row of 32-bit values, terminated by NOTACHAR
  except        character to omit; this is used when adding lists of
                  case-equivalent characters to avoid including the one we
                  already know about

Returns:        the number of < 256 characters added
                the pointer to extra data is updated
*/

static unsigned int
add_list_to_class_internal(uint8_t *classbits, PCRE2_UCHAR **uchardptr,
  uint32_t options, compile_block *cb, const uint32_t *p, unsigned int except)
{
unsigned int n8 = 0;
while (p[0] < NOTACHAR)
  {
  unsigned int n = 0;
  if (p[0] != except)
    {
    while(p[n+1] == p[0] + n + 1) n++;
    n8 += add_to_class_internal(classbits, uchardptr, options, cb, p[0], p[n]);
    }
  p += n + 1;
  }
return n8;
}
#endif



/*************************************************
*   External entry point for add range to class  *
*************************************************/

/* This function sets the overall range so that the internal functions can try
to avoid duplication when handling case-independence.

Arguments:
  classbits     the bit map for characters < 256
  uchardptr     points to the pointer for extra data
  options       the options word
  cb            compile data
  start         start of range character
  end           end of range character

Returns:        the number of < 256 characters added
                the pointer to extra data is updated
*/

static unsigned int
add_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr, uint32_t options,
  compile_block *cb, uint32_t start, uint32_t end)
{
cb->class_range_start = start;
cb->class_range_end = end;
return add_to_class_internal(classbits, uchardptr, options, cb, start, end);
}


/*************************************************
*   External entry point for add list to class   *
*************************************************/

/* This function is used for adding a list of horizontal or vertical whitespace
characters to a class. The list must be in order so that ranges of characters
can be detected and handled appropriately. This function sets the overall range
so that the internal functions can try to avoid duplication when handling
case-independence.

Arguments:
  classbits     the bit map for characters < 256
  uchardptr     points to the pointer for extra data
  options       the options word
  cb            contains pointers to tables etc.
  p             points to row of 32-bit values, terminated by NOTACHAR
  except        character to omit; this is used when adding lists of
                  case-equivalent characters to avoid including the one we
                  already know about

Returns:        the number of < 256 characters added
                the pointer to extra data is updated
*/

static unsigned int
add_list_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr, uint32_t options,
  compile_block *cb, const uint32_t *p, unsigned int except)
{
unsigned int n8 = 0;
while (p[0] < NOTACHAR)
  {
  unsigned int n = 0;
  if (p[0] != except)
    {
    while(p[n+1] == p[0] + n + 1) n++;
    cb->class_range_start = p[0];
    cb->class_range_end = p[n];
    n8 += add_to_class_internal(classbits, uchardptr, options, cb, p[0], p[n]);
    }
  p += n + 1;
  }
return n8;
}



/*************************************************
*    Add characters not in a list to a class     *
*************************************************/

/* This function is used for adding the complement of a list of horizontal or
vertical whitespace to a class. The list must be in order.

Arguments:
  classbits     the bit map for characters < 256
  uchardptr     points to the pointer for extra data
  options       the options word
  cb            contains pointers to tables etc.
  p             points to row of 32-bit values, terminated by NOTACHAR

Returns:        the number of < 256 characters added
                the pointer to extra data is updated
*/

static unsigned int
add_not_list_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr,
  uint32_t options, compile_block *cb, const uint32_t *p)
{
BOOL utf = (options & PCRE2_UTF) != 0;
unsigned int n8 = 0;
if (p[0] > 0)
  n8 += add_to_class(classbits, uchardptr, options, cb, 0, p[0] - 1);
while (p[0] < NOTACHAR)
  {
  while (p[1] == p[0] + 1) p++;
  n8 += add_to_class(classbits, uchardptr, options, cb, p[0] + 1,
    (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
  p++;
  }
return n8;
}



/*************************************************
*    Find details of duplicate group names       *
*************************************************/

/* This is called from compile_branch() when it needs to know the index and
count of duplicates in the names table when processing named backreferences,
either directly, or as conditions.

Arguments:
  name          points to the name
  length        the length of the name
  indexptr      where to put the index
  countptr      where to put the count of duplicates
  errorcodeptr  where to put an error code
  cb            the compile block

Returns:        TRUE if OK, FALSE if not, error code set
*/

static BOOL
find_dupname_details(PCRE2_SPTR name, uint32_t length, int *indexptr,
  int *countptr, int *errorcodeptr, compile_block *cb)
{
uint32_t i, groupnumber;
int count;
PCRE2_UCHAR *slot = cb->name_table;

/* Find the first entry in the table */

for (i = 0; i < cb->names_found; i++)
  {
  if (PRIV(strncmp)(name, slot+IMM2_SIZE, length) == 0 &&
      slot[IMM2_SIZE+length] == 0) break;
  slot += cb->name_entry_size;
  }

/* This should not occur, because this function is called only when we know we
have duplicate names. Give an internal error. */

if (i >= cb->names_found)
  {
  *errorcodeptr = ERR53;
  cb->erroroffset = name - cb->start_pattern;
  return FALSE;
  }

/* Record the index and then see how many duplicates there are, updating the
backref map and maximum back reference as we do. */

*indexptr = i;
count = 0;

for (;;)
  {
  count++;
  groupnumber = GET2(slot,0);
  cb->backref_map |= (groupnumber < 32)? (1u << groupnumber) : 1;
  if (groupnumber > cb->top_backref) cb->top_backref = groupnumber;
  if (++i >= cb->names_found) break;
  slot += cb->name_entry_size;
  if (PRIV(strncmp)(name, slot+IMM2_SIZE, length) != 0 ||
    (slot+IMM2_SIZE)[length] != 0) break;
  }

*countptr = count;
return TRUE;
}



/*************************************************
*           Compile one branch                   *
*************************************************/

/* Scan the parsed pattern, compiling it into the a vector of PCRE2_UCHAR. If
the options are changed during the branch, the pointer is used to change the
external options bits. This function is used during the pre-compile phase when
we are trying to find out the amount of memory needed, as well as during the
real compile phase. The value of lengthptr distinguishes the two phases.

Arguments:
  optionsptr        pointer to the option bits
  codeptr           points to the pointer to the current code point
  pptrptr           points to the current parsed pattern pointer
  errorcodeptr      points to error code variable
  firstcuptr        place to put the first required code unit
  firstcuflagsptr   place to put the first code unit flags, or a negative number
  reqcuptr          place to put the last required code unit
  reqcuflagsptr     place to put the last required code unit flags, or a negative number
  bcptr             points to current branch chain
  cb                contains pointers to tables etc.
  lengthptr         NULL during the real compile phase
                    points to length accumulator during pre-compile phase

Returns:            0 There's been an error, *errorcodeptr is non-zero
                   +1 Success, this branch must match at least one character
                   -1 Success, this branch may match an empty string
*/

static int
compile_branch(uint32_t *optionsptr, PCRE2_UCHAR **codeptr, uint32_t **pptrptr,
  int *errorcodeptr, uint32_t *firstcuptr, int32_t *firstcuflagsptr,
  uint32_t *reqcuptr, int32_t *reqcuflagsptr, branch_chain *bcptr,
  compile_block *cb, PCRE2_SIZE *lengthptr)
{
int bravalue = 0;
int okreturn = -1;
int group_return = 0;
uint32_t repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
uint32_t greedy_default, greedy_non_default;
uint32_t repeat_type, op_type;
uint32_t options = *optionsptr;               /* May change dynamically */
uint32_t firstcu, reqcu;
uint32_t zeroreqcu, zerofirstcu;
uint32_t escape;
uint32_t *pptr = *pptrptr;
uint32_t meta, meta_arg;
int32_t firstcuflags, reqcuflags;
int32_t zeroreqcuflags, zerofirstcuflags;
int32_t req_caseopt, reqvary, tempreqvary;
PCRE2_SIZE offset = 0;
PCRE2_SIZE length_prevgroup = 0;
PCRE2_UCHAR *code = *codeptr;
PCRE2_UCHAR *last_code = code;
PCRE2_UCHAR *orig_code = code;
PCRE2_UCHAR *tempcode;
PCRE2_UCHAR *previous = NULL;
PCRE2_UCHAR op_previous;
BOOL groupsetfirstcu = FALSE;
BOOL matched_char = FALSE;
BOOL previous_matched_char = FALSE;
const uint8_t *cbits = cb->cbits;
uint8_t classbits[32];

/* We can fish out the UTF setting once and for all into a BOOL, but we must
not do this for other options (e.g. PCRE2_EXTENDED) because they may change
dynamically as we process the pattern. */

#ifdef SUPPORT_UNICODE
BOOL utf = (options & PCRE2_UTF) != 0;
#else  /* No UTF support */
BOOL utf = FALSE;
#endif

/* Helper variables for OP_XCLASS opcode (for characters > 255). We define
class_uchardata always so that it can be passed to add_to_class() always,
though it will not be used in non-UTF 8-bit cases. This avoids having to supply
alternative calls for the different cases. */

PCRE2_UCHAR *class_uchardata;
#ifdef SUPPORT_WIDE_CHARS
BOOL xclass;
PCRE2_UCHAR *class_uchardata_base;
#endif

/* Set up the default and non-default settings for greediness */

greedy_default = ((options & PCRE2_UNGREEDY) != 0);
greedy_non_default = greedy_default ^ 1;

/* Initialize no first unit, no required unit. REQ_UNSET means "no char
matching encountered yet". It gets changed to REQ_NONE if we hit something that
matches a non-fixed first unit; reqcu just remains unset if we never find one.

When we hit a repeat whose minimum is zero, we may have to adjust these values
to take the zero repeat into account. This is implemented by setting them to
zerofirstcu and zeroreqcu when such a repeat is encountered. The individual
item types that can be repeated set these backoff variables appropriately. */

firstcu = reqcu = zerofirstcu = zeroreqcu = 0;
firstcuflags = reqcuflags = zerofirstcuflags = zeroreqcuflags = REQ_UNSET;

/* The variable req_caseopt contains either the REQ_CASELESS value or zero,
according to the current setting of the caseless flag. The REQ_CASELESS value
leaves the lower 28 bit empty. It is added into the firstcu or reqcu variables
to record the case status of the value. This is used only for ASCII characters.
*/

req_caseopt = ((options & PCRE2_CASELESS) != 0)? REQ_CASELESS:0;

/* Switch on next META item until the end of the branch */

for (;; pptr++)
  {
#ifdef SUPPORT_WIDE_CHARS
  BOOL xclass_has_prop;
#endif
  BOOL negate_class;
  BOOL should_flip_negation;
  BOOL match_all_or_no_wide_chars;
  BOOL possessive_quantifier;
  BOOL note_group_empty;
  int class_has_8bitchar;
  int i;
  uint32_t mclength;
  uint32_t skipunits;
  uint32_t subreqcu, subfirstcu;
  uint32_t groupnumber;
  uint32_t verbarglen, verbculen;
  int32_t subreqcuflags, subfirstcuflags;  /* Must be signed */
  open_capitem *oc;
  PCRE2_UCHAR mcbuffer[8];

  /* Get next META item in the pattern and its potential argument. */

  meta = META_CODE(*pptr);
  meta_arg = META_DATA(*pptr);

  /* If we are in the pre-compile phase, accumulate the length used for the
  previous cycle of this loop, unless the next item is a quantifier. */

  if (lengthptr != NULL)
    {
    if (code > cb->start_workspace + cb->workspace_size -
        WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
      {
      *errorcodeptr = (code >= cb->start_workspace + cb->workspace_size)?
        ERR52 : ERR86;
      return 0;
      }

    /* There is at least one situation where code goes backwards: this is the
    case of a zero quantifier after a class (e.g. [ab]{0}). When the quantifier
    is processed, the whole class is eliminated. However, it is created first,
    so we have to allow memory for it. Therefore, don't ever reduce the length
    at this point. */

    if (code < last_code) code = last_code;

    /* If the next thing is not a quantifier, we add the length of the previous
    item into the total, and reset the code pointer to the start of the
    workspace. Otherwise leave the previous item available to be quantified. */

    if (meta < META_ASTERISK || meta > META_MINMAX_QUERY)
      {
      if (OFLOW_MAX - *lengthptr < (PCRE2_SIZE)(code - orig_code))
        {
        *errorcodeptr = ERR20;   /* Integer overflow */
        return 0;
        }
      *lengthptr += (PCRE2_SIZE)(code - orig_code);
      if (*lengthptr > MAX_PATTERN_SIZE)
        {
        *errorcodeptr = ERR20;   /* Pattern is too large */
        return 0;
        }
      code = orig_code;
      }

    /* Remember where this code item starts so we can catch the "backwards"
    case above next time round. */

    last_code = code;
    }

  /* Process the next parsed pattern item. If it is not a quantifier, remember
  where it starts so that it can be quantified when a quantifier follows.
  Checking for the legality of quantifiers happens in parse_regex(), except for
  a quantifier after an assertion that is a condition. */

  if (meta < META_ASTERISK || meta > META_MINMAX_QUERY)
    {
    previous = code;
    if (matched_char) okreturn = 1;
    }

  previous_matched_char = matched_char;
  matched_char = FALSE;
  note_group_empty = FALSE;
  skipunits = 0;         /* Default value for most subgroups */

  switch(meta)
    {
    /* ===================================================================*/
    /* The branch terminates at pattern end or | or ) */

    case META_END:
    case META_ALT:
    case META_KET:
    *firstcuptr = firstcu;
    *firstcuflagsptr = firstcuflags;
    *reqcuptr = reqcu;
    *reqcuflagsptr = reqcuflags;
    *codeptr = code;
    *pptrptr = pptr;
    return okreturn;


    /* ===================================================================*/
    /* Handle single-character metacharacters. In multiline mode, ^ disables
    the setting of any following char as a first character. */

    case META_CIRCUMFLEX:
    if ((options & PCRE2_MULTILINE) != 0)
      {
      if (firstcuflags == REQ_UNSET)
        zerofirstcuflags = firstcuflags = REQ_NONE;
      *code++ = OP_CIRCM;
      }
    else *code++ = OP_CIRC;
    break;

    case META_DOLLAR:
    *code++ = ((options & PCRE2_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
    break;

    /* There can never be a first char if '.' is first, whatever happens about
    repeats. The value of reqcu doesn't change either. */

    case META_DOT:
    matched_char = TRUE;
    if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    zeroreqcu = reqcu;
    zeroreqcuflags = reqcuflags;
    *code++ = ((options & PCRE2_DOTALL) != 0)? OP_ALLANY: OP_ANY;
    break;


    /* ===================================================================*/
    /* Empty character classes are allowed if PCRE2_ALLOW_EMPTY_CLASS is set.
    Otherwise, an initial ']' is taken as a data character. When empty classes
    are allowed, [] must always fail, so generate OP_FAIL, whereas [^] must
    match any character, so generate OP_ALLANY. */

    case META_CLASS_EMPTY:
    case META_CLASS_EMPTY_NOT:
    matched_char = TRUE;
    *code++ = (meta == META_CLASS_EMPTY_NOT)? OP_ALLANY : OP_FAIL;
    if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    break;


    /* ===================================================================*/
    /* Non-empty character class. If the included characters are all < 256, we
    build a 32-byte bitmap of the permitted characters, except in the special
    case where there is only one such character. For negated classes, we build
    the map as usual, then invert it at the end. However, we use a different
    opcode so that data characters > 255 can be handled correctly.

    If the class contains characters outside the 0-255 range, a different
    opcode is compiled. It may optionally have a bit map for characters < 256,
    but those above are are explicitly listed afterwards. A flag code unit
    tells whether the bitmap is present, and whether this is a negated class or
    not. */

    case META_CLASS_NOT:
    case META_CLASS:
    matched_char = TRUE;
    negate_class = meta == META_CLASS_NOT;

    /* We can optimize the case of a single character in a class by generating
    OP_CHAR or OP_CHARI if it's positive, or OP_NOT or OP_NOTI if it's
    negative. In the negative case there can be no first char if this item is
    first, whatever repeat count may follow. In the case of reqcu, save the
    previous value for reinstating. */

    /* NOTE: at present this optimization is not effective if the only
    character in a class in 32-bit, non-UCP mode has its top bit set. */

    if (pptr[1] < META_END && pptr[2] == META_CLASS_END)
      {
#ifdef SUPPORT_UNICODE
      uint32_t d;
#endif
      uint32_t c = pptr[1];

      pptr += 2;                 /* Move on to class end */
      if (meta == META_CLASS)    /* A positive one-char class can be */
        {                        /* handled as a normal literal character. */
        meta = c;                /* Set up the character */
        goto NORMAL_CHAR_SET;
        }

      /* Handle a negative one-character class */

      zeroreqcu = reqcu;
      zeroreqcuflags = reqcuflags;
      if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
      zerofirstcu = firstcu;
      zerofirstcuflags = firstcuflags;

      /* For caseless UTF mode, check whether this character has more than
      one other case. If so, generate a special OP_NOTPROP item instead of
      OP_NOTI. */

#ifdef SUPPORT_UNICODE
      if (utf && (options & PCRE2_CASELESS) != 0 &&
          (d = UCD_CASESET(c)) != 0)
        {
        *code++ = OP_NOTPROP;
        *code++ = PT_CLIST;
        *code++ = d;
        break;   /* We are finished with this class */
        }
#endif
      /* Char has only one other case, or UCP not available */

      *code++ = ((options & PCRE2_CASELESS) != 0)? OP_NOTI: OP_NOT;
      code += PUTCHAR(c, code);
      break;   /* We are finished with this class */
      }        /* End of 1-char optimization */

    /* Handle character classes that contain more than just one literal
    character. */

    /* If a non-extended class contains a negative special such as \S, we need
    to flip the negation flag at the end, so that support for characters > 255
    works correctly (they are all included in the class). An extended class may
    need to insert specific matching or non-matching code for wide characters.
    */

    should_flip_negation = match_all_or_no_wide_chars = FALSE;

    /* Extended class (xclass) will be used when characters > 255
    might match. */

#ifdef SUPPORT_WIDE_CHARS
    xclass = FALSE;
    class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
    class_uchardata_base = class_uchardata;   /* Save the start */
#endif

    /* For optimization purposes, we track some properties of the class:
    class_has_8bitchar will be non-zero if the class contains at least one
    character with a code point less than 256; xclass_has_prop will be TRUE if
    Unicode property checks are present in the class. */

    class_has_8bitchar = 0;
#ifdef SUPPORT_WIDE_CHARS
    xclass_has_prop = FALSE;
#endif

    /* Initialize the 256-bit (32-byte) bit map to all zeros. We build the map
    in a temporary bit of memory, in case the class contains fewer than two
    8-bit characters because in that case the compiled code doesn't use the bit
    map. */

    memset(classbits, 0, 32 * sizeof(uint8_t));

    /* Process items until META_CLASS_END is reached. */

    while ((meta = *(++pptr)) != META_CLASS_END)
      {
      /* Handle POSIX classes such as [:alpha:] etc. */

      if (meta == META_POSIX || meta == META_POSIX_NEG)
        {
        BOOL local_negate = (meta == META_POSIX_NEG);
        int posix_class = *(++pptr);
        int taboffset, tabopt;
        uint8_t pbits[32];

        should_flip_negation = local_negate;  /* Note negative special */

        /* If matching is caseless, upper and lower are converted to alpha.
        This relies on the fact that the class table starts with alpha,
        lower, upper as the first 3 entries. */

        if ((options & PCRE2_CASELESS) != 0 && posix_class <= 2)
          posix_class = 0;

        /* When PCRE2_UCP is set, some of the POSIX classes are converted to
        different escape sequences that use Unicode properties \p or \P.
        Others that are not available via \p or \P have to generate
        XCL_PROP/XCL_NOTPROP directly, which is done here. */

#ifdef SUPPORT_UNICODE
        if ((options & PCRE2_UCP) != 0) switch(posix_class)
          {
          case PC_GRAPH:
          case PC_PRINT:
          case PC_PUNCT:
          *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
          *class_uchardata++ = (PCRE2_UCHAR)
            ((posix_class == PC_GRAPH)? PT_PXGRAPH :
             (posix_class == PC_PRINT)? PT_PXPRINT : PT_PXPUNCT);
          *class_uchardata++ = 0;
          xclass_has_prop = TRUE;
          goto CONTINUE_CLASS;

          /* For the other POSIX classes (ascii, xdigit) we are going to
          fall through to the non-UCP case and build a bit map for
          characters with code points less than 256. However, if we are in
          a negated POSIX class, characters with code points greater than
          255 must either all match or all not match, depending on whether
          the whole class is not or is negated. For example, for
          [[:^ascii:]... they must all match, whereas for [^[:^xdigit:]...
          they must not.

          In the special case where there are no xclass items, this is
          automatically handled by the use of OP_CLASS or OP_NCLASS, but an
          explicit range is needed for OP_XCLASS. Setting a flag here
          causes the range to be generated later when it is known that
          OP_XCLASS is required. In the 8-bit library this is relevant only in
          utf mode, since no wide characters can exist otherwise. */

          default:
#if PCRE2_CODE_UNIT_WIDTH == 8
          if (utf)
#endif
          match_all_or_no_wide_chars |= local_negate;
          break;
          }
#endif  /* SUPPORT_UNICODE */

        /* In the non-UCP case, or when UCP makes no difference, we build the
        bit map for the POSIX class in a chunk of local store because we may
        be adding and subtracting from it, and we don't want to subtract bits
        that may be in the main map already. At the end we or the result into
        the bit map that is being built. */

        posix_class *= 3;

        /* Copy in the first table (always present) */

        memcpy(pbits, cbits + posix_class_maps[posix_class],
          32 * sizeof(uint8_t));

        /* If there is a second table, add or remove it as required. */

        taboffset = posix_class_maps[posix_class + 1];
        tabopt = posix_class_maps[posix_class + 2];

        if (taboffset >= 0)
          {
          if (tabopt >= 0)
            for (i = 0; i < 32; i++) pbits[i] |= cbits[(int)i + taboffset];
          else
            for (i = 0; i < 32; i++) pbits[i] &= ~cbits[(int)i + taboffset];
          }

        /* Now see if we need to remove any special characters. An option
        value of 1 removes vertical space and 2 removes underscore. */

        if (tabopt < 0) tabopt = -tabopt;
        if (tabopt == 1) pbits[1] &= ~0x3c;
          else if (tabopt == 2) pbits[11] &= 0x7f;

        /* Add the POSIX table or its complement into the main table that is
        being built and we are done. */

        if (local_negate)
          for (i = 0; i < 32; i++) classbits[i] |= ~pbits[i];
        else
          for (i = 0; i < 32; i++) classbits[i] |= pbits[i];

        /* Every class contains at least one < 256 character. */

        class_has_8bitchar = 1;
        goto CONTINUE_CLASS;    /* End of POSIX handling */
        }

      /* Other than POSIX classes, the only items we should encounter are
      \d-type escapes and literal characters (possibly as ranges). */

      if (meta == META_BIGVALUE)
        {
        meta = *(++pptr);
        goto CLASS_LITERAL;
        }

      /* Any other non-literal must be an escape */

      if (meta >= META_END)
        {
        if (META_CODE(meta) != META_ESCAPE)
          {
#ifdef DEBUG_SHOW_PARSED
          fprintf(stderr, "** Unrecognized parsed pattern item 0x%.8x "
                          "in character class\n", meta);
#endif
          *errorcodeptr = ERR89;  /* Internal error - unrecognized. */
          return 0;
          }
        escape = META_DATA(meta);

        /* Every class contains at least one < 256 character. */

        class_has_8bitchar++;

        switch(escape)
          {
          case ESC_d:
          for (i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_digit];
          break;

          case ESC_D:
          should_flip_negation = TRUE;
          for (i = 0; i < 32; i++) classbits[i] |= ~cbits[i+cbit_digit];
          break;

          case ESC_w:
          for (i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_word];
          break;

          case ESC_W:
          should_flip_negation = TRUE;
          for (i = 0; i < 32; i++) classbits[i] |= ~cbits[i+cbit_word];
          break;

          /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
          5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
          previously set by something earlier in the character class.
          Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
          we could just adjust the appropriate bit. From PCRE 8.34 we no
          longer treat \s and \S specially. */

          case ESC_s:
          for (i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_space];
          break;

          case ESC_S:
          should_flip_negation = TRUE;
          for (i = 0; i < 32; i++) classbits[i] |= ~cbits[i+cbit_space];
          break;

          /* When adding the horizontal or vertical space lists to a class, or
          their complements, disable PCRE2_CASELESS, because it justs wastes
          time, and in the "not-x" UTF cases can create unwanted duplicates in
          the XCLASS list (provoked by characters that have more than one other
          case and by both cases being in the same "not-x" sublist). */

          case ESC_h:
          (void)add_list_to_class(classbits, &class_uchardata,
            options & ~PCRE2_CASELESS, cb, PRIV(hspace_list), NOTACHAR);
          break;

          case ESC_H:
          (void)add_not_list_to_class(classbits, &class_uchardata,
            options & ~PCRE2_CASELESS, cb, PRIV(hspace_list));
          break;

          case ESC_v:
          (void)add_list_to_class(classbits, &class_uchardata,
            options & ~PCRE2_CASELESS, cb, PRIV(vspace_list), NOTACHAR);
          break;

          case ESC_V:
          (void)add_not_list_to_class(classbits, &class_uchardata,
            options & ~PCRE2_CASELESS, cb, PRIV(vspace_list));
          break;

          /* If Unicode is not supported, \P and \p are not allowed and are
          faulted at parse time, so will never appear here. */

#ifdef SUPPORT_UNICODE
          case ESC_p:
          case ESC_P:
            {
            uint32_t ptype = *(++pptr) >> 16;
            uint32_t pdata = *pptr & 0xffff;
            *class_uchardata++ = (escape == ESC_p)? XCL_PROP : XCL_NOTPROP;
            *class_uchardata++ = ptype;
            *class_uchardata++ = pdata;
            xclass_has_prop = TRUE;
            class_has_8bitchar--;                /* Undo! */
            }
          break;
#endif
          }

        goto CONTINUE_CLASS;
        }  /* End handling \d-type escapes */

      /* A literal character may be followed by a range meta. At parse time
      there are checks for out-of-order characters, for ranges where the two
      characters are equal, and for hyphens that cannot indicate a range. At
      this point, therefore, no checking is needed. */

      else
        {
        uint32_t c, d;

        CLASS_LITERAL:
        c = d = meta;

        /* Remember if \r or \n were explicitly used */

        if (c == CHAR_CR || c == CHAR_NL) cb->external_flags |= PCRE2_HASCRORLF;

        /* Process a character range */

        if (pptr[1] == META_RANGE_LITERAL || pptr[1] == META_RANGE_ESCAPED)
          {
#ifdef EBCDIC
          BOOL range_is_literal = (pptr[1] == META_RANGE_LITERAL);
#endif
          pptr += 2;
          d = *pptr;
          if (d == META_BIGVALUE) d = *(++pptr);

          /* Remember an explicit \r or \n, and add the range to the class. */

          if (d == CHAR_CR || d == CHAR_NL) cb->external_flags |= PCRE2_HASCRORLF;

          /* In an EBCDIC environment, Perl treats alphabetic ranges specially
          because there are holes in the encoding, and simply using the range
          A-Z (for example) would include the characters in the holes. This
          applies only to literal ranges; [\xC1-\xE9] is different to [A-Z]. */

#ifdef EBCDIC
          if (range_is_literal &&
               (cb->ctypes[c] & ctype_letter) != 0 &&
               (cb->ctypes[d] & ctype_letter) != 0 &&
               (c <= CHAR_z) == (d <= CHAR_z))
            {
            uint32_t uc = (d <= CHAR_z)? 0 : 64;
            uint32_t C = c - uc;
            uint32_t D = d - uc;

            if (C <= CHAR_i)
              {
              class_has_8bitchar +=
                add_to_class(classbits, &class_uchardata, options, cb, C + uc,
                  ((D < CHAR_i)? D : CHAR_i) + uc);
              C = CHAR_j;
              }

            if (C <= D && C <= CHAR_r)
              {
              class_has_8bitchar +=
                add_to_class(classbits, &class_uchardata, options, cb, C + uc,
                  ((D < CHAR_r)? D : CHAR_r) + uc);
              C = CHAR_s;
              }

            if (C <= D)
              {
              class_has_8bitchar +=
                add_to_class(classbits, &class_uchardata, options, cb, C + uc,
                  D + uc);
              }
            }
          else
#endif
          /* Not an EBCDIC special range */

          class_has_8bitchar +=
            add_to_class(classbits, &class_uchardata, options, cb, c, d);
          goto CONTINUE_CLASS;   /* Go get the next char in the class */
          }  /* End of range handling */


        /* Handle a single character. */

        class_has_8bitchar +=
          add_to_class(classbits, &class_uchardata, options, cb, meta, meta);
        }

      /* Continue to the next item in the class. */

      CONTINUE_CLASS:

#ifdef SUPPORT_WIDE_CHARS
      /* If any wide characters or Unicode properties have been encountered,
      set xclass = TRUE. Then, in the pre-compile phase, accumulate the length
      of the extra data and reset the pointer. This is so that very large
      classes that contain a zillion wide characters or Unicode property tests
      do not overwrite the workspace (which is on the stack). */

      if (class_uchardata > class_uchardata_base)
        {
        xclass = TRUE;
        if (lengthptr != NULL)
          {
          *lengthptr += class_uchardata - class_uchardata_base;
          class_uchardata = class_uchardata_base;
          }
        }
#endif

      continue;  /* Needed to avoid error when not supporting wide chars */
      }   /* End of main class-processing loop */

    /* If this class is the first thing in the branch, there can be no first
    char setting, whatever the repeat count. Any reqcu setting must remain
    unchanged after any kind of repeat. */

    if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    zeroreqcu = reqcu;
    zeroreqcuflags = reqcuflags;

    /* If there are characters with values > 255, or Unicode property settings
    (\p or \P), we have to compile an extended class, with its own opcode,
    unless there were no property settings and there was a negated special such
    as \S in the class, and PCRE2_UCP is not set, because in that case all
    characters > 255 are in or not in the class, so any that were explicitly
    given as well can be ignored.

    In the UCP case, if certain negated POSIX classes ([:^ascii:] or
    [^:xdigit:]) were present in a class, we either have to match or not match
    all wide characters (depending on whether the whole class is or is not
    negated). This requirement is indicated by match_all_or_no_wide_chars being
    true. We do this by including an explicit range, which works in both cases.
    This applies only in UTF and 16-bit and 32-bit non-UTF modes, since there
    cannot be any wide characters in 8-bit non-UTF mode.

    When there *are* properties in a positive UTF-8 or any 16-bit or 32_bit
    class where \S etc is present without PCRE2_UCP, causing an extended class
    to be compiled, we make sure that all characters > 255 are included by
    forcing match_all_or_no_wide_chars to be true.

    If, when generating an xclass, there are no characters < 256, we can omit
    the bitmap in the actual compiled code. */

#ifdef SUPPORT_WIDE_CHARS  /* Defined for 16/32 bits, or 8-bit with Unicode */
    if (xclass && (
#ifdef SUPPORT_UNICODE
        (options & PCRE2_UCP) != 0 ||
#endif
        xclass_has_prop || !should_flip_negation))
      {
      if (match_all_or_no_wide_chars || (
#if PCRE2_CODE_UNIT_WIDTH == 8
           utf &&
#endif
           should_flip_negation && !negate_class && (options & PCRE2_UCP) == 0))
        {
        *class_uchardata++ = XCL_RANGE;
        if (utf)   /* Will always be utf in the 8-bit library */
          {
          class_uchardata += PRIV(ord2utf)(0x100, class_uchardata);
          class_uchardata += PRIV(ord2utf)(MAX_UTF_CODE_POINT, class_uchardata);
          }
        else       /* Can only happen for the 16-bit & 32-bit libraries */
          {
#if PCRE2_CODE_UNIT_WIDTH == 16
          *class_uchardata++ = 0x100;
          *class_uchardata++ = 0xffffu;
#elif PCRE2_CODE_UNIT_WIDTH == 32
          *class_uchardata++ = 0x100;
          *class_uchardata++ = 0xffffffffu;
#endif
          }
        }
      *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
      *code++ = OP_XCLASS;
      code += LINK_SIZE;
      *code = negate_class? XCL_NOT:0;
      if (xclass_has_prop) *code |= XCL_HASPROP;

      /* If the map is required, move up the extra data to make room for it;
      otherwise just move the code pointer to the end of the extra data. */

      if (class_has_8bitchar > 0)
        {
        *code++ |= XCL_MAP;
        (void)memmove(code + (32 / sizeof(PCRE2_UCHAR)), code,
          CU2BYTES(class_uchardata - code));
        if (negate_class && !xclass_has_prop)
          {
          /* Using 255 ^ instead of ~ avoids clang sanitize warning. */
          for (i = 0; i < 32; i++) classbits[i] = 255 ^ classbits[i];
          }
        memcpy(code, classbits, 32);
        code = class_uchardata + (32 / sizeof(PCRE2_UCHAR));
        }
      else code = class_uchardata;

      /* Now fill in the complete length of the item */

      PUT(previous, 1, (int)(code - previous));
      break;   /* End of class handling */
      }
#endif  /* SUPPORT_WIDE_CHARS */

    /* If there are no characters > 255, or they are all to be included or
    excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
    whole class was negated and whether there were negative specials such as \S
    (non-UCP) in the class. Then copy the 32-byte map into the code vector,
    negating it if necessary. */

    *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
    if (lengthptr == NULL)    /* Save time in the pre-compile phase */
      {
      if (negate_class)
        {
       /* Using 255 ^ instead of ~ avoids clang sanitize warning. */
       for (i = 0; i < 32; i++) classbits[i] = 255 ^ classbits[i];
       }
      memcpy(code, classbits, 32);
      }
    code += 32 / sizeof(PCRE2_UCHAR);
    break;  /* End of class processing */


    /* ===================================================================*/
    /* Deal with (*VERB)s. */

    /* Check for open captures before ACCEPT and close those that are within
    the same assertion level, also converting ACCEPT to ASSERT_ACCEPT in an
    assertion. In the first pass, just accumulate the length required;
    otherwise hitting (*ACCEPT) inside many nested parentheses can cause
    workspace overflow. Do not set firstcu after *ACCEPT. */

    case META_ACCEPT:
    cb->had_accept = TRUE;
    for (oc = cb->open_caps;
         oc != NULL && oc->assert_depth >= cb->assert_depth;
         oc = oc->next)
      {
      if (lengthptr != NULL)
        {
        *lengthptr += CU2BYTES(1) + IMM2_SIZE;
        }
      else
        {
        *code++ = OP_CLOSE;
        PUT2INC(code, 0, oc->number);
        }
      }
    *code++ = (cb->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
    if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
    break;

    case META_PRUNE:
    case META_SKIP:
    cb->had_pruneorskip = TRUE;
    /* Fall through */
    case META_COMMIT:
    case META_FAIL:
    *code++ = verbops[(meta - META_MARK) >> 16];
    break;

    case META_THEN:
    cb->external_flags |= PCRE2_HASTHEN;
    *code++ = OP_THEN;
    break;

    /* Handle verbs with arguments. Arguments can be very long, especially in
    16- and 32-bit modes, and can overflow the workspace in the first pass.
    However, the argument length is constrained to be small enough to fit in
    one code unit. This check happens in parse_regex(). In the first pass,
    instead of putting the argument into memory, we just update the length
    counter and set up an empty argument. */

    case META_THEN_ARG:
    cb->external_flags |= PCRE2_HASTHEN;
    goto VERB_ARG;

    case META_PRUNE_ARG:
    case META_SKIP_ARG:
    cb->had_pruneorskip = TRUE;
    /* Fall through */
    case META_MARK:
    case META_COMMIT_ARG:
    VERB_ARG:
    *code++ = verbops[(meta - META_MARK) >> 16];
    /* The length is in characters. */
    verbarglen = *(++pptr);
    verbculen = 0;
    tempcode = code++;
    for (i = 0; i < (int)verbarglen; i++)
      {
      meta = *(++pptr);
#ifdef SUPPORT_UNICODE
      if (utf) mclength = PRIV(ord2utf)(meta, mcbuffer); else
#endif
        {
        mclength = 1;
        mcbuffer[0] = meta;
        }
      if (lengthptr != NULL) *lengthptr += mclength; else
        {
        memcpy(code, mcbuffer, CU2BYTES(mclength));
        code += mclength;
        verbculen += mclength;
        }
      }

    *tempcode = verbculen;   /* Fill in the code unit length */
    *code++ = 0;             /* Terminating zero */
    break;


    /* ===================================================================*/
    /* Handle options change. The new setting must be passed back for use in
    subsequent branches. Reset the greedy defaults and the case value for
    firstcu and reqcu. */

    case META_OPTIONS:
    *optionsptr = options = *(++pptr);
    greedy_default = ((options & PCRE2_UNGREEDY) != 0);
    greedy_non_default = greedy_default ^ 1;
    req_caseopt = ((options & PCRE2_CASELESS) != 0)? REQ_CASELESS : 0;
    break;


    /* ===================================================================*/
    /* Handle conditional subpatterns. The case of (?(Rdigits) is ambiguous
    because it could be a numerical check on recursion, or a name check on a
    group's being set. The pre-pass sets up META_COND_RNUMBER as a name so that
    we can handle it either way. We first try for a name; if not found, process
    the number. */

    case META_COND_RNUMBER:   /* (?(Rdigits) */
    case META_COND_NAME:      /* (?(name) or (?'name') or ?(<name>) */
    case META_COND_RNAME:     /* (?(R&name) - test for recursion */
    bravalue = OP_COND;
      {
      int count, index;
      PCRE2_SPTR name;
      named_group *ng = cb->named_groups;
      uint32_t length = *(++pptr);

      GETPLUSOFFSET(offset, pptr);
      name = cb->start_pattern + offset;

      /* In the first pass, the names generated in the pre-pass are available,
      but the main name table has not yet been created. Scan the list of names
      generated in the pre-pass in order to get a number and whether or not
      this name is duplicated. If it is not duplicated, we can handle it as a
      numerical group. */

      for (i = 0; i < cb->names_found; i++, ng++)
        {
        if (length == ng->length &&
            PRIV(strncmp)(name, ng->name, length) == 0)
          {
          if (!ng->isdup)
            {
            code[1+LINK_SIZE] = (meta == META_COND_RNAME)? OP_RREF : OP_CREF;
            PUT2(code, 2+LINK_SIZE, ng->number);
            if (ng->number > cb->top_backref) cb->top_backref = ng->number;
            skipunits = 1+IMM2_SIZE;
            goto GROUP_PROCESS_NOTE_EMPTY;
            }
          break;  /* Found a duplicated name */
          }
        }

      /* If the name was not found we have a bad reference, unless we are
      dealing with R<digits>, which is treated as a recursion test by number.
      */

      if (i >= cb->names_found)
        {
        groupnumber = 0;
        if (meta == META_COND_RNUMBER)
          {
          for (i = 1; i < (int)length; i++)
            {
            groupnumber = groupnumber * 10 + name[i] - CHAR_0;
            if (groupnumber > MAX_GROUP_NUMBER)
              {
              *errorcodeptr = ERR61;
              cb->erroroffset = offset + i;
              return 0;
              }
            }
          }

        if (meta != META_COND_RNUMBER || groupnumber > cb->bracount)
          {
          *errorcodeptr = ERR15;
          cb->erroroffset = offset;
          return 0;
          }

        /* (?Rdigits) treated as a recursion reference by number. A value of
        zero (which is the result of both (?R) and (?R0)) means "any", and is
        translated into RREF_ANY (which is 0xffff). */

        if (groupnumber == 0) groupnumber = RREF_ANY;
        code[1+LINK_SIZE] = OP_RREF;
        PUT2(code, 2+LINK_SIZE, groupnumber);
        skipunits = 1+IMM2_SIZE;
        goto GROUP_PROCESS_NOTE_EMPTY;
        }

      /* A duplicated name was found. Note that if an R<digits> name is found
      (META_COND_RNUMBER), it is a reference test, not a recursion test. */

      code[1+LINK_SIZE] = (meta == META_COND_RNAME)? OP_RREF : OP_CREF;

      /* We have a duplicated name. In the compile pass we have to search the
      main table in order to get the index and count values. */

      count = 0;  /* Values for first pass (avoids compiler warning) */
      index = 0;
      if (lengthptr == NULL && !find_dupname_details(name, length, &index,
            &count, errorcodeptr, cb)) return 0;

      /* Add one to the opcode to change CREF/RREF into DNCREF/DNRREF and
      insert appropriate data values. */

      code[1+LINK_SIZE]++;
      skipunits = 1+2*IMM2_SIZE;
      PUT2(code, 2+LINK_SIZE, index);
      PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
      }
    goto GROUP_PROCESS_NOTE_EMPTY;

    /* The DEFINE condition is always false. Its internal groups may never
    be called, so matched_char must remain false, hence the jump to
    GROUP_PROCESS rather than GROUP_PROCESS_NOTE_EMPTY. */

    case META_COND_DEFINE:
    bravalue = OP_COND;
    GETPLUSOFFSET(offset, pptr);
    code[1+LINK_SIZE] = OP_DEFINE;
    skipunits = 1;
    goto GROUP_PROCESS;

    /* Conditional test of a group's being set. */

    case META_COND_NUMBER:
    bravalue = OP_COND;
    GETPLUSOFFSET(offset, pptr);
    groupnumber = *(++pptr);
    if (groupnumber > cb->bracount)
      {
      *errorcodeptr = ERR15;
      cb->erroroffset = offset;
      return 0;
      }
    if (groupnumber > cb->top_backref) cb->top_backref = groupnumber;
    offset -= 2;   /* Point at initial ( for too many branches error */
    code[1+LINK_SIZE] = OP_CREF;
    skipunits = 1+IMM2_SIZE;
    PUT2(code, 2+LINK_SIZE, groupnumber);
    goto GROUP_PROCESS_NOTE_EMPTY;

    /* Test for the PCRE2 version. */

    case META_COND_VERSION:
    bravalue = OP_COND;
    if (pptr[1] > 0)
      code[1+LINK_SIZE] = ((PCRE2_MAJOR > pptr[2]) ||
        (PCRE2_MAJOR == pptr[2] && PCRE2_MINOR >= pptr[3]))?
          OP_TRUE : OP_FALSE;
    else
      code[1+LINK_SIZE] = (PCRE2_MAJOR == pptr[2] && PCRE2_MINOR == pptr[3])?
        OP_TRUE : OP_FALSE;
    skipunits = 1;
    pptr += 3;
    goto GROUP_PROCESS_NOTE_EMPTY;

    /* The condition is an assertion, possibly preceded by a callout. */

    case META_COND_ASSERT:
    bravalue = OP_COND;
    goto GROUP_PROCESS_NOTE_EMPTY;


    /* ===================================================================*/
    /* Handle all kinds of nested bracketed groups. The non-capturing,
    non-conditional cases are here; others come to GROUP_PROCESS via goto. */

    case META_LOOKAHEAD:
    bravalue = OP_ASSERT;
    cb->assert_depth += 1;
    goto GROUP_PROCESS;

    /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
    thing to do, but Perl allows all assertions to be quantified, and when
    they contain capturing parentheses there may be a potential use for
    this feature. Not that that applies to a quantified (?!) but we allow
    it for uniformity. */

    case META_LOOKAHEADNOT:
    if (pptr[1] == META_KET &&
         (pptr[2] < META_ASTERISK || pptr[2] > META_MINMAX_QUERY))
      {
      *code++ = OP_FAIL;
      pptr++;
      }
    else
      {
      bravalue = OP_ASSERT_NOT;
      cb->assert_depth += 1;
      goto GROUP_PROCESS;
      }
    break;

    case META_LOOKBEHIND:
    bravalue = OP_ASSERTBACK;
    cb->assert_depth += 1;
    goto GROUP_PROCESS;

    case META_LOOKBEHINDNOT:
    bravalue = OP_ASSERTBACK_NOT;
    cb->assert_depth += 1;
    goto GROUP_PROCESS;

    case META_ATOMIC:
    bravalue = OP_ONCE;
    goto GROUP_PROCESS_NOTE_EMPTY;

    case META_SCRIPT_RUN:
    bravalue = OP_SCRIPT_RUN;
    goto GROUP_PROCESS_NOTE_EMPTY;

    case META_NOCAPTURE:
    bravalue = OP_BRA;
    /* Fall through */

    /* Process nested bracketed regex. The nesting depth is maintained for the
    benefit of the stackguard function. The test for too deep nesting is now
    done in parse_regex(). Assertion and DEFINE groups come to GROUP_PROCESS;
    others come to GROUP_PROCESS_NOTE_EMPTY, to indicate that we need to take
    note of whether or not they may match an empty string. */

    GROUP_PROCESS_NOTE_EMPTY:
    note_group_empty = TRUE;

    GROUP_PROCESS:
    cb->parens_depth += 1;
    *code = bravalue;
    pptr++;
    tempcode = code;
    tempreqvary = cb->req_varyopt;        /* Save value before group */
    length_prevgroup = 0;                 /* Initialize for pre-compile phase */

    if ((group_return =
         compile_regex(
         options,                         /* The option state */
         &tempcode,                       /* Where to put code (updated) */
         &pptr,                           /* Input pointer (updated) */
         errorcodeptr,                    /* Where to put an error message */
         skipunits,                       /* Skip over bracket number */
         &subfirstcu,                     /* For possible first char */
         &subfirstcuflags,
         &subreqcu,                       /* For possible last char */
         &subreqcuflags,
         bcptr,                           /* Current branch chain */
         cb,                              /* Compile data block */
         (lengthptr == NULL)? NULL :      /* Actual compile phase */
           &length_prevgroup              /* Pre-compile phase */
         )) == 0)
      return 0;  /* Error */

    cb->parens_depth -= 1;

    /* If that was a non-conditional significant group (not an assertion, not a
    DEFINE) that matches at least one character, then the current item matches
    a character. Conditionals are handled below. */

    if (note_group_empty && bravalue != OP_COND && group_return > 0)
      matched_char = TRUE;

    /* If we've just compiled an assertion, pop the assert depth. */

    if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
      cb->assert_depth -= 1;

    /* At the end of compiling, code is still pointing to the start of the
    group, while tempcode has been updated to point past the end of the group.
    The parsed pattern pointer (pptr) is on the closing META_KET.

    If this is a conditional bracket, check that there are no more than
    two branches in the group, or just one if it's a DEFINE group. We do this
    in the real compile phase, not in the pre-pass, where the whole group may
    not be available. */

    if (bravalue == OP_COND && lengthptr == NULL)
      {
      PCRE2_UCHAR *tc = code;
      int condcount = 0;

      do {
         condcount++;
         tc += GET(tc,1);
         }
      while (*tc != OP_KET);

      /* A DEFINE group is never obeyed inline (the "condition" is always
      false). It must have only one branch. Having checked this, change the
      opcode to OP_FALSE. */

      if (code[LINK_SIZE+1] == OP_DEFINE)
        {
        if (condcount > 1)
          {
          cb->erroroffset = offset;
          *errorcodeptr = ERR54;
          return 0;
          }
        code[LINK_SIZE+1] = OP_FALSE;
        bravalue = OP_DEFINE;   /* A flag to suppress char handling below */
        }

      /* A "normal" conditional group. If there is just one branch, we must not
      make use of its firstcu or reqcu, because this is equivalent to an
      empty second branch. Also, it may match an empty string. If there are two
      branches, this item must match a character if the group must. */

      else
        {
        if (condcount > 2)
          {
          cb->erroroffset = offset;
          *errorcodeptr = ERR27;
          return 0;
          }
        if (condcount == 1) subfirstcuflags = subreqcuflags = REQ_NONE;
          else if (group_return > 0) matched_char = TRUE;
        }
      }

    /* In the pre-compile phase, update the length by the length of the group,
    less the brackets at either end. Then reduce the compiled code to just a
    set of non-capturing brackets so that it doesn't use much memory if it is
    duplicated by a quantifier.*/

    if (lengthptr != NULL)
      {
      if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
        {
        *errorcodeptr = ERR20;
        return 0;
        }
      *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
      code++;   /* This already contains bravalue */
      PUTINC(code, 0, 1 + LINK_SIZE);
      *code++ = OP_KET;
      PUTINC(code, 0, 1 + LINK_SIZE);
      break;    /* No need to waste time with special character handling */
      }

    /* Otherwise update the main code pointer to the end of the group. */

    code = tempcode;

    /* For a DEFINE group, required and first character settings are not
    relevant. */

    if (bravalue == OP_DEFINE) break;

    /* Handle updating of the required and first code units for other types of
    group. Update for normal brackets of all kinds, and conditions with two
    branches (see code above). If the bracket is followed by a quantifier with
    zero repeat, we have to back off. Hence the definition of zeroreqcu and
    zerofirstcu outside the main loop so that they can be accessed for the back
    off. */

    zeroreqcu = reqcu;
    zeroreqcuflags = reqcuflags;
    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    groupsetfirstcu = FALSE;

    if (bravalue >= OP_ONCE)  /* Not an assertion */
      {
      /* If we have not yet set a firstcu in this branch, take it from the
      subpattern, remembering that it was set here so that a repeat of more
      than one can replicate it as reqcu if necessary. If the subpattern has
      no firstcu, set "none" for the whole branch. In both cases, a zero
      repeat forces firstcu to "none". */

      if (firstcuflags == REQ_UNSET && subfirstcuflags != REQ_UNSET)
        {
        if (subfirstcuflags >= 0)
          {
          firstcu = subfirstcu;
          firstcuflags = subfirstcuflags;
          groupsetfirstcu = TRUE;
          }
        else firstcuflags = REQ_NONE;
        zerofirstcuflags = REQ_NONE;
        }

      /* If firstcu was previously set, convert the subpattern's firstcu
      into reqcu if there wasn't one, using the vary flag that was in
      existence beforehand. */

      else if (subfirstcuflags >= 0 && subreqcuflags < 0)
        {
        subreqcu = subfirstcu;
        subreqcuflags = subfirstcuflags | tempreqvary;
        }

      /* If the subpattern set a required code unit (or set a first code unit
      that isn't really the first code unit - see above), set it. */

      if (subreqcuflags >= 0)
        {
        reqcu = subreqcu;
        reqcuflags = subreqcuflags;
        }
      }

    /* For a forward assertion, we take the reqcu, if set, provided that the
    group has also set a firstcu. This can be helpful if the pattern that
    follows the assertion doesn't set a different char. For example, it's
    useful for /(?=abcde).+/. We can't set firstcu for an assertion, however
    because it leads to incorrect effect for patterns such as /(?=a)a.+/ when
    the "real" "a" would then become a reqcu instead of a firstcu. This is
    overcome by a scan at the end if there's no firstcu, looking for an
    asserted first char. A similar effect for patterns like /(?=.*X)X$/ means
    we must only take the reqcu when the group also set a firstcu. Otherwise,
    in that example, 'X' ends up set for both. */

    else if (bravalue == OP_ASSERT && subreqcuflags >= 0 &&
             subfirstcuflags >= 0)
      {
      reqcu = subreqcu;
      reqcuflags = subreqcuflags;
      }

    break;  /* End of nested group handling */


    /* ===================================================================*/
    /* Handle named backreferences and recursions. */

    case META_BACKREF_BYNAME:
    case META_RECURSE_BYNAME:
      {
      int count, index;
      PCRE2_SPTR name;
      BOOL is_dupname = FALSE;
      named_group *ng = cb->named_groups;
      uint32_t length = *(++pptr);

      GETPLUSOFFSET(offset, pptr);
      name = cb->start_pattern + offset;

      /* In the first pass, the names generated in the pre-pass are available,
      but the main name table has not yet been created. Scan the list of names
      generated in the pre-pass in order to get a number and whether or not
      this name is duplicated. */

      groupnumber = 0;
      for (i = 0; i < cb->names_found; i++, ng++)
        {
        if (length == ng->length &&
            PRIV(strncmp)(name, ng->name, length) == 0)
          {
          is_dupname = ng->isdup;
          groupnumber = ng->number;

          /* For a recursion, that's all that is needed. We can now go to
          the code that handles numerical recursion, applying it to the first
          group with the given name. */

          if (meta == META_RECURSE_BYNAME)
            {
            meta_arg = groupnumber;
            goto HANDLE_NUMERICAL_RECURSION;
            }

          /* For a back reference, update the back reference map and the
          maximum back reference. Then, for each group, we must check to
          see if it is recursive, that is, it is inside the group that it
          references. A flag is set so that the group can be made atomic.
          */

          cb->backref_map |= (groupnumber < 32)? (1u << groupnumber) : 1;
          if (groupnumber > cb->top_backref)
            cb->top_backref = groupnumber;

          for (oc = cb->open_caps; oc != NULL; oc = oc->next)
            {
            if (oc->number == groupnumber)
              {
              oc->flag = TRUE;
              break;
              }
            }
          }
        }

      /* If the name was not found we have a bad reference. */

      if (groupnumber == 0)
        {
        *errorcodeptr = ERR15;
        cb->erroroffset = offset;
        return 0;
        }

      /* If a back reference name is not duplicated, we can handle it as
      a numerical reference. */

      if (!is_dupname)
        {
        meta_arg = groupnumber;
        goto HANDLE_SINGLE_REFERENCE;
        }

      /* If a back reference name is duplicated, we generate a different
      opcode to a numerical back reference. In the second pass we must
      search for the index and count in the final name table. */

      count = 0;  /* Values for first pass (avoids compiler warning) */
      index = 0;
      if (lengthptr == NULL && !find_dupname_details(name, length, &index,
            &count, errorcodeptr, cb)) return 0;

      if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
      *code++ = ((options & PCRE2_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
      PUT2INC(code, 0, index);
      PUT2INC(code, 0, count);
      }
    break;


    /* ===================================================================*/
    /* Handle a numerical callout. */

    case META_CALLOUT_NUMBER:
    code[0] = OP_CALLOUT;
    PUT(code, 1, pptr[1]);               /* Offset to next pattern item */
    PUT(code, 1 + LINK_SIZE, pptr[2]);   /* Length of next pattern item */
    code[1 + 2*LINK_SIZE] = pptr[3];
    pptr += 3;
    code += PRIV(OP_lengths)[OP_CALLOUT];
    break;


    /* ===================================================================*/
    /* Handle a callout with a string argument. In the pre-pass we just compute
    the length without generating anything. The length in pptr[3] includes both
    delimiters; in the actual compile only the first one is copied, but a
    terminating zero is added. Any doubled delimiters within the string make
    this an overestimate, but it is not worth bothering about. */

    case META_CALLOUT_STRING:
    if (lengthptr != NULL)
      {
      *lengthptr += pptr[3] + (1 + 4*LINK_SIZE);
      pptr += 3;
      SKIPOFFSET(pptr);
      }

    /* In the real compile we can copy the string. The starting delimiter is
     included so that the client can discover it if they want. We also pass the
     start offset to help a script language give better error messages. */

    else
      {
      PCRE2_SPTR pp;
      uint32_t delimiter;
      uint32_t length = pptr[3];
      PCRE2_UCHAR *callout_string = code + (1 + 4*LINK_SIZE);

      code[0] = OP_CALLOUT_STR;
      PUT(code, 1, pptr[1]);               /* Offset to next pattern item */
      PUT(code, 1 + LINK_SIZE, pptr[2]);   /* Length of next pattern item */

      pptr += 3;
      GETPLUSOFFSET(offset, pptr);         /* Offset to string in pattern */
      pp = cb->start_pattern + offset;
      delimiter = *callout_string++ = *pp++;
      if (delimiter == CHAR_LEFT_CURLY_BRACKET)
        delimiter = CHAR_RIGHT_CURLY_BRACKET;
      PUT(code, 1 + 3*LINK_SIZE, (int)(offset + 1));  /* One after delimiter */

      /* The syntax of the pattern was checked in the parsing scan. The length
      includes both delimiters, but we have passed the opening one just above,
      so we reduce length before testing it. The test is for > 1 because we do
      not want to copy the final delimiter. This also ensures that pp[1] is
      accessible. */

      while (--length > 1)
        {
        if (*pp == delimiter && pp[1] == delimiter)
          {
          *callout_string++ = delimiter;
          pp += 2;
          length--;
          }
        else *callout_string++ = *pp++;
        }
      *callout_string++ = CHAR_NUL;

      /* Set the length of the entire item, the advance to its end. */

      PUT(code, 1 + 2*LINK_SIZE, (int)(callout_string - code));
      code = callout_string;
      }
    break;


    /* ===================================================================*/
    /* Handle repetition. The different types are all sorted out in the parsing
    pass. */

    case META_MINMAX_PLUS:
    case META_MINMAX_QUERY:
    case META_MINMAX:
    repeat_min = *(++pptr);
    repeat_max = *(++pptr);
    goto REPEAT;

    case META_ASTERISK:
    case META_ASTERISK_PLUS:
    case META_ASTERISK_QUERY:
    repeat_min = 0;
    repeat_max = REPEAT_UNLIMITED;
    goto REPEAT;

    case META_PLUS:
    case META_PLUS_PLUS:
    case META_PLUS_QUERY:
    repeat_min = 1;
    repeat_max = REPEAT_UNLIMITED;
    goto REPEAT;

    case META_QUERY:
    case META_QUERY_PLUS:
    case META_QUERY_QUERY:
    repeat_min = 0;
    repeat_max = 1;

    REPEAT:
    if (previous_matched_char && repeat_min > 0) matched_char = TRUE;

    /* Remember whether this is a variable length repeat, and default to
    single-char opcodes. */

    reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
    op_type = 0;

    /* If the repeat is {1} we can ignore it. */

    if (repeat_max == 1 && repeat_min == 1) goto END_REPEAT;

    /* Adjust first and required code units for a zero repeat. */

    if (repeat_min == 0)
      {
      firstcu = zerofirstcu;
      firstcuflags = zerofirstcuflags;
      reqcu = zeroreqcu;
      reqcuflags = zeroreqcuflags;
      }

    /* Note the greediness and possessiveness. */

    switch (meta)
      {
      case META_MINMAX_PLUS:
      case META_ASTERISK_PLUS:
      case META_PLUS_PLUS:
      case META_QUERY_PLUS:
      repeat_type = 0;                  /* Force greedy */
      possessive_quantifier = TRUE;
      break;

      case META_MINMAX_QUERY:
      case META_ASTERISK_QUERY:
      case META_PLUS_QUERY:
      case META_QUERY_QUERY:
      repeat_type = greedy_non_default;
      possessive_quantifier = FALSE;
      break;

      default:
      repeat_type = greedy_default;
      possessive_quantifier = FALSE;
      break;
      }

    /* Save start of previous item, in case we have to move it up in order to
    insert something before it, and remember what it was. */

    tempcode = previous;
    op_previous = *previous;

    /* Now handle repetition for the different types of item. */

    switch (op_previous)
      {
      /* If previous was a character or negated character match, abolish the
      item and generate a repeat item instead. If a char item has a minimum of
      more than one, ensure that it is set in reqcu - it might not be if a
      sequence such as x{3} is the first thing in a branch because the x will
      have gone into firstcu instead.  */

      case OP_CHAR:
      case OP_CHARI:
      case OP_NOT:
      case OP_NOTI:
      op_type = chartypeoffset[op_previous - OP_CHAR];

      /* Deal with UTF characters that take up more than one code unit. */

#ifdef MAYBE_UTF_MULTI
      if (utf && NOT_FIRSTCU(code[-1]))
        {
        PCRE2_UCHAR *lastchar = code - 1;
        BACKCHAR(lastchar);
        mclength = (uint32_t)(code - lastchar);   /* Length of UTF character */
        memcpy(mcbuffer, lastchar, CU2BYTES(mclength));  /* Save the char */
        }
      else
#endif  /* MAYBE_UTF_MULTI */

      /* Handle the case of a single code unit - either with no UTF support, or
      with UTF disabled, or for a single-code-unit UTF character. */
        {
        mcbuffer[0] = code[-1];
        mclength = 1;
        if (op_previous <= OP_CHARI && repeat_min > 1)
          {
          reqcu = mcbuffer[0];
          reqcuflags = req_caseopt | cb->req_varyopt;
          }
        }
      goto OUTPUT_SINGLE_REPEAT;  /* Code shared with single character types */

      /* If previous was a character class or a back reference, we put the
      repeat stuff after it, but just skip the item if the repeat was {0,0}. */

#ifdef SUPPORT_WIDE_CHARS
      case OP_XCLASS:
#endif
      case OP_CLASS:
      case OP_NCLASS:
      case OP_REF:
      case OP_REFI:
      case OP_DNREF:
      case OP_DNREFI:

      if (repeat_max == 0)
        {
        code = previous;
        goto END_REPEAT;
        }

      if (repeat_min == 0 && repeat_max == REPEAT_UNLIMITED)
        *code++ = OP_CRSTAR + repeat_type;
      else if (repeat_min == 1 && repeat_max == REPEAT_UNLIMITED)
        *code++ = OP_CRPLUS + repeat_type;
      else if (repeat_min == 0 && repeat_max == 1)
        *code++ = OP_CRQUERY + repeat_type;
      else
        {
        *code++ = OP_CRRANGE + repeat_type;
        PUT2INC(code, 0, repeat_min);
        if (repeat_max == REPEAT_UNLIMITED) repeat_max = 0;  /* 2-byte encoding for max */
        PUT2INC(code, 0, repeat_max);
        }
      break;

      /* If previous is OP_FAIL, it was generated by an empty class []
      (PCRE2_ALLOW_EMPTY_CLASS is set). The other ways in which OP_FAIL can be
      generated, that is by (*FAIL) or (?!), disallow a quantifier at parse
      time. We can just ignore this repeat. */

      case OP_FAIL:
      goto END_REPEAT;

      /* Prior to 10.30, repeated recursions were wrapped in OP_ONCE brackets
      because pcre2_match() could not handle backtracking into recursively
      called groups. Now that this backtracking is available, we no longer need
      to do this. However, we still need to replicate recursions as we do for
      groups so as to have independent backtracking points. We can replicate
      for the minimum number of repeats directly. For optional repeats we now
      wrap the recursion in OP_BRA brackets and make use of the bracket
      repetition. */

      case OP_RECURSE:

      /* Generate unwrapped repeats for a non-zero minimum, except when the
      minimum is 1 and the maximum unlimited, because that can be handled with
      OP_BRA terminated by OP_KETRMAX/MIN. When the maximum is equal to the
      minimum, we just need to generate the appropriate additional copies.
      Otherwise we need to generate one more, to simulate the situation when
      the minimum is zero. */

      if (repeat_min > 0 && (repeat_min != 1 || repeat_max != REPEAT_UNLIMITED))
        {
        int replicate = repeat_min;
        if (repeat_min == repeat_max) replicate--;

        /* In the pre-compile phase, we don't actually do the replication. We
        just adjust the length as if we had. Do some paranoid checks for
        potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
        integer type when available, otherwise double. */

        if (lengthptr != NULL)
          {
          PCRE2_SIZE delta = replicate*(1 + LINK_SIZE);
          if ((INT64_OR_DOUBLE)replicate*
                (INT64_OR_DOUBLE)(1 + LINK_SIZE) >
                  (INT64_OR_DOUBLE)INT_MAX ||
              OFLOW_MAX - *lengthptr < delta)
            {
            *errorcodeptr = ERR20;
            return 0;
            }
          *lengthptr += delta;
          }

        else for (i = 0; i < replicate; i++)
          {
          memcpy(code, previous, CU2BYTES(1 + LINK_SIZE));
          previous = code;
          code += 1 + LINK_SIZE;
          }

        /* If the number of repeats is fixed, we are done. Otherwise, adjust
        the counts and fall through. */

        if (repeat_min == repeat_max) break;
        if (repeat_max != REPEAT_UNLIMITED) repeat_max -= repeat_min;
        repeat_min = 0;
        }

      /* Wrap the recursion call in OP_BRA brackets. */

      (void)memmove(previous + 1 + LINK_SIZE, previous, CU2BYTES(1 + LINK_SIZE));
      op_previous = *previous = OP_BRA;
      PUT(previous, 1, 2 + 2*LINK_SIZE);
      previous[2 + 2*LINK_SIZE] = OP_KET;
      PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
      code += 2 + 2 * LINK_SIZE;
      length_prevgroup = 3 + 3*LINK_SIZE;
      group_return = -1;  /* Set "may match empty string" */

      /* Now treat as a repeated OP_BRA. */
      /* Fall through */

      /* If previous was a bracket group, we may have to replicate it in
      certain cases. Note that at this point we can encounter only the "basic"
      bracket opcodes such as BRA and CBRA, as this is the place where they get
      converted into the more special varieties such as BRAPOS and SBRA.
      Originally, PCRE did not allow repetition of assertions, but now it does,
      for Perl compatibility. */

      case OP_ASSERT:
      case OP_ASSERT_NOT:
      case OP_ASSERTBACK:
      case OP_ASSERTBACK_NOT:
      case OP_ONCE:
      case OP_SCRIPT_RUN:
      case OP_BRA:
      case OP_CBRA:
      case OP_COND:
        {
        int len = (int)(code - previous);
        PCRE2_UCHAR *bralink = NULL;
        PCRE2_UCHAR *brazeroptr = NULL;

        /* Repeating a DEFINE group (or any group where the condition is always
        FALSE and there is only one branch) is pointless, but Perl allows the
        syntax, so we just ignore the repeat. */

        if (op_previous == OP_COND && previous[LINK_SIZE+1] == OP_FALSE &&
            previous[GET(previous, 1)] != OP_ALT)
          goto END_REPEAT;

        /* There is no sense in actually repeating assertions. The only
        potential use of repetition is in cases when the assertion is optional.
        Therefore, if the minimum is greater than zero, just ignore the repeat.
        If the maximum is not zero or one, set it to 1. */

        if (op_previous < OP_ONCE)    /* Assertion */
          {
          if (repeat_min > 0) goto END_REPEAT;
          if (repeat_max > 1) repeat_max = 1;
          }

        /* The case of a zero minimum is special because of the need to stick
        OP_BRAZERO in front of it, and because the group appears once in the
        data, whereas in other cases it appears the minimum number of times. For
        this reason, it is simplest to treat this case separately, as otherwise
        the code gets far too messy. There are several special subcases when the
        minimum is zero. */

        if (repeat_min == 0)
          {
          /* If the maximum is also zero, we used to just omit the group from
          the output altogether, like this:

          ** if (repeat_max == 0)
          **   {
          **   code = previous;
          **   goto END_REPEAT;
          **   }

          However, that fails when a group or a subgroup within it is
          referenced as a subroutine from elsewhere in the pattern, so now we
          stick in OP_SKIPZERO in front of it so that it is skipped on
          execution. As we don't have a list of which groups are referenced, we
          cannot do this selectively.

          If the maximum is 1 or unlimited, we just have to stick in the
          BRAZERO and do no more at this point. */

          if (repeat_max <= 1 || repeat_max == REPEAT_UNLIMITED)
            {
            (void)memmove(previous + 1, previous, CU2BYTES(len));
            code++;
            if (repeat_max == 0)
              {
              *previous++ = OP_SKIPZERO;
              goto END_REPEAT;
              }
            brazeroptr = previous;    /* Save for possessive optimizing */
            *previous++ = OP_BRAZERO + repeat_type;
            }

          /* If the maximum is greater than 1 and limited, we have to replicate
          in a nested fashion, sticking OP_BRAZERO before each set of brackets.
          The first one has to be handled carefully because it's the original
          copy, which has to be moved up. The remainder can be handled by code
          that is common with the non-zero minimum case below. We have to
          adjust the value or repeat_max, since one less copy is required. */

          else
            {
            int linkoffset;
            (void)memmove(previous + 2 + LINK_SIZE, previous, CU2BYTES(len));
            code += 2 + LINK_SIZE;
            *previous++ = OP_BRAZERO + repeat_type;
            *previous++ = OP_BRA;

            /* We chain together the bracket link offset fields that have to be
            filled in later when the ends of the brackets are reached. */

            linkoffset = (bralink == NULL)? 0 : (int)(previous - bralink);
            bralink = previous;
            PUTINC(previous, 0, linkoffset);
            }

          if (repeat_max != REPEAT_UNLIMITED) repeat_max--;
          }

        /* If the minimum is greater than zero, replicate the group as many
        times as necessary, and adjust the maximum to the number of subsequent
        copies that we need. */

        else
          {
          if (repeat_min > 1)
            {
            /* In the pre-compile phase, we don't actually do the replication.
            We just adjust the length as if we had. Do some paranoid checks for
            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
            integer type when available, otherwise double. */

            if (lengthptr != NULL)
              {
              PCRE2_SIZE delta = (repeat_min - 1)*length_prevgroup;
              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
                    (INT64_OR_DOUBLE)length_prevgroup >
                      (INT64_OR_DOUBLE)INT_MAX ||
                  OFLOW_MAX - *lengthptr < delta)
                {
                *errorcodeptr = ERR20;
                return 0;
                }
              *lengthptr += delta;
              }

            /* This is compiling for real. If there is a set first code unit
            for the group, and we have not yet set a "required code unit", set
            it. */

            else
              {
              if (groupsetfirstcu && reqcuflags < 0)
                {
                reqcu = firstcu;
                reqcuflags = firstcuflags;
                }
              for (i = 1; (uint32_t)i < repeat_min; i++)
                {
                memcpy(code, previous, CU2BYTES(len));
                code += len;
                }
              }
            }

          if (repeat_max != REPEAT_UNLIMITED) repeat_max -= repeat_min;
          }

        /* This code is common to both the zero and non-zero minimum cases. If
        the maximum is limited, it replicates the group in a nested fashion,
        remembering the bracket starts on a stack. In the case of a zero
        minimum, the first one was set up above. In all cases the repeat_max
        now specifies the number of additional copies needed. Again, we must
        remember to replicate entries on the forward reference list. */

        if (repeat_max != REPEAT_UNLIMITED)
          {
          /* In the pre-compile phase, we don't actually do the replication. We
          just adjust the length as if we had. For each repetition we must add
          1 to the length for BRAZERO and for all but the last repetition we
          must add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type
          is a 64-bit integer type when available, otherwise double. */

          if (lengthptr != NULL && repeat_max > 0)
            {
            PCRE2_SIZE delta = repeat_max*(length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
            if ((INT64_OR_DOUBLE)repeat_max *
                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
                    > (INT64_OR_DOUBLE)INT_MAX ||
                OFLOW_MAX - *lengthptr < delta)
              {
              *errorcodeptr = ERR20;
              return 0;
              }
            *lengthptr += delta;
            }

          /* This is compiling for real */

          else for (i = repeat_max - 1; i >= 0; i--)
            {
            *code++ = OP_BRAZERO + repeat_type;

            /* All but the final copy start a new nesting, maintaining the
            chain of brackets outstanding. */

            if (i != 0)
              {
              int linkoffset;
              *code++ = OP_BRA;
              linkoffset = (bralink == NULL)? 0 : (int)(code - bralink);
              bralink = code;
              PUTINC(code, 0, linkoffset);
              }

            memcpy(code, previous, CU2BYTES(len));
            code += len;
            }

          /* Now chain through the pending brackets, and fill in their length
          fields (which are holding the chain links pro tem). */

          while (bralink != NULL)
            {
            int oldlinkoffset;
            int linkoffset = (int)(code - bralink + 1);
            PCRE2_UCHAR *bra = code - linkoffset;
            oldlinkoffset = GET(bra, 1);
            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
            *code++ = OP_KET;
            PUTINC(code, 0, linkoffset);
            PUT(bra, 1, linkoffset);
            }
          }

        /* If the maximum is unlimited, set a repeater in the final copy. For
        SCRIPT_RUN and ONCE brackets, that's all we need to do. However,
        possessively repeated ONCE brackets can be converted into non-capturing
        brackets, as the behaviour of (?:xx)++ is the same as (?>xx)++ and this
        saves having to deal with possessive ONCEs specially.

        Otherwise, when we are doing the actual compile phase, check to see
        whether this group is one that could match an empty string. If so,
        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
        that runtime checking can be done. [This check is also applied to ONCE
        and SCRIPT_RUN groups at runtime, but in a different way.]

        Then, if the quantifier was possessive and the bracket is not a
        conditional, we convert the BRA code to the POS form, and the KET code to
        KETRPOS. (It turns out to be convenient at runtime to detect this kind of
        subpattern at both the start and at the end.) The use of special opcodes
        makes it possible to reduce greatly the stack usage in pcre2_match(). If
        the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.

        Then, if the minimum number of matches is 1 or 0, cancel the possessive
        flag so that the default action below, of wrapping everything inside
        atomic brackets, does not happen. When the minimum is greater than 1,
        there will be earlier copies of the group, and so we still have to wrap
        the whole thing. */

        else
          {
          PCRE2_UCHAR *ketcode = code - 1 - LINK_SIZE;
          PCRE2_UCHAR *bracode = ketcode - GET(ketcode, 1);

          /* Convert possessive ONCE brackets to non-capturing */

          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;

          /* For non-possessive ONCE and for SCRIPT_RUN brackets, all we need
          to do is to set the KET. */

          if (*bracode == OP_ONCE || *bracode == OP_SCRIPT_RUN)
            *ketcode = OP_KETRMAX + repeat_type;

          /* Handle non-SCRIPT_RUN and non-ONCE brackets and possessive ONCEs
          (which have been converted to non-capturing above). */

          else
            {
            /* In the compile phase, adjust the opcode if the group can match
            an empty string. For a conditional group with only one branch, the
            value of group_return will not show "could be empty", so we must
            check that separately. */

            if (lengthptr == NULL)
              {
              if (group_return < 0) *bracode += OP_SBRA - OP_BRA;
              if (*bracode == OP_COND && bracode[GET(bracode,1)] != OP_ALT)
                *bracode = OP_SCOND;
              }

            /* Handle possessive quantifiers. */

            if (possessive_quantifier)
              {
              /* For COND brackets, we wrap the whole thing in a possessively
              repeated non-capturing bracket, because we have not invented POS
              versions of the COND opcodes. */

              if (*bracode == OP_COND || *bracode == OP_SCOND)
                {
                int nlen = (int)(code - bracode);
                (void)memmove(bracode + 1 + LINK_SIZE, bracode, CU2BYTES(nlen));
                code += 1 + LINK_SIZE;
                nlen += 1 + LINK_SIZE;
                *bracode = (*bracode == OP_COND)? OP_BRAPOS : OP_SBRAPOS;
                *code++ = OP_KETRPOS;
                PUTINC(code, 0, nlen);
                PUT(bracode, 1, nlen);
                }

              /* For non-COND brackets, we modify the BRA code and use KETRPOS. */

              else
                {
                *bracode += 1;              /* Switch to xxxPOS opcodes */
                *ketcode = OP_KETRPOS;
                }

              /* If the minimum is zero, mark it as possessive, then unset the
              possessive flag when the minimum is 0 or 1. */

              if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
              if (repeat_min < 2) possessive_quantifier = FALSE;
              }

            /* Non-possessive quantifier */

            else *ketcode = OP_KETRMAX + repeat_type;
            }
          }
        }
      break;

      /* If previous was a character type match (\d or similar), abolish it and
      create a suitable repeat item. The code is shared with single-character
      repeats by setting op_type to add a suitable offset into repeat_type.
      Note the the Unicode property types will be present only when
      SUPPORT_UNICODE is defined, but we don't wrap the little bits of code
      here because it just makes it horribly messy. */

      default:
      if (op_previous >= OP_EODN)   /* Not a character type - internal error */
        {
        *errorcodeptr = ERR10;
        return 0;
        }
      else
        {
        int prop_type, prop_value;
        PCRE2_UCHAR *oldcode;

        op_type = OP_TYPESTAR - OP_STAR;      /* Use type opcodes */
        mclength = 0;                         /* Not a character */

        if (op_previous == OP_PROP || op_previous == OP_NOTPROP)
          {
          prop_type = previous[1];
          prop_value = previous[2];
          }
        else
          {
          /* Come here from just above with a character in mcbuffer/mclength. */
          OUTPUT_SINGLE_REPEAT:
          prop_type = prop_value = -1;
          }

        /* At this point, if prop_type == prop_value == -1 we either have a
        character in mcbuffer when mclength is greater than zero, or we have
        mclength zero, in which case there is a non-property character type in
        op_previous. If prop_type/value are not negative, we have a property
        character type in op_previous. */

        oldcode = code;                   /* Save where we were */
        code = previous;                  /* Usually overwrite previous item */

        /* If the maximum is zero then the minimum must also be zero; Perl allows
        this case, so we do too - by simply omitting the item altogether. */

        if (repeat_max == 0) goto END_REPEAT;

        /* Combine the op_type with the repeat_type */

        repeat_type += op_type;

        /* A minimum of zero is handled either as the special case * or ?, or as
        an UPTO, with the maximum given. */

        if (repeat_min == 0)
          {
          if (repeat_max == REPEAT_UNLIMITED) *code++ = OP_STAR + repeat_type;
            else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
          else
            {
            *code++ = OP_UPTO + repeat_type;
            PUT2INC(code, 0, repeat_max);
            }
          }

        /* A repeat minimum of 1 is optimized into some special cases. If the
        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
        left in place and, if the maximum is greater than 1, we use OP_UPTO with
        one less than the maximum. */

        else if (repeat_min == 1)
          {
          if (repeat_max == REPEAT_UNLIMITED)
            *code++ = OP_PLUS + repeat_type;
          else
            {
            code = oldcode;  /* Leave previous item in place */
            if (repeat_max == 1) goto END_REPEAT;
            *code++ = OP_UPTO + repeat_type;
            PUT2INC(code, 0, repeat_max - 1);
            }
          }

        /* The case {n,n} is just an EXACT, while the general case {n,m} is
        handled as an EXACT followed by an UPTO or STAR or QUERY. */

        else
          {
          *code++ = OP_EXACT + op_type;  /* NB EXACT doesn't have repeat_type */
          PUT2INC(code, 0, repeat_min);

          /* Unless repeat_max equals repeat_min, fill in the data for EXACT,
          and then generate the second opcode. For a repeated Unicode property
          match, there are two extra values that define the required property,
          and mclength is set zero to indicate this. */

          if (repeat_max != repeat_min)
            {
            if (mclength > 0)
              {
              memcpy(code, mcbuffer, CU2BYTES(mclength));
              code += mclength;
              }
            else
              {
              *code++ = op_previous;
              if (prop_type >= 0)
                {
                *code++ = prop_type;
                *code++ = prop_value;
                }
              }

            /* Now set up the following opcode */

            if (repeat_max == REPEAT_UNLIMITED)
              *code++ = OP_STAR + repeat_type;
            else
              {
              repeat_max -= repeat_min;
              if (repeat_max == 1)
                {
                *code++ = OP_QUERY + repeat_type;
                }
              else
                {
                *code++ = OP_UPTO + repeat_type;
                PUT2INC(code, 0, repeat_max);
                }
              }
            }
          }

        /* Fill in the character or character type for the final opcode. */

        if (mclength > 0)
          {
          memcpy(code, mcbuffer, CU2BYTES(mclength));
          code += mclength;
          }
        else
          {
          *code++ = op_previous;
          if (prop_type >= 0)
            {
            *code++ = prop_type;
            *code++ = prop_value;
            }
          }
        }
      break;
      }  /* End of switch on different op_previous values */


    /* If the character following a repeat is '+', possessive_quantifier is
    TRUE. For some opcodes, there are special alternative opcodes for this
    case. For anything else, we wrap the entire repeated item inside OP_ONCE
    brackets. Logically, the '+' notation is just syntactic sugar, taken from
    Sun's Java package, but the special opcodes can optimize it.

    Some (but not all) possessively repeated subpatterns have already been
    completely handled in the code just above. For them, possessive_quantifier
    is always FALSE at this stage. Note that the repeated item starts at
    tempcode, not at previous, which might be the first part of a string whose
    (former) last char we repeated. */

    if (possessive_quantifier)
      {
      int len;

      /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
      However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
      {5,}, or {5,10}). We skip over an EXACT item; if the length of what
      remains is greater than zero, there's a further opcode that can be
      handled. If not, do nothing, leaving the EXACT alone. */

      switch(*tempcode)
        {
        case OP_TYPEEXACT:
        tempcode += PRIV(OP_lengths)[*tempcode] +
          ((tempcode[1 + IMM2_SIZE] == OP_PROP
          || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
        break;

        /* CHAR opcodes are used for exacts whose count is 1. */

        case OP_CHAR:
        case OP_CHARI:
        case OP_NOT:
        case OP_NOTI:
        case OP_EXACT:
        case OP_EXACTI:
        case OP_NOTEXACT:
        case OP_NOTEXACTI:
        tempcode += PRIV(OP_lengths)[*tempcode];
#ifdef SUPPORT_UNICODE
        if (utf && HAS_EXTRALEN(tempcode[-1]))
          tempcode += GET_EXTRALEN(tempcode[-1]);
#endif
        break;

        /* For the class opcodes, the repeat operator appears at the end;
        adjust tempcode to point to it. */

        case OP_CLASS:
        case OP_NCLASS:
        tempcode += 1 + 32/sizeof(PCRE2_UCHAR);
        break;

#ifdef SUPPORT_WIDE_CHARS
        case OP_XCLASS:
        tempcode += GET(tempcode, 1);
        break;
#endif
        }

      /* If tempcode is equal to code (which points to the end of the repeated
      item), it means we have skipped an EXACT item but there is no following
      QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
      all other cases, tempcode will be pointing to the repeat opcode, and will
      be less than code, so the value of len will be greater than 0. */

      len = (int)(code - tempcode);
      if (len > 0)
        {
        unsigned int repcode = *tempcode;

        /* There is a table for possessifying opcodes, all of which are less
        than OP_CALLOUT. A zero entry means there is no possessified version.
        */

        if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
          *tempcode = opcode_possessify[repcode];

        /* For opcode without a special possessified version, wrap the item in
        ONCE brackets. */

        else
          {
          (void)memmove(tempcode + 1 + LINK_SIZE, tempcode, CU2BYTES(len));
          code += 1 + LINK_SIZE;
          len += 1 + LINK_SIZE;
          tempcode[0] = OP_ONCE;
          *code++ = OP_KET;
          PUTINC(code, 0, len);
          PUT(tempcode, 1, len);
          }
        }
      }

    /* We set the "follows varying string" flag for subsequently encountered
    reqcus if it isn't already set and we have just passed a varying length
    item. */

    END_REPEAT:
    cb->req_varyopt |= reqvary;
    break;


    /* ===================================================================*/
    /* Handle a 32-bit data character with a value greater than META_END. */

    case META_BIGVALUE:
    pptr++;
    goto NORMAL_CHAR;


    /* ===============================================================*/
    /* Handle a back reference by number, which is the meta argument. The
    pattern offsets for back references to group numbers less than 10 are held
    in a special vector, to avoid using more than two parsed pattern elements
    in 64-bit environments. We only need the offset to the first occurrence,
    because if that doesn't fail, subsequent ones will also be OK. */

    case META_BACKREF:
    if (meta_arg < 10) offset = cb->small_ref_offset[meta_arg];
      else GETPLUSOFFSET(offset, pptr);

    if (meta_arg > cb->bracount)
      {
      cb->erroroffset = offset;
      *errorcodeptr = ERR15;  /* Non-existent subpattern */
      return 0;
      }

    /* Come here from named backref handling when the reference is to a
    single group (that is, not to a duplicated name). The back reference
    data will have already been updated. We must disable firstcu if not
    set, to cope with cases like (?=(\w+))\1: which would otherwise set ':'
    later. */

    HANDLE_SINGLE_REFERENCE:
    if (firstcuflags == REQ_UNSET) zerofirstcuflags = firstcuflags = REQ_NONE;
    *code++ = ((options & PCRE2_CASELESS) != 0)? OP_REFI : OP_REF;
    PUT2INC(code, 0, meta_arg);

    /* Update the map of back references, and keep the highest one. We
    could do this in parse_regex() for numerical back references, but not
    for named back references, because we don't know the numbers to which
    named back references refer. So we do it all in this function. */

    cb->backref_map |= (meta_arg < 32)? (1u << meta_arg) : 1;
    if (meta_arg > cb->top_backref) cb->top_backref = meta_arg;

    /* Check to see if this back reference is recursive, that it, it
    is inside the group that it references. A flag is set so that the
    group can be made atomic. */

    for (oc = cb->open_caps; oc != NULL; oc = oc->next)
      {
      if (oc->number == meta_arg)
        {
        oc->flag = TRUE;
        break;
        }
      }
    break;


    /* ===============================================================*/
    /* Handle recursion by inserting the number of the called group (which is
    the meta argument) after OP_RECURSE. At the end of compiling the pattern is
    scanned and these numbers are replaced by offsets within the pattern. It is
    done like this to avoid problems with forward references and adjusting
    offsets when groups are duplicated and moved (as discovered in previous
    implementations). Note that a recursion does not have a set first
    character. */

    case META_RECURSE:
    GETPLUSOFFSET(offset, pptr);
    if (meta_arg > cb->bracount)
      {
      cb->erroroffset = offset;
      *errorcodeptr = ERR15;  /* Non-existent subpattern */
      return 0;
      }
    HANDLE_NUMERICAL_RECURSION:
    *code = OP_RECURSE;
    PUT(code, 1, meta_arg);
    code += 1 + LINK_SIZE;
    groupsetfirstcu = FALSE;
    cb->had_recurse = TRUE;
    if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    break;


    /* ===============================================================*/
    /* Handle capturing parentheses; the number is the meta argument. */

    case META_CAPTURE:
    bravalue = OP_CBRA;
    skipunits = IMM2_SIZE;
    PUT2(code, 1+LINK_SIZE, meta_arg);
    cb->lastcapture = meta_arg;
    goto GROUP_PROCESS_NOTE_EMPTY;


    /* ===============================================================*/
    /* Handle escape sequence items. For ones like \d, the ESC_values are
    arranged to be the same as the corresponding OP_values in the default case
    when PCRE2_UCP is not set (which is the only case in which they will appear
    here).

    Note: \Q and \E are never seen here, as they were dealt with in
    parse_pattern(). Neither are numerical back references or recursions, which
    were turned into META_BACKREF or META_RECURSE items, respectively. \k and
    \g, when followed by names, are turned into META_BACKREF_BYNAME or
    META_RECURSE_BYNAME. */

    case META_ESCAPE:

    /* We can test for escape sequences that consume a character because their
    values lie between ESC_b and ESC_Z; this may have to change if any new ones
    are ever created. For these sequences, we disable the setting of a first
    character if it hasn't already been set. */

    if (meta_arg > ESC_b && meta_arg < ESC_Z)
      {
      matched_char = TRUE;
      if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE;
      }

    /* Set values to reset to if this is followed by a zero repeat. */

    zerofirstcu = firstcu;
    zerofirstcuflags = firstcuflags;
    zeroreqcu = reqcu;
    zeroreqcuflags = reqcuflags;

    /* If Unicode is not supported, \P and \p are not allowed and are
    faulted at parse time, so will never appear here. */

#ifdef SUPPORT_UNICODE
    if (meta_arg == ESC_P || meta_arg == ESC_p)
      {
      uint32_t ptype = *(++pptr) >> 16;
      uint32_t pdata = *pptr & 0xffff;

      /* The special case of \p{Any} is compiled to OP_ALLANY so as to benefit
      from the auto-anchoring code. */

      if (meta_arg == ESC_p && ptype == PT_ANY)
        {
        *code++ = OP_ALLANY;
        }
      else
        {
        *code++ = (meta_arg == ESC_p)? OP_PROP : OP_NOTPROP;
        *code++ = ptype;
        *code++ = pdata;
        }
      break;  /* End META_ESCAPE */
      }
#endif

    /* For the rest (including \X when Unicode is supported - if not it's
    faulted at parse time), the OP value is the escape value when PCRE2_UCP is
    not set; if it is set, these escapes do not show up here because they are
    converted into Unicode property tests in parse_regex(). Note that \b and \B
    do a one-character lookbehind, and \A also behaves as if it does. */

    if (meta_arg == ESC_C) cb->external_flags |= PCRE2_HASBKC; /* Record */
    if ((meta_arg == ESC_b || meta_arg == ESC_B || meta_arg == ESC_A) &&
         cb->max_lookbehind == 0)
      cb->max_lookbehind = 1;

    /* In non-UTF mode, and for both 32-bit modes, we turn \C into OP_ALLANY
    instead of OP_ANYBYTE so that it works in DFA mode and in lookbehinds. */

#if PCRE2_CODE_UNIT_WIDTH == 32
    *code++ = (meta_arg == ESC_C)? OP_ALLANY : meta_arg;
#else
    *code++ = (!utf && meta_arg == ESC_C)? OP_ALLANY : meta_arg;
#endif
    break;  /* End META_ESCAPE */


    /* ===================================================================*/
    /* Handle an unrecognized meta value. A parsed pattern value less than
    META_END is a literal. Otherwise we have a problem. */

    default:
    if (meta >= META_END)
      {
#ifdef DEBUG_SHOW_PARSED
      fprintf(stderr, "** Unrecognized parsed pattern item 0x%.8x\n", *pptr);
#endif
      *errorcodeptr = ERR89;  /* Internal error - unrecognized. */
      return 0;
      }

    /* Handle a literal character. We come here by goto in the case of a
    32-bit, non-UTF character whose value is greater than META_END. */

    NORMAL_CHAR:
    meta = *pptr;     /* Get the full 32 bits */
    NORMAL_CHAR_SET:  /* Character is already in meta */
    matched_char = TRUE;

    /* For caseless UTF mode, check whether this character has more than one
    other case. If so, generate a special OP_PROP item instead of OP_CHARI. */

#ifdef SUPPORT_UNICODE
    if (utf && (options & PCRE2_CASELESS) != 0)
      {
      uint32_t caseset = UCD_CASESET(meta);
      if (caseset != 0)
        {
        *code++ = OP_PROP;
        *code++ = PT_CLIST;
        *code++ = caseset;
        if (firstcuflags == REQ_UNSET)
          firstcuflags = zerofirstcuflags = REQ_NONE;
        break;  /* End handling this meta item */
        }
      }
#endif

    /* Caseful matches, or not one of the multicase characters. Get the
    character's code units into mcbuffer, with the length in mclength. When not
    in UTF mode, the length is always 1. */

#ifdef SUPPORT_UNICODE
    if (utf) mclength = PRIV(ord2utf)(meta, mcbuffer); else
#endif
      {
      mclength = 1;
      mcbuffer[0] = meta;
      }

    /* Generate the appropriate code */

    *code++ = ((options & PCRE2_CASELESS) != 0)? OP_CHARI : OP_CHAR;
    memcpy(code, mcbuffer, CU2BYTES(mclength));
    code += mclength;

    /* Remember if \r or \n were seen */

    if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
      cb->external_flags |= PCRE2_HASCRORLF;

    /* Set the first and required code units appropriately. If no previous
    first code unit, set it from this character, but revert to none on a zero
    repeat. Otherwise, leave the firstcu value alone, and don't change it on
    a zero repeat. */

    if (firstcuflags == REQ_UNSET)
      {
      zerofirstcuflags = REQ_NONE;
      zeroreqcu = reqcu;
      zeroreqcuflags = reqcuflags;

      /* If the character is more than one code unit long, we can set firstcu
      only if it is not to be matched caselessly. */

      if (mclength == 1 || req_caseopt == 0)
        {
        firstcu = mcbuffer[0];
        firstcuflags = req_caseopt;
        if (mclength != 1)
          {
          reqcu = code[-1];
          reqcuflags = cb->req_varyopt;
          }
        }
      else firstcuflags = reqcuflags = REQ_NONE;
      }

    /* firstcu was previously set; we can set reqcu only if the length is
    1 or the matching is caseful. */

    else
      {
      zerofirstcu = firstcu;
      zerofirstcuflags = firstcuflags;
      zeroreqcu = reqcu;
      zeroreqcuflags = reqcuflags;
      if (mclength == 1 || req_caseopt == 0)
        {
        reqcu = code[-1];
        reqcuflags = req_caseopt | cb->req_varyopt;
        }
      }
    break;    /* End default meta handling */
    }         /* End of big switch */
  }           /* End of big loop */

/* Control never reaches here. */
}



/*************************************************
*   Compile regex: a sequence of alternatives    *
*************************************************/

/* On entry, pptr is pointing past the bracket meta, but on return it points to
the closing bracket or META_END. The code variable is pointing at the code unit
into which the BRA operator has been stored. This function is used during the
pre-compile phase when we are trying to find out the amount of memory needed,
as well as during the real compile phase. The value of lengthptr distinguishes
the two phases.

Arguments:
  options           option bits, including any changes for this subpattern
  codeptr           -> the address of the current code pointer
  pptrptr           -> the address of the current parsed pattern pointer
  errorcodeptr      -> pointer to error code variable
  skipunits         skip this many code units at start (for brackets and OP_COND)
  firstcuptr        place to put the first required code unit
  firstcuflagsptr   place to put the first code unit flags, or a negative number
  reqcuptr          place to put the last required code unit
  reqcuflagsptr     place to put the last required code unit flags, or a negative number
  bcptr             pointer to the chain of currently open branches
  cb                points to the data block with tables pointers etc.
  lengthptr         NULL during the real compile phase
                    points to length accumulator during pre-compile phase

Returns:            0 There has been an error
                   +1 Success, this group must match at least one character
                   -1 Success, this group may match an empty string
*/

static int
compile_regex(uint32_t options, PCRE2_UCHAR **codeptr, uint32_t **pptrptr,
  int *errorcodeptr, uint32_t skipunits, uint32_t *firstcuptr,
  int32_t *firstcuflagsptr, uint32_t *reqcuptr,int32_t *reqcuflagsptr,
  branch_chain *bcptr, compile_block *cb, PCRE2_SIZE *lengthptr)
{
PCRE2_UCHAR *code = *codeptr;
PCRE2_UCHAR *last_branch = code;
PCRE2_UCHAR *start_bracket = code;
BOOL lookbehind;
open_capitem capitem;
int capnumber = 0;
int okreturn = 1;
uint32_t *pptr = *pptrptr;
uint32_t firstcu, reqcu;
uint32_t lookbehindlength;
int32_t firstcuflags, reqcuflags;
uint32_t branchfirstcu, branchreqcu;
int32_t branchfirstcuflags, branchreqcuflags;
PCRE2_SIZE length;
branch_chain bc;

/* If set, call the external function that checks for stack availability. */

if (cb->cx->stack_guard != NULL &&
    cb->cx->stack_guard(cb->parens_depth, cb->cx->stack_guard_data))
  {
  *errorcodeptr= ERR33;
  return 0;
  }

/* Miscellaneous initialization */

bc.outer = bcptr;
bc.current_branch = code;

firstcu = reqcu = 0;
firstcuflags = reqcuflags = REQ_UNSET;

/* Accumulate the length for use in the pre-compile phase. Start with the
length of the BRA and KET and any extra code units that are required at the
beginning. We accumulate in a local variable to save frequent testing of
lengthptr for NULL. We cannot do this by looking at the value of 'code' at the
start and end of each alternative, because compiled items are discarded during
the pre-compile phase so that the workspace is not exceeded. */

length = 2 + 2*LINK_SIZE + skipunits;

/* Remember if this is a lookbehind assertion, and if it is, save its length
and skip over the pattern offset. */

lookbehind = *code == OP_ASSERTBACK || *code == OP_ASSERTBACK_NOT;
if (lookbehind)
  {
  lookbehindlength = META_DATA(pptr[-1]);
  pptr += SIZEOFFSET;
  }
else lookbehindlength = 0;

/* If this is a capturing subpattern, add to the chain of open capturing items
so that we can detect them if (*ACCEPT) is encountered. Note that only OP_CBRA
need be tested here; changing this opcode to one of its variants, e.g.
OP_SCBRAPOS, happens later, after the group has been compiled. */

if (*code == OP_CBRA)
  {
  capnumber = GET2(code, 1 + LINK_SIZE);
  capitem.number = capnumber;
  capitem.next = cb->open_caps;
  capitem.flag = FALSE;
  capitem.assert_depth = cb->assert_depth;
  cb->open_caps = &capitem;
  }

/* Offset is set zero to mark that this bracket is still open */

PUT(code, 1, 0);
code += 1 + LINK_SIZE + skipunits;

/* Loop for each alternative branch */

for (;;)
  {
  int branch_return;

  /* Insert OP_REVERSE if this is as lookbehind assertion. */

  if (lookbehind && lookbehindlength > 0)
    {
    *code++ = OP_REVERSE;
    PUTINC(code, 0, lookbehindlength);
    length += 1 + LINK_SIZE;
    }

  /* Now compile the branch; in the pre-compile phase its length gets added
  into the length. */

  if ((branch_return =
        compile_branch(&options, &code, &pptr, errorcodeptr, &branchfirstcu,
          &branchfirstcuflags, &branchreqcu, &branchreqcuflags, &bc,
          cb, (lengthptr == NULL)? NULL : &length)) == 0)
    return 0;

  /* If a branch can match an empty string, so can the whole group. */

  if (branch_return < 0) okreturn = -1;

  /* In the real compile phase, there is some post-processing to be done. */

  if (lengthptr == NULL)
    {
    /* If this is the first branch, the firstcu and reqcu values for the
    branch become the values for the regex. */

    if (*last_branch != OP_ALT)
      {
      firstcu = branchfirstcu;
      firstcuflags = branchfirstcuflags;
      reqcu = branchreqcu;
      reqcuflags = branchreqcuflags;
      }

    /* If this is not the first branch, the first char and reqcu have to
    match the values from all the previous branches, except that if the
    previous value for reqcu didn't have REQ_VARY set, it can still match,
    and we set REQ_VARY for the regex. */

    else
      {
      /* If we previously had a firstcu, but it doesn't match the new branch,
      we have to abandon the firstcu for the regex, but if there was
      previously no reqcu, it takes on the value of the old firstcu. */

      if (firstcuflags != branchfirstcuflags || firstcu != branchfirstcu)
        {
        if (firstcuflags >= 0)
          {
          if (reqcuflags < 0)
            {
            reqcu = firstcu;
            reqcuflags = firstcuflags;
            }
          }
        firstcuflags = REQ_NONE;
        }

      /* If we (now or from before) have no firstcu, a firstcu from the
      branch becomes a reqcu if there isn't a branch reqcu. */

      if (firstcuflags < 0 && branchfirstcuflags >= 0 &&
          branchreqcuflags < 0)
        {
        branchreqcu = branchfirstcu;
        branchreqcuflags = branchfirstcuflags;
        }

      /* Now ensure that the reqcus match */

      if (((reqcuflags & ~REQ_VARY) != (branchreqcuflags & ~REQ_VARY)) ||
          reqcu != branchreqcu)
        reqcuflags = REQ_NONE;
      else
        {
        reqcu = branchreqcu;
        reqcuflags |= branchreqcuflags; /* To "or" REQ_VARY */
        }
      }
    }

  /* Handle reaching the end of the expression, either ')' or end of pattern.
  In the real compile phase, go back through the alternative branches and
  reverse the chain of offsets, with the field in the BRA item now becoming an
  offset to the first alternative. If there are no alternatives, it points to
  the end of the group. The length in the terminating ket is always the length
  of the whole bracketed item. Return leaving the pointer at the terminating
  char. */

  if (META_CODE(*pptr) != META_ALT)
    {
    if (lengthptr == NULL)
      {
      PCRE2_SIZE branch_length = code - last_branch;
      do
        {
        PCRE2_SIZE prev_length = GET(last_branch, 1);
        PUT(last_branch, 1, branch_length);
        branch_length = prev_length;
        last_branch -= branch_length;
        }
      while (branch_length > 0);
      }

    /* Fill in the ket */

    *code = OP_KET;
    PUT(code, 1, (int)(code - start_bracket));
    code += 1 + LINK_SIZE;

    /* If it was a capturing subpattern, check to see if it contained any
    recursive back references. If so, we must wrap it in atomic brackets. In
    any event, remove the block from the chain. */

    if (capnumber > 0)
      {
      if (cb->open_caps->flag)
        {
        (void)memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
          CU2BYTES(code - start_bracket));
        *start_bracket = OP_ONCE;
        code += 1 + LINK_SIZE;
        PUT(start_bracket, 1, (int)(code - start_bracket));
        *code = OP_KET;
        PUT(code, 1, (int)(code - start_bracket));
        code += 1 + LINK_SIZE;
        length += 2 + 2*LINK_SIZE;
        }
      cb->open_caps = cb->open_caps->next;
      }

    /* Set values to pass back */

    *codeptr = code;
    *pptrptr = pptr;
    *firstcuptr = firstcu;
    *firstcuflagsptr = firstcuflags;
    *reqcuptr = reqcu;
    *reqcuflagsptr = reqcuflags;
    if (lengthptr != NULL)
      {
      if (OFLOW_MAX - *lengthptr < length)
        {
        *errorcodeptr = ERR20;
        return 0;
        }
      *lengthptr += length;
      }
    return okreturn;
    }

  /* Another branch follows. In the pre-compile phase, we can move the code
  pointer back to where it was for the start of the first branch. (That is,
  pretend that each branch is the only one.)

  In the real compile phase, insert an ALT node. Its length field points back
  to the previous branch while the bracket remains open. At the end the chain
  is reversed. It's done like this so that the start of the bracket has a
  zero offset until it is closed, making it possible to detect recursion. */

  if (lengthptr != NULL)
    {
    code = *codeptr + 1 + LINK_SIZE + skipunits;
    length += 1 + LINK_SIZE;
    }
  else
    {
    *code = OP_ALT;
    PUT(code, 1, (int)(code - last_branch));
    bc.current_branch = last_branch = code;
    code += 1 + LINK_SIZE;
    }

  /* Set the lookbehind length (if not in a lookbehind the value will be zero)
  and then advance past the vertical bar. */

  lookbehindlength = META_DATA(*pptr);
  pptr++;
  }
/* Control never reaches here */
}



/*************************************************
*          Check for anchored pattern            *
*************************************************/

/* Try to find out if this is an anchored regular expression. Consider each
alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
it's anchored. However, if this is a multiline pattern, then only OP_SOD will
be found, because ^ generates OP_CIRCM in that mode.

We can also consider a regex to be anchored if OP_SOM starts all its branches.
This is the code for \G, which means "match at start of match position, taking
into account the match offset".

A branch is also implicitly anchored if it starts with .* and DOTALL is set,
because that will try the rest of the pattern at all possible matching points,
so there is no point trying again.... er ....

.... except when the .* appears inside capturing parentheses, and there is a
subsequent back reference to those parentheses. We haven't enough information
to catch that case precisely.

At first, the best we could do was to detect when .* was in capturing brackets
and the highest back reference was greater than or equal to that level.
However, by keeping a bitmap of the first 31 back references, we can catch some
of the more common cases more precisely.

... A second exception is when the .* appears inside an atomic group, because
this prevents the number of characters it matches from being adjusted.

Arguments:
  code           points to start of the compiled pattern
  bracket_map    a bitmap of which brackets we are inside while testing; this
                   handles up to substring 31; after that we just have to take
                   the less precise approach
  cb             points to the compile data block
  atomcount      atomic group level
  inassert       TRUE if in an assertion

Returns:     TRUE or FALSE
*/

static BOOL
is_anchored(PCRE2_SPTR code, unsigned int bracket_map, compile_block *cb,
  int atomcount, BOOL inassert)
{
do {
   PCRE2_SPTR scode = first_significant_code(
     code + PRIV(OP_lengths)[*code], FALSE);
   int op = *scode;

   /* Non-capturing brackets */

   if (op == OP_BRA  || op == OP_BRAPOS ||
       op == OP_SBRA || op == OP_SBRAPOS)
     {
     if (!is_anchored(scode, bracket_map, cb, atomcount, inassert))
       return FALSE;
     }

   /* Capturing brackets */

   else if (op == OP_CBRA  || op == OP_CBRAPOS ||
            op == OP_SCBRA || op == OP_SCBRAPOS)
     {
     int n = GET2(scode, 1+LINK_SIZE);
     int new_map = bracket_map | ((n < 32)? (1u << n) : 1);
     if (!is_anchored(scode, new_map, cb, atomcount, inassert)) return FALSE;
     }

   /* Positive forward assertion */

   else if (op == OP_ASSERT)
     {
     if (!is_anchored(scode, bracket_map, cb, atomcount, TRUE)) return FALSE;
     }

   /* Condition. If there is no second branch, it can't be anchored. */

   else if (op == OP_COND || op == OP_SCOND)
     {
     if (scode[GET(scode,1)] != OP_ALT) return FALSE;
     if (!is_anchored(scode, bracket_map, cb, atomcount, inassert))
       return FALSE;
     }

   /* Atomic groups */

   else if (op == OP_ONCE)
     {
     if (!is_anchored(scode, bracket_map, cb, atomcount + 1, inassert))
       return FALSE;
     }

   /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
   it isn't in brackets that are or may be referenced or inside an atomic
   group or an assertion. Also the pattern must not contain *PRUNE or *SKIP,
   because these break the feature. Consider, for example, /(?s).*?(*PRUNE)b/
   with the subject "aab", which matches "b", i.e. not at the start of a line.
   There is also an option that disables auto-anchoring. */

   else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
             op == OP_TYPEPOSSTAR))
     {
     if (scode[1] != OP_ALLANY || (bracket_map & cb->backref_map) != 0 ||
         atomcount > 0 || cb->had_pruneorskip || inassert ||
         (cb->external_options & PCRE2_NO_DOTSTAR_ANCHOR) != 0)
       return FALSE;
     }

   /* Check for explicit anchoring */

   else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;

   code += GET(code, 1);
   }
while (*code == OP_ALT);   /* Loop for each alternative */
return TRUE;
}



/*************************************************
*         Check for starting with ^ or .*        *
*************************************************/

/* This is called to find out if every branch starts with ^ or .* so that
"first char" processing can be done to speed things up in multiline
matching and for non-DOTALL patterns that start with .* (which must start at
the beginning or after \n). As in the case of is_anchored() (see above), we
have to take account of back references to capturing brackets that contain .*
because in that case we can't make the assumption. Also, the appearance of .*
inside atomic brackets or in an assertion, or in a pattern that contains *PRUNE
or *SKIP does not count, because once again the assumption no longer holds.

Arguments:
  code           points to start of the compiled pattern or a group
  bracket_map    a bitmap of which brackets we are inside while testing; this
                   handles up to substring 31; after that we just have to take
                   the less precise approach
  cb             points to the compile data
  atomcount      atomic group level
  inassert       TRUE if in an assertion

Returns:         TRUE or FALSE
*/

static BOOL
is_startline(PCRE2_SPTR code, unsigned int bracket_map, compile_block *cb,
  int atomcount, BOOL inassert)
{
do {
   PCRE2_SPTR scode = first_significant_code(
     code + PRIV(OP_lengths)[*code], FALSE);
   int op = *scode;

   /* If we are at the start of a conditional assertion group, *both* the
   conditional assertion *and* what follows the condition must satisfy the test
   for start of line. Other kinds of condition fail. Note that there may be an
   auto-callout at the start of a condition. */

   if (op == OP_COND)
     {
     scode += 1 + LINK_SIZE;

     if (*scode == OP_CALLOUT) scode += PRIV(OP_lengths)[OP_CALLOUT];
       else if (*scode == OP_CALLOUT_STR) scode += GET(scode, 1 + 2*LINK_SIZE);

     switch (*scode)
       {
       case OP_CREF:
       case OP_DNCREF:
       case OP_RREF:
       case OP_DNRREF:
       case OP_FAIL:
       case OP_FALSE:
       case OP_TRUE:
       return FALSE;

       default:     /* Assertion */
       if (!is_startline(scode, bracket_map, cb, atomcount, TRUE)) return FALSE;
       do scode += GET(scode, 1); while (*scode == OP_ALT);
       scode += 1 + LINK_SIZE;
       break;
       }
     scode = first_significant_code(scode, FALSE);
     op = *scode;
     }

   /* Non-capturing brackets */

   if (op == OP_BRA  || op == OP_BRAPOS ||
       op == OP_SBRA || op == OP_SBRAPOS)
     {
     if (!is_startline(scode, bracket_map, cb, atomcount, inassert))
       return FALSE;
     }

   /* Capturing brackets */

   else if (op == OP_CBRA  || op == OP_CBRAPOS ||
            op == OP_SCBRA || op == OP_SCBRAPOS)
     {
     int n = GET2(scode, 1+LINK_SIZE);
     int new_map = bracket_map | ((n < 32)? (1u << n) : 1);
     if (!is_startline(scode, new_map, cb, atomcount, inassert)) return FALSE;
     }

   /* Positive forward assertions */

   else if (op == OP_ASSERT)
     {
     if (!is_startline(scode, bracket_map, cb, atomcount, TRUE))
       return FALSE;
     }

   /* Atomic brackets */

   else if (op == OP_ONCE)
     {
     if (!is_startline(scode, bracket_map, cb, atomcount + 1, inassert))
       return FALSE;
     }

   /* .* means "start at start or after \n" if it isn't in atomic brackets or
   brackets that may be referenced or an assertion, and as long as the pattern
   does not contain *PRUNE or *SKIP, because these break the feature. Consider,
   for example, /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab",
   i.e. not at the start of a line. There is also an option that disables this
   optimization. */

   else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR || op == OP_TYPEPOSSTAR)
     {
     if (scode[1] != OP_ANY || (bracket_map & cb->backref_map) != 0 ||
         atomcount > 0 || cb->had_pruneorskip || inassert ||
         (cb->external_options & PCRE2_NO_DOTSTAR_ANCHOR) != 0)
       return FALSE;
     }

   /* Check for explicit circumflex; anything else gives a FALSE result. Note
   in particular that this includes atomic brackets OP_ONCE because the number
   of characters matched by .* cannot be adjusted inside them. */

   else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;

   /* Move on to the next alternative */

   code += GET(code, 1);
   }
while (*code == OP_ALT);  /* Loop for each alternative */
return TRUE;
}



/*************************************************
*   Scan compiled regex for recursion reference  *
*************************************************/

/* This function scans through a compiled pattern until it finds an instance of
OP_RECURSE.

Arguments:
  code        points to start of expression
  utf         TRUE in UTF mode

Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
*/

static PCRE2_SPTR
find_recurse(PCRE2_SPTR code, BOOL utf)
{
for (;;)
  {
  PCRE2_UCHAR c = *code;
  if (c == OP_END) return NULL;
  if (c == OP_RECURSE) return code;

  /* XCLASS is used for classes that cannot be represented just by a bit map.
  This includes negated single high-valued characters. CALLOUT_STR is used for
  callouts with string arguments. In both cases the length in the table is
  zero; the actual length is stored in the compiled code. */

  if (c == OP_XCLASS) code += GET(code, 1);
    else if (c == OP_CALLOUT_STR) code += GET(code, 1 + 2*LINK_SIZE);

  /* Otherwise, we can get the item's length from the table, except that for
  repeated character types, we have to test for \p and \P, which have an extra
  two code units of parameters, and for MARK/PRUNE/SKIP/THEN with an argument,
  we must add in its length. */

  else
    {
    switch(c)
      {
      case OP_TYPESTAR:
      case OP_TYPEMINSTAR:
      case OP_TYPEPLUS:
      case OP_TYPEMINPLUS:
      case OP_TYPEQUERY:
      case OP_TYPEMINQUERY:
      case OP_TYPEPOSSTAR:
      case OP_TYPEPOSPLUS:
      case OP_TYPEPOSQUERY:
      if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
      break;

      case OP_TYPEPOSUPTO:
      case OP_TYPEUPTO:
      case OP_TYPEMINUPTO:
      case OP_TYPEEXACT:
      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
        code += 2;
      break;

      case OP_MARK:
      case OP_COMMIT_ARG:
      case OP_PRUNE_ARG:
      case OP_SKIP_ARG:
      case OP_THEN_ARG:
      code += code[1];
      break;
      }

    /* Add in the fixed length from the table */

    code += PRIV(OP_lengths)[c];

    /* In UTF-8 and UTF-16 modes, opcodes that are followed by a character may
    be followed by a multi-unit character. The length in the table is a
    minimum, so we have to arrange to skip the extra units. */

#ifdef MAYBE_UTF_MULTI
    if (utf) switch(c)
      {
      case OP_CHAR:
      case OP_CHARI:
      case OP_NOT:
      case OP_NOTI:
      case OP_EXACT:
      case OP_EXACTI:
      case OP_NOTEXACT:
      case OP_NOTEXACTI:
      case OP_UPTO:
      case OP_UPTOI:
      case OP_NOTUPTO:
      case OP_NOTUPTOI:
      case OP_MINUPTO:
      case OP_MINUPTOI:
      case OP_NOTMINUPTO:
      case OP_NOTMINUPTOI:
      case OP_POSUPTO:
      case OP_POSUPTOI:
      case OP_NOTPOSUPTO:
      case OP_NOTPOSUPTOI:
      case OP_STAR:
      case OP_STARI:
      case OP_NOTSTAR:
      case OP_NOTSTARI:
      case OP_MINSTAR:
      case OP_MINSTARI:
      case OP_NOTMINSTAR:
      case OP_NOTMINSTARI:
      case OP_POSSTAR:
      case OP_POSSTARI:
      case OP_NOTPOSSTAR:
      case OP_NOTPOSSTARI:
      case OP_PLUS:
      case OP_PLUSI:
      case OP_NOTPLUS:
      case OP_NOTPLUSI:
      case OP_MINPLUS:
      case OP_MINPLUSI:
      case OP_NOTMINPLUS:
      case OP_NOTMINPLUSI:
      case OP_POSPLUS:
      case OP_POSPLUSI:
      case OP_NOTPOSPLUS:
      case OP_NOTPOSPLUSI:
      case OP_QUERY:
      case OP_QUERYI:
      case OP_NOTQUERY:
      case OP_NOTQUERYI:
      case OP_MINQUERY:
      case OP_MINQUERYI:
      case OP_NOTMINQUERY:
      case OP_NOTMINQUERYI:
      case OP_POSQUERY:
      case OP_POSQUERYI:
      case OP_NOTPOSQUERY:
      case OP_NOTPOSQUERYI:
      if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
      break;
      }
#else
    (void)(utf);  /* Keep compiler happy by referencing function argument */
#endif  /* MAYBE_UTF_MULTI */
    }
  }
}



/*************************************************
*    Check for asserted fixed first code unit    *
*************************************************/

/* During compilation, the "first code unit" settings from forward assertions
are discarded, because they can cause conflicts with actual literals that
follow. However, if we end up without a first code unit setting for an
unanchored pattern, it is worth scanning the regex to see if there is an
initial asserted first code unit. If all branches start with the same asserted
code unit, or with a non-conditional bracket all of whose alternatives start
with the same asserted code unit (recurse ad lib), then we return that code
unit, with the flags set to zero or REQ_CASELESS; otherwise return zero with
REQ_NONE in the flags.

Arguments:
  code       points to start of compiled pattern
  flags      points to the first code unit flags
  inassert   non-zero if in an assertion

Returns:     the fixed first code unit, or 0 with REQ_NONE in flags
*/

static uint32_t
find_firstassertedcu(PCRE2_SPTR code, int32_t *flags, uint32_t inassert)
{
uint32_t c = 0;
int cflags = REQ_NONE;

*flags = REQ_NONE;
do {
   uint32_t d;
   int dflags;
   int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
             *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? IMM2_SIZE:0;
   PCRE2_SPTR scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
   PCRE2_UCHAR op = *scode;

   switch(op)
     {
     default:
     return 0;

     case OP_BRA:
     case OP_BRAPOS:
     case OP_CBRA:
     case OP_SCBRA:
     case OP_CBRAPOS:
     case OP_SCBRAPOS:
     case OP_ASSERT:
     case OP_ONCE:
     case OP_SCRIPT_RUN:
     d = find_firstassertedcu(scode, &dflags, inassert + ((op==OP_ASSERT)?1:0));
     if (dflags < 0)
       return 0;
     if (cflags < 0) { c = d; cflags = dflags; }
       else if (c != d || cflags != dflags) return 0;
     break;

     case OP_EXACT:
     scode += IMM2_SIZE;
     /* Fall through */

     case OP_CHAR:
     case OP_PLUS:
     case OP_MINPLUS:
     case OP_POSPLUS:
     if (inassert == 0) return 0;
     if (cflags < 0) { c = scode[1]; cflags = 0; }
       else if (c != scode[1]) return 0;
     break;

     case OP_EXACTI:
     scode += IMM2_SIZE;
     /* Fall through */

     case OP_CHARI:
     case OP_PLUSI:
     case OP_MINPLUSI:
     case OP_POSPLUSI:
     if (inassert == 0) return 0;
     if (cflags < 0) { c = scode[1]; cflags = REQ_CASELESS; }
       else if (c != scode[1]) return 0;
     break;
     }

   code += GET(code, 1);
   }
while (*code == OP_ALT);

*flags = cflags;
return c;
}



/*************************************************
*     Add an entry to the name/number table      *
*************************************************/

/* This function is called between compiling passes to add an entry to the
name/number table, maintaining alphabetical order. Checking for permitted
and forbidden duplicates has already been done.

Arguments:
  cb           the compile data block
  name         the name to add
  length       the length of the name
  groupno      the group number
  tablecount   the count of names in the table so far

Returns:       nothing
*/

static void
add_name_to_table(compile_block *cb, PCRE2_SPTR name, int length,
  unsigned int groupno, uint32_t tablecount)
{
uint32_t i;
PCRE2_UCHAR *slot = cb->name_table;

for (i = 0; i < tablecount; i++)
  {
  int crc = memcmp(name, slot+IMM2_SIZE, CU2BYTES(length));
  if (crc == 0 && slot[IMM2_SIZE+length] != 0)
    crc = -1; /* Current name is a substring */

  /* Make space in the table and break the loop for an earlier name. For a
  duplicate or later name, carry on. We do this for duplicates so that in the
  simple case (when ?(| is not used) they are in order of their numbers. In all
  cases they are in the order in which they appear in the pattern. */

  if (crc < 0)
    {
    (void)memmove(slot + cb->name_entry_size, slot,
      CU2BYTES((tablecount - i) * cb->name_entry_size));
    break;
    }

  /* Continue the loop for a later or duplicate name */

  slot += cb->name_entry_size;
  }

PUT2(slot, 0, groupno);
memcpy(slot + IMM2_SIZE, name, CU2BYTES(length));

/* Add a terminating zero and fill the rest of the slot with zeroes so that
the memory is all initialized. Otherwise valgrind moans about uninitialized
memory when saving serialized compiled patterns. */

memset(slot + IMM2_SIZE + length, 0,
  CU2BYTES(cb->name_entry_size - length - IMM2_SIZE));
}



/*************************************************
*             Skip in parsed pattern             *
*************************************************/

/* This function is called to skip parts of the parsed pattern when finding the
length of a lookbehind branch. It is called after (*ACCEPT) and (*FAIL) to find
the end of the branch, it is called to skip over an internal lookaround, and it
is also called to skip to the end of a class, during which it will never
encounter nested groups (but there's no need to have special code for that).

When called to find the end of a branch or group, pptr must point to the first
meta code inside the branch, not the branch-starting code. In other cases it
can point to the item that causes the function to be called.

Arguments:
  pptr       current pointer to skip from
  skiptype   PSKIP_CLASS when skipping to end of class
             PSKIP_ALT when META_ALT ends the skip
             PSKIP_KET when only META_KET ends the skip

Returns:     new value of pptr
             NULL if META_END is reached - should never occur
               or for an unknown meta value - likewise
*/

static uint32_t *
parsed_skip(uint32_t *pptr, uint32_t skiptype)
{
uint32_t nestlevel = 0;

for (;; pptr++)
  {
  uint32_t meta = META_CODE(*pptr);

  switch(meta)
    {
    default:  /* Just skip over most items */
    if (meta < META_END) continue;  /* Literal */
    break;

    /* This should never occur. */

    case META_END:
    return NULL;

    /* The data for these items is variable in length. */

    case META_BACKREF:  /* Offset is present only if group >= 10 */
    if (META_DATA(*pptr) >= 10) pptr += SIZEOFFSET;
    break;

    case META_ESCAPE:   /* A few escapes are followed by data items. */
    switch (META_DATA(*pptr))
      {
      case ESC_P:
      case ESC_p:
      pptr += 1;
      break;

      case ESC_g:
      case ESC_k:
      pptr += 1 + SIZEOFFSET;
      break;
      }
    break;

    case META_MARK:     /* Add the length of the name. */
    case META_COMMIT_ARG:
    case META_PRUNE_ARG:
    case META_SKIP_ARG:
    case META_THEN_ARG:
    pptr += pptr[1];
    break;

    /* These are the "active" items in this loop. */

    case META_CLASS_END:
    if (skiptype == PSKIP_CLASS) return pptr;
    break;

    case META_ATOMIC:
    case META_CAPTURE:
    case META_COND_ASSERT:
    case META_COND_DEFINE:
    case META_COND_NAME:
    case META_COND_NUMBER:
    case META_COND_RNAME:
    case META_COND_RNUMBER:
    case META_COND_VERSION:
    case META_LOOKAHEAD:
    case META_LOOKAHEADNOT:
    case META_LOOKBEHIND:
    case META_LOOKBEHINDNOT:
    case META_NOCAPTURE:
    case META_SCRIPT_RUN:
    nestlevel++;
    break;

    case META_ALT:
    if (nestlevel == 0 && skiptype == PSKIP_ALT) return pptr;
    break;

    case META_KET:
    if (nestlevel == 0) return pptr;
    nestlevel--;
    break;
    }

  /* The extra data item length for each meta is in a table. */

  meta = (meta >> 16) & 0x7fff;
  if (meta >= sizeof(meta_extra_lengths)) return NULL;
  pptr += meta_extra_lengths[meta];
  }
/* Control never reaches here */
return pptr;
}



/*************************************************
*       Find length of a parsed group            *
*************************************************/

/* This is called for nested groups within a branch of a lookbehind whose
length is being computed. If all the branches in the nested group have the same
length, that is OK. On entry, the pointer must be at the first element after
the group initializing code. On exit it points to OP_KET. Caching is used to
improve processing speed when the same capturing group occurs many times.

Arguments:
  pptrptr     pointer to pointer in the parsed pattern
  isinline    FALSE if a reference or recursion; TRUE for inline group
  errcodeptr  pointer to the errorcode
  lcptr       pointer to the loop counter
  group       number of captured group or -1 for a non-capturing group
  recurses    chain of recurse_check to catch mutual recursion
  cb          pointer to the compile data

Returns:      the group length or a negative number
*/

static int
get_grouplength(uint32_t **pptrptr, BOOL isinline, int *errcodeptr, int *lcptr,
  int group, parsed_recurse_check *recurses, compile_block *cb)
{
int branchlength;
int grouplength = -1;

/* The cache can be used only if there is no possibility of there being two
groups with the same number. We do not need to set the end pointer for a group
that is being processed as a back reference or recursion, but we must do so for
an inline group. */

if (group > 0 && (cb->external_flags & PCRE2_DUPCAPUSED) == 0)
  {
  uint32_t groupinfo = cb->groupinfo[group];
  if ((groupinfo & GI_NOT_FIXED_LENGTH) != 0) return -1;
  if ((groupinfo & GI_SET_FIXED_LENGTH) != 0)
    {
    if (isinline) *pptrptr = parsed_skip(*pptrptr, PSKIP_KET);
    return groupinfo & GI_FIXED_LENGTH_MASK;
    }
  }

/* Scan the group. In this case we find the end pointer of necessity. */

for(;;)
  {
  branchlength = get_branchlength(pptrptr, errcodeptr, lcptr, recurses, cb);
  if (branchlength < 0) goto ISNOTFIXED;
  if (grouplength == -1) grouplength = branchlength;
    else if (grouplength != branchlength) goto ISNOTFIXED;
  if (**pptrptr == META_KET) break;
  *pptrptr += 1;   /* Skip META_ALT */
  }

if (group > 0)
  cb->groupinfo[group] |= (uint32_t)(GI_SET_FIXED_LENGTH | grouplength);
return grouplength;

ISNOTFIXED:
if (group > 0) cb->groupinfo[group] |= GI_NOT_FIXED_LENGTH;
return -1;
}



/*************************************************
*        Find length of a parsed branch          *
*************************************************/

/* Return a fixed length for a branch in a lookbehind, giving an error if the
length is not fixed. If any lookbehinds are encountered on the way, they get
their length set. On entry, *pptrptr points to the first element inside the
branch. On exit it is set to point to the ALT or KET.

Arguments:
  pptrptr     pointer to pointer in the parsed pattern
  errcodeptr  pointer to error code
  lcptr       pointer to loop counter
  recurses    chain of recurse_check to catch mutual recursion
  cb          pointer to compile block

Returns:      the length, or a negative value on error
*/

static int
get_branchlength(uint32_t **pptrptr, int *errcodeptr, int *lcptr,
  parsed_recurse_check *recurses, compile_block *cb)
{
int branchlength = 0;
int grouplength;
uint32_t lastitemlength = 0;
uint32_t *pptr = *pptrptr;
PCRE2_SIZE offset;
parsed_recurse_check this_recurse;

/* A large and/or complex regex can take too long to process. This can happen
more often when (?| groups are present in the pattern because their length
cannot be cached. */

if ((*lcptr)++ > 2000)
  {
  *errcodeptr = ERR35;  /* Lookbehind is too complicated */
  return -1;
  }

/* Scan the branch, accumulating the length. */

for (;; pptr++)
  {
  parsed_recurse_check *r;
  uint32_t *gptr, *gptrend;
  uint32_t escape;
  uint32_t group = 0;
  uint32_t itemlength = 0;

  if (*pptr < META_END)
    {
    itemlength = 1;
    }

  else switch (META_CODE(*pptr))
    {
    case META_KET:
    case META_ALT:
    goto EXIT;

    /* (*ACCEPT) and (*FAIL) terminate the branch, but we must skip to the
    actual termination. */

    case META_ACCEPT:
    case META_FAIL:
    pptr = parsed_skip(pptr, PSKIP_ALT);
    if (pptr == NULL) goto PARSED_SKIP_FAILED;
    goto EXIT;

    case META_MARK:
    case META_COMMIT_ARG:
    case META_PRUNE_ARG:
    case META_SKIP_ARG:
    case META_THEN_ARG:
    pptr += pptr[1] + 1;
    break;

    case META_CIRCUMFLEX:
    case META_COMMIT:
    case META_DOLLAR:
    case META_PRUNE:
    case META_SKIP:
    case META_THEN:
    break;

    case META_OPTIONS:
    pptr += 1;
    break;

    case META_BIGVALUE:
    itemlength = 1;
    pptr += 1;
    break;

    case META_CLASS:
    case META_CLASS_NOT:
    itemlength = 1;
    pptr = parsed_skip(pptr, PSKIP_CLASS);
    if (pptr == NULL) goto PARSED_SKIP_FAILED;
    break;

    case META_CLASS_EMPTY_NOT:
    case META_DOT:
    itemlength = 1;
    break;

    case META_CALLOUT_NUMBER:
    pptr += 3;
    break;

    case META_CALLOUT_STRING:
    pptr += 3 + SIZEOFFSET;
    break;

    /* Only some escapes consume a character. Of those, \R and \X are never
    allowed because they might match more than character. \C is allowed only in
    32-bit and non-UTF 8/16-bit modes. */

    case META_ESCAPE:
    escape = META_DATA(*pptr);
    if (escape == ESC_R || escape == ESC_X) return -1;
    if (escape > ESC_b && escape < ESC_Z)
      {
#if PCRE2_CODE_UNIT_WIDTH != 32
      if ((cb->external_options & PCRE2_UTF) != 0 && escape == ESC_C)
        {
        *errcodeptr = ERR36;
        return -1;
        }
#endif
      itemlength = 1;
      if (escape == ESC_p || escape == ESC_P) pptr++;  /* Skip prop data */
      }
    break;

    /* Lookaheads can be ignored, but we must start the skip inside the group
    so that it isn't treated as a group within the branch. */

    case META_LOOKAHEAD:
    case META_LOOKAHEADNOT:
    pptr = parsed_skip(pptr + 1, PSKIP_KET);
    if (pptr == NULL) goto PARSED_SKIP_FAILED;

    /* Also ignore any qualifiers that follow a lookahead assertion. */

    switch (pptr[1])
      {
      case META_ASTERISK:
      case META_ASTERISK_PLUS:
      case META_ASTERISK_QUERY:
      case META_PLUS:
      case META_PLUS_PLUS:
      case META_PLUS_QUERY:
      case META_QUERY:
      case META_QUERY_PLUS:
      case META_QUERY_QUERY:
      pptr++;
      break;

      case META_MINMAX:
      case META_MINMAX_PLUS:
      case META_MINMAX_QUERY:
      pptr += 3;
      break;

      default:
      break;
      }
    break;

    /* Lookbehinds can be ignored, but must themselves be checked. */

    case META_LOOKBEHIND:
    case META_LOOKBEHINDNOT:
    if (!set_lookbehind_lengths(&pptr, errcodeptr, lcptr, recurses, cb))
      return -1;
    break;

    /* Back references and recursions are handled by very similar code. At this
    stage, the names generated in the parsing pass are available, but the main
    name table has not yet been created. So for the named varieties, scan the
    list of names in order to get the number of the first one in the pattern,
    and whether or not this name is duplicated. */

    case META_BACKREF_BYNAME:
    if ((cb->external_options & PCRE2_MATCH_UNSET_BACKREF) != 0)
      goto ISNOTFIXED;
    /* Fall through */

    case META_RECURSE_BYNAME:
      {
      int i;
      PCRE2_SPTR name;
      BOOL is_dupname = FALSE;
      named_group *ng = cb->named_groups;
      uint32_t meta_code = META_CODE(*pptr);
      uint32_t length = *(++pptr);

      GETPLUSOFFSET(offset, pptr);
      name = cb->start_pattern + offset;
      for (i = 0; i < cb->names_found; i++, ng++)
        {
        if (length == ng->length && PRIV(strncmp)(name, ng->name, length) == 0)
          {
          group = ng->number;
          is_dupname = ng->isdup;
          break;
          }
        }

      if (group == 0)
        {
        *errcodeptr = ERR15;  /* Non-existent subpattern */
        cb->erroroffset = offset;
        return -1;
        }

      /* A numerical back reference can be fixed length if duplicate capturing
      groups are not being used. A non-duplicate named back reference can also
      be handled. */

      if (meta_code == META_RECURSE_BYNAME ||
          (!is_dupname && (cb->external_flags & PCRE2_DUPCAPUSED) == 0))
        goto RECURSE_OR_BACKREF_LENGTH;  /* Handle as a numbered version. */
      }
    goto ISNOTFIXED;                     /* Duplicate name or number */

    /* The offset values for back references < 10 are in a separate vector
    because otherwise they would use more than two parsed pattern elements on
    64-bit systems. */

    case META_BACKREF:
    if ((cb->external_options & PCRE2_MATCH_UNSET_BACKREF) != 0 ||
        (cb->external_flags & PCRE2_DUPCAPUSED) != 0)
      goto ISNOTFIXED;
    group = META_DATA(*pptr);
    if (group < 10)
      {
      offset = cb->small_ref_offset[group];
      goto RECURSE_OR_BACKREF_LENGTH;
      }

    /* Fall through */
    /* For groups >= 10 - picking up group twice does no harm. */

    /* A true recursion implies not fixed length, but a subroutine call may
    be OK. Back reference "recursions" are also failed. */

    case META_RECURSE:
    group = META_DATA(*pptr);
    GETPLUSOFFSET(offset, pptr);

    RECURSE_OR_BACKREF_LENGTH:
    if (group > cb->bracount)
      {
      cb->erroroffset = offset;
      *errcodeptr = ERR15;  /* Non-existent subpattern */
      return -1;
      }
    if (group == 0) goto ISNOTFIXED;  /* Local recursion */
    for (gptr = cb->parsed_pattern; *gptr != META_END; gptr++)
      {
      if (META_CODE(*gptr) == META_BIGVALUE) gptr++;
        else if (*gptr == (META_CAPTURE | group)) break;
      }

    /* We must start the search for the end of the group at the first meta code
    inside the group. Otherwise it will be treated as an enclosed group. */

    gptrend = parsed_skip(gptr + 1, PSKIP_KET);
    if (gptrend == NULL) goto PARSED_SKIP_FAILED;
    if (pptr > gptr && pptr < gptrend) goto ISNOTFIXED;  /* Local recursion */
    for (r = recurses; r != NULL; r = r->prev) if (r->groupptr == gptr) break;
    if (r != NULL) goto ISNOTFIXED;   /* Mutual recursion */
    this_recurse.prev = recurses;
    this_recurse.groupptr = gptr;

    /* We do not need to know the position of the end of the group, that is,
    gptr is not used after the call to get_grouplength(). Setting the second
    argument FALSE stops it scanning for the end when the length can be found
    in the cache. */

    gptr++;
    grouplength = get_grouplength(&gptr, FALSE, errcodeptr, lcptr, group,
      &this_recurse, cb);
    if (grouplength < 0)
      {
      if (*errcodeptr == 0) goto ISNOTFIXED;
      return -1;  /* Error already set */
      }
    itemlength = grouplength;
    break;

    /* Check nested groups - advance past the initial data for each type and
    then seek a fixed length with get_grouplength(). */

    case META_COND_NAME:
    case META_COND_NUMBER:
    case META_COND_RNAME:
    case META_COND_RNUMBER:
    case META_COND_DEFINE:
    pptr += 2 + SIZEOFFSET;
    goto CHECK_GROUP;

    case META_COND_ASSERT:
    pptr += 1;
    goto CHECK_GROUP;

    case META_COND_VERSION:
    pptr += 4;
    goto CHECK_GROUP;

    case META_CAPTURE:
    group = META_DATA(*pptr);
    /* Fall through */

    case META_ATOMIC:
    case META_NOCAPTURE:
    case META_SCRIPT_RUN:
    pptr++;
    CHECK_GROUP:
    grouplength = get_grouplength(&pptr, TRUE, errcodeptr, lcptr, group,
      recurses, cb);
    if (grouplength < 0) return -1;
    itemlength = grouplength;
    break;

    /* Exact repetition is OK; variable repetition is not. A repetition of zero
    must subtract the length that has already been added. */

    case META_MINMAX:
    case META_MINMAX_PLUS:
    case META_MINMAX_QUERY:
    if (pptr[1] == pptr[2])
      {
      if (pptr[1] == 0) branchlength -= lastitemlength;
        else itemlength = (pptr[1] - 1) * lastitemlength;
      pptr += 2;
      break;
      }
    /* Fall through */

    /* Any other item means this branch does not have a fixed length. */

    default:
    ISNOTFIXED:
    *errcodeptr = ERR25;   /* Not fixed length */
    return -1;
    }

  /* Add the item length to the branchlength, and save it for use if the next
  thing is a quantifier. */

  branchlength += itemlength;
  lastitemlength = itemlength;

  /* Ensure that the length does not overflow the limit. */

  if (branchlength > LOOKBEHIND_MAX)
    {
    *errcodeptr = ERR87;
    return -1;
    }
  }

EXIT:
*pptrptr = pptr;
if (branchlength > cb->max_lookbehind) cb->max_lookbehind = branchlength;
return branchlength;

PARSED_SKIP_FAILED:
*errcodeptr = ERR90;
return -1;
}



/*************************************************
*        Set lengths in a lookbehind             *
*************************************************/

/* This function is called for each lookbehind, to set the lengths in its
branches. An error occurs if any branch does not have a fixed length that is
less than the maximum (65535). On exit, the pointer must be left on the final
ket.

Arguments:
  pptrptr     pointer to pointer in the parsed pattern
  errcodeptr  pointer to error code
  lcptr       pointer to loop counter
  recurses    chain of recurse_check to catch mutual recursion
  cb          pointer to compile block

Returns:      TRUE if all is well
              FALSE otherwise, with error code and offset set
*/

static BOOL
set_lookbehind_lengths(uint32_t **pptrptr, int *errcodeptr, int *lcptr,
  parsed_recurse_check *recurses, compile_block *cb)
{
PCRE2_SIZE offset;
int branchlength;
uint32_t *bptr = *pptrptr;

READPLUSOFFSET(offset, bptr);  /* Offset for error messages */
*pptrptr += SIZEOFFSET;

do
  {
  *pptrptr += 1;
  branchlength = get_branchlength(pptrptr, errcodeptr, lcptr, recurses, cb);
  if (branchlength < 0)
    {
    /* The errorcode and offset may already be set from a nested lookbehind. */
    if (*errcodeptr == 0) *errcodeptr = ERR25;
    if (cb->erroroffset == PCRE2_UNSET) cb->erroroffset = offset;
    return FALSE;
    }
  *bptr |= branchlength;  /* branchlength never more than 65535 */
  bptr = *pptrptr;
  }
while (*bptr == META_ALT);

return TRUE;
}



/*************************************************
*         Check parsed pattern lookbehinds       *
*************************************************/

/* This function is called at the end of parsing a pattern if any lookbehinds
were encountered. It scans the parsed pattern for them, calling
set_lookbehind_lengths() for each one. At the start, the errorcode is zero and
the error offset is marked unset. The enables the functions above not to
override settings from deeper nestings.

Arguments cb      points to the compile block
Returns:          0 on success, or an errorcode (cb->erroroffset will be set)
*/

static int
check_lookbehinds(compile_block *cb)
{
uint32_t *pptr;
int errorcode = 0;
int loopcount = 0;

cb->erroroffset = PCRE2_UNSET;

for (pptr = cb->parsed_pattern; *pptr != META_END; pptr++)
  {
  if (*pptr < META_END) continue;  /* Literal */

  switch (META_CODE(*pptr))
    {
    default:
    return ERR70;  /* Unrecognized meta code */

    case META_ESCAPE:
    if (*pptr - META_ESCAPE == ESC_P || *pptr - META_ESCAPE == ESC_p)
      pptr += 1;
    break;

    case META_ACCEPT:
    case META_ALT:
    case META_ASTERISK:
    case META_ASTERISK_PLUS:
    case META_ASTERISK_QUERY:
    case META_ATOMIC:
    case META_BACKREF:
    case META_CAPTURE:
    case META_CIRCUMFLEX:
    case META_CLASS:
    case META_CLASS_EMPTY:
    case META_CLASS_EMPTY_NOT:
    case META_CLASS_END:
    case META_CLASS_NOT:
    case META_COMMIT:
    case META_COND_ASSERT:
    case META_DOLLAR:
    case META_DOT:
    case META_FAIL:
    case META_KET:
    case META_LOOKAHEAD:
    case META_LOOKAHEADNOT:
    case META_NOCAPTURE:
    case META_PLUS:
    case META_PLUS_PLUS:
    case META_PLUS_QUERY:
    case META_PRUNE:
    case META_QUERY:
    case META_QUERY_PLUS:
    case META_QUERY_QUERY:
    case META_RANGE_ESCAPED:
    case META_RANGE_LITERAL:
    case META_SCRIPT_RUN:
    case META_SKIP:
    case META_THEN:
    break;

    case META_RECURSE:
    pptr += SIZEOFFSET;
    break;

    case META_BACKREF_BYNAME:
    case META_COND_DEFINE:
    case META_COND_NAME:
    case META_COND_NUMBER:
    case META_COND_RNAME:
    case META_COND_RNUMBER:
    case META_RECURSE_BYNAME:
    pptr += 1 + SIZEOFFSET;
    break;

    case META_CALLOUT_STRING:
    pptr += 3 + SIZEOFFSET;
    break;

    case META_BIGVALUE:
    case META_OPTIONS:
    case META_POSIX:
    case META_POSIX_NEG:
    pptr += 1;
    break;

    case META_MINMAX:
    case META_MINMAX_QUERY:
    case META_MINMAX_PLUS:
    pptr += 2;
    break;

    case META_CALLOUT_NUMBER:
    case META_COND_VERSION:
    pptr += 3;
    break;

    case META_MARK:
    case META_COMMIT_ARG:
    case META_PRUNE_ARG:
    case META_SKIP_ARG:
    case META_THEN_ARG:
    pptr += 1 + pptr[1];
    break;

    case META_LOOKBEHIND:
    case META_LOOKBEHINDNOT:
    if (!set_lookbehind_lengths(&pptr, &errorcode, &loopcount, NULL, cb))
      return errorcode;
    break;
    }
  }

return 0;
}



/*************************************************
*     External function to compile a pattern     *
*************************************************/

/* This function reads a regular expression in the form of a string and returns
a pointer to a block of store holding a compiled version of the expression.

Arguments:
  pattern       the regular expression
  patlen        the length of the pattern, or PCRE2_ZERO_TERMINATED
  options       option bits
  errorptr      pointer to errorcode
  erroroffset   pointer to error offset
  ccontext      points to a compile context or is NULL

Returns:        pointer to compiled data block, or NULL on error,
                with errorcode and erroroffset set
*/

PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION
pcre2_compile(PCRE2_SPTR pattern, PCRE2_SIZE patlen, uint32_t options,
   int *errorptr, PCRE2_SIZE *erroroffset, pcre2_compile_context *ccontext)
{
BOOL utf;                             /* Set TRUE for UTF mode */
BOOL has_lookbehind = FALSE;          /* Set TRUE if a lookbehind is found */
BOOL zero_terminated;                 /* Set TRUE for zero-terminated pattern */
pcre2_real_code *re = NULL;           /* What we will return */
compile_block cb;                     /* "Static" compile-time data */
const uint8_t *tables;                /* Char tables base pointer */

PCRE2_UCHAR *code;                    /* Current pointer in compiled code */
PCRE2_SPTR codestart;                 /* Start of compiled code */
PCRE2_SPTR ptr;                       /* Current pointer in pattern */
uint32_t *pptr;                       /* Current pointer in parsed pattern */

PCRE2_SIZE length = 1;                /* Allow for final END opcode */
PCRE2_SIZE usedlength;                /* Actual length used */
PCRE2_SIZE re_blocksize;              /* Size of memory block */
PCRE2_SIZE big32count = 0;            /* 32-bit literals >= 0x80000000 */
PCRE2_SIZE parsed_size_needed;        /* Needed for parsed pattern */

int32_t firstcuflags, reqcuflags;     /* Type of first/req code unit */
uint32_t firstcu, reqcu;              /* Value of first/req code unit */
uint32_t setflags = 0;                /* NL and BSR set flags */

uint32_t skipatstart;                 /* When checking (*UTF) etc */
uint32_t limit_heap  = UINT32_MAX;
uint32_t limit_match = UINT32_MAX;    /* Unset match limits */
uint32_t limit_depth = UINT32_MAX;

int newline = 0;                      /* Unset; can be set by the pattern */
int bsr = 0;                          /* Unset; can be set by the pattern */
int errorcode = 0;                    /* Initialize to avoid compiler warn */
int regexrc;                          /* Return from compile */

uint32_t i;                           /* Local loop counter */

/* Comments at the head of this file explain about these variables. */

uint32_t stack_groupinfo[GROUPINFO_DEFAULT_SIZE];
uint32_t stack_parsed_pattern[PARSED_PATTERN_DEFAULT_SIZE];
named_group named_groups[NAMED_GROUP_LIST_SIZE];

/* The workspace is used in different ways in the different compiling phases.
It needs to be 16-bit aligned for the preliminary parsing scan. */

uint32_t c16workspace[C16_WORK_SIZE];
PCRE2_UCHAR *cworkspace = (PCRE2_UCHAR *)c16workspace;


/* -------------- Check arguments and set up the pattern ----------------- */

/* There must be error code and offset pointers. */

if (errorptr == NULL || erroroffset == NULL) return NULL;
*errorptr = ERR0;
*erroroffset = 0;

/* There must be a pattern! */

if (pattern == NULL)
  {
  *errorptr = ERR16;
  return NULL;
  }

/* A NULL compile context means "use a default context" */

if (ccontext == NULL)
  ccontext = (pcre2_compile_context *)(&PRIV(default_compile_context));

/* Check that all undefined public option bits are zero. */

if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0 ||
    (ccontext->extra_options & ~PUBLIC_COMPILE_EXTRA_OPTIONS) != 0)
  {
  *errorptr = ERR17;
  return NULL;
  }

if ((options & PCRE2_LITERAL) != 0 &&
    ((options & ~PUBLIC_LITERAL_COMPILE_OPTIONS) != 0 ||
     (ccontext->extra_options & ~PUBLIC_LITERAL_COMPILE_EXTRA_OPTIONS) != 0))
  {
  *errorptr = ERR92;
  return NULL;
  }

/* A zero-terminated pattern is indicated by the special length value
PCRE2_ZERO_TERMINATED. Check for an overlong pattern. */

if ((zero_terminated = (patlen == PCRE2_ZERO_TERMINATED)))
  patlen = PRIV(strlen)(pattern);

if (patlen > ccontext->max_pattern_length)
  {
  *errorptr = ERR88;
  return NULL;
  }

/* From here on, all returns from this function should end up going via the
EXIT label. */


/* ------------ Initialize the "static" compile data -------------- */

tables = (ccontext->tables != NULL)? ccontext->tables : PRIV(default_tables);

cb.lcc = tables + lcc_offset;          /* Individual */
cb.fcc = tables + fcc_offset;          /*   character */
cb.cbits = tables + cbits_offset;      /*      tables */
cb.ctypes = tables + ctypes_offset;

cb.assert_depth = 0;
cb.bracount = 0;
cb.cx = ccontext;
cb.dupnames = FALSE;
cb.end_pattern = pattern + patlen;
cb.erroroffset = 0;
cb.external_flags = 0;
cb.external_options = options;
cb.groupinfo = stack_groupinfo;
cb.had_recurse = FALSE;
cb.lastcapture = 0;
cb.max_lookbehind = 0;
cb.name_entry_size = 0;
cb.name_table = NULL;
cb.named_groups = named_groups;
cb.named_group_list_size = NAMED_GROUP_LIST_SIZE;
cb.names_found = 0;
cb.open_caps = NULL;
cb.parens_depth = 0;
cb.parsed_pattern = stack_parsed_pattern;
cb.req_varyopt = 0;
cb.start_code = cworkspace;
cb.start_pattern = pattern;
cb.start_workspace = cworkspace;
cb.workspace_size = COMPILE_WORK_SIZE;

/* Maximum back reference and backref bitmap. The bitmap records up to 31 back
references to help in deciding whether (.*) can be treated as anchored or not.
*/

cb.top_backref = 0;
cb.backref_map = 0;

/* Escape sequences \1 to \9 are always back references, but as they are only
two characters long, only two elements can be used in the parsed_pattern
vector. The first contains the reference, and we'd like to use the second to
record the offset in the pattern, so that forward references to non-existent
groups can be diagnosed later with an offset. However, on 64-bit systems,
PCRE2_SIZE won't fit. Instead, we have a vector of offsets for the first
occurrence of \1 to \9, indexed by the second parsed_pattern value. All other
references have enough space for the offset to be put into the parsed pattern.
*/

for (i = 0; i < 10; i++) cb.small_ref_offset[i] = PCRE2_UNSET;


/* --------------- Start looking at the pattern --------------- */

/* Unless PCRE2_LITERAL is set, check for global one-time option settings at
the start of the pattern, and remember the offset to the actual regex. With
valgrind support, make the terminator of a zero-terminated pattern
inaccessible. This catches bugs that would otherwise only show up for
non-zero-terminated patterns. */

#ifdef SUPPORT_VALGRIND
if (zero_terminated) VALGRIND_MAKE_MEM_NOACCESS(pattern + patlen, CU2BYTES(1));
#endif

ptr = pattern;
skipatstart = 0;

if ((options & PCRE2_LITERAL) == 0)
  {
  while (patlen - skipatstart >= 2 &&
         ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
         ptr[skipatstart+1] == CHAR_ASTERISK)
    {
    for (i = 0; i < sizeof(pso_list)/sizeof(pso); i++)
      {
      uint32_t c, pp;
      pso *p = pso_list + i;

      if (patlen - skipatstart - 2 >= p->length &&
          PRIV(strncmp_c8)(ptr + skipatstart + 2, (char *)(p->name),
            p->length) == 0)
        {
        skipatstart += p->length + 2;
        switch(p->type)
          {
          case PSO_OPT:
          cb.external_options |= p->value;
          break;

          case PSO_FLG:
          setflags |= p->value;
          break;

          case PSO_NL:
          newline = p->value;
          setflags |= PCRE2_NL_SET;
          break;

          case PSO_BSR:
          bsr = p->value;
          setflags |= PCRE2_BSR_SET;
          break;

          case PSO_LIMM:
          case PSO_LIMD:
          case PSO_LIMH:
          c = 0;
          pp = skipatstart;
          if (!IS_DIGIT(ptr[pp]))
            {
            errorcode = ERR60;
            ptr += pp;
            goto HAD_EARLY_ERROR;
            }
          while (IS_DIGIT(ptr[pp]))
            {
            if (c > UINT32_MAX / 10 - 1) break;   /* Integer overflow */
            c = c*10 + (ptr[pp++] - CHAR_0);
            }
          if (ptr[pp++] != CHAR_RIGHT_PARENTHESIS)
            {
            errorcode = ERR60;
            ptr += pp;
            goto HAD_EARLY_ERROR;
            }
          if (p->type == PSO_LIMH) limit_heap = c;
            else if (p->type == PSO_LIMM) limit_match = c;
            else limit_depth = c;
          skipatstart += pp - skipatstart;
          break;
          }
        break;   /* Out of the table scan loop */
        }
      }
    if (i >= sizeof(pso_list)/sizeof(pso)) break;   /* Out of pso loop */
    }
  }

/* End of pattern-start options; advance to start of real regex. */

ptr += skipatstart;

/* Can't support UTF or UCP unless PCRE2 has been compiled with UTF support. */

#ifndef SUPPORT_UNICODE
if ((cb.external_options & (PCRE2_UTF|PCRE2_UCP)) != 0)
  {
  errorcode = ERR32;
  goto HAD_EARLY_ERROR;
  }
#endif

/* Check UTF. We have the original options in 'options', with that value as
modified by (*UTF) etc in cb->external_options. The extra option
PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES is not permitted in UTF-16 mode because the
surrogate code points cannot be represented in UTF-16. */

utf = (cb.external_options & PCRE2_UTF) != 0;
if (utf)
  {
  if ((options & PCRE2_NEVER_UTF) != 0)
    {
    errorcode = ERR74;
    goto HAD_EARLY_ERROR;
    }
  if ((options & PCRE2_NO_UTF_CHECK) == 0 &&
       (errorcode = PRIV(valid_utf)(pattern, patlen, erroroffset)) != 0)
    goto HAD_ERROR;  /* Offset was set by valid_utf() */

#if PCRE2_CODE_UNIT_WIDTH == 16
  if ((ccontext->extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) != 0)
    {
    errorcode = ERR91;
    goto HAD_EARLY_ERROR;
    }
#endif
  }

/* Check UCP lockout. */

if ((cb.external_options & (PCRE2_UCP|PCRE2_NEVER_UCP)) ==
    (PCRE2_UCP|PCRE2_NEVER_UCP))
  {
  errorcode = ERR75;
  goto HAD_EARLY_ERROR;
  }

/* Process the BSR setting. */

if (bsr == 0) bsr = ccontext->bsr_convention;

/* Process the newline setting. */

if (newline == 0) newline = ccontext->newline_convention;
cb.nltype = NLTYPE_FIXED;
switch(newline)
  {
  case PCRE2_NEWLINE_CR:
  cb.nllen = 1;
  cb.nl[0] = CHAR_CR;
  break;

  case PCRE2_NEWLINE_LF:
  cb.nllen = 1;
  cb.nl[0] = CHAR_NL;
  break;

  case PCRE2_NEWLINE_NUL:
  cb.nllen = 1;
  cb.nl[0] = CHAR_NUL;
  break;

  case PCRE2_NEWLINE_CRLF:
  cb.nllen = 2;
  cb.nl[0] = CHAR_CR;
  cb.nl[1] = CHAR_NL;
  break;

  case PCRE2_NEWLINE_ANY:
  cb.nltype = NLTYPE_ANY;
  break;

  case PCRE2_NEWLINE_ANYCRLF:
  cb.nltype = NLTYPE_ANYCRLF;
  break;

  default:
  errorcode = ERR56;
  goto HAD_EARLY_ERROR;
  }

/* Pre-scan the pattern to do two things: (1) Discover the named groups and
their numerical equivalents, so that this information is always available for
the remaining processing. (2) At the same time, parse the pattern and put a
processed version into the parsed_pattern vector. This has escapes interpreted
and comments removed (amongst other things).

In all but one case, when PCRE2_AUTO_CALLOUT is not set, the number of unsigned
32-bit ints in the parsed pattern is bounded by the length of the pattern plus
one (for the terminator) plus four if PCRE2_EXTRA_WORD or PCRE2_EXTRA_LINE is
set. The exceptional case is when running in 32-bit, non-UTF mode, when literal
characters greater than META_END (0x80000000) have to be coded as two units. In
this case, therefore, we scan the pattern to check for such values. */

#if PCRE2_CODE_UNIT_WIDTH == 32
if (!utf)
  {
  PCRE2_SPTR p;
  for (p = ptr; p < cb.end_pattern; p++) if (*p >= META_END) big32count++;
  }
#endif

/* Ensure that the parsed pattern buffer is big enough. When PCRE2_AUTO_CALLOUT
is set we have to assume a numerical callout (4 elements) for each character
plus one at the end. This is overkill, but memory is plentiful these days. For
many smaller patterns the vector on the stack (which was set up above) can be
used. */

parsed_size_needed = patlen - skipatstart + big32count;

if ((ccontext->extra_options &
     (PCRE2_EXTRA_MATCH_WORD|PCRE2_EXTRA_MATCH_LINE)) != 0)
  parsed_size_needed += 4;

if ((options & PCRE2_AUTO_CALLOUT) != 0)
  parsed_size_needed = (parsed_size_needed + 1) * 5;

if (parsed_size_needed >= PARSED_PATTERN_DEFAULT_SIZE)
  {
  uint32_t *heap_parsed_pattern = ccontext->memctl.malloc(
    (parsed_size_needed + 1) * sizeof(uint32_t), ccontext->memctl.memory_data);
  if (heap_parsed_pattern == NULL)
    {
    *errorptr = ERR21;
    goto EXIT;
    }
  cb.parsed_pattern = heap_parsed_pattern;
  }
cb.parsed_pattern_end = cb.parsed_pattern + parsed_size_needed + 1;

/* Do the parsing scan. */

errorcode = parse_regex(ptr, cb.external_options, &has_lookbehind, &cb);
if (errorcode != 0) goto HAD_CB_ERROR;

/* Workspace is needed to remember information about numbered groups: whether a
group can match an empty string and what its fixed length is. This is done to
avoid the possibility of recursive references causing very long compile times
when checking these features. Unnumbered groups do not have this exposure since
they cannot be referenced. We use an indexed vector for this purpose. If there
are sufficiently few groups, the default vector on the stack, as set up above,
can be used. Otherwise we have to get/free a special vector. The vector must be
initialized to zero. */

if (cb.bracount >= GROUPINFO_DEFAULT_SIZE)
  {
  cb.groupinfo = ccontext->memctl.malloc(
    (cb.bracount + 1)*sizeof(uint32_t), ccontext->memctl.memory_data);
  if (cb.groupinfo == NULL)
    {
    errorcode = ERR21;
    cb.erroroffset = 0;
    goto HAD_CB_ERROR;
    }
  }
memset(cb.groupinfo, 0, (cb.bracount + 1) * sizeof(uint32_t));

/* If there were any lookbehinds, scan the parsed pattern to figure out their
lengths. */

if (has_lookbehind)
  {
  errorcode = check_lookbehinds(&cb);
  if (errorcode != 0) goto HAD_CB_ERROR;
  }

/* For debugging, there is a function that shows the parsed data vector. */

#ifdef DEBUG_SHOW_PARSED
fprintf(stderr, "+++ Pre-scan complete:\n");
show_parsed(&cb);
#endif

/* For debugging capturing information this code can be enabled. */

#ifdef DEBUG_SHOW_CAPTURES
  {
  named_group *ng = cb.named_groups;
  fprintf(stderr, "+++Captures: %d\n", cb.bracount);
  for (i = 0; i < cb.names_found; i++, ng++)
    {
    fprintf(stderr, "+++%3d %.*s\n", ng->number, ng->length, ng->name);
    }
  }
#endif

/* Pretend to compile the pattern while actually just accumulating the amount
of memory required in the 'length' variable. This behaviour is triggered by
passing a non-NULL final argument to compile_regex(). We pass a block of
workspace (cworkspace) for it to compile parts of the pattern into; the
compiled code is discarded when it is no longer needed, so hopefully this
workspace will never overflow, though there is a test for its doing so.

On error, errorcode will be set non-zero, so we don't need to look at the
result of the function. The initial options have been put into the cb block,
but we still have to pass a separate options variable (the first argument)
because the options may change as the pattern is processed. */

cb.erroroffset = patlen;   /* For any subsequent errors that do not set it */
pptr = cb.parsed_pattern;
code = cworkspace;
*code = OP_BRA;

(void)compile_regex(cb.external_options, &code, &pptr, &errorcode, 0, &firstcu,
   &firstcuflags, &reqcu, &reqcuflags, NULL, &cb, &length);

if (errorcode != 0) goto HAD_CB_ERROR;  /* Offset is in cb.erroroffset */

/* This should be caught in compile_regex(), but just in case... */

if (length > MAX_PATTERN_SIZE)
  {
  errorcode = ERR20;
  goto HAD_CB_ERROR;
  }

/* Compute the size of, and then get and initialize, the data block for storing
the compiled pattern and names table. Integer overflow should no longer be
possible because nowadays we limit the maximum value of cb.names_found and
cb.name_entry_size. */

re_blocksize = sizeof(pcre2_real_code) +
  CU2BYTES(length +
  (PCRE2_SIZE)cb.names_found * (PCRE2_SIZE)cb.name_entry_size);
re = (pcre2_real_code *)
  ccontext->memctl.malloc(re_blocksize, ccontext->memctl.memory_data);
if (re == NULL)
  {
  errorcode = ERR21;
  goto HAD_CB_ERROR;
  }

/* The compiler may put padding at the end of the pcre2_real_code structure in
order to round it up to a multiple of 4 or 8 bytes. This means that when a
compiled pattern is copied (for example, when serialized) undefined bytes are
read, and this annoys debuggers such as valgrind. To avoid this, we explicitly
write to the last 8 bytes of the structure before setting the fields. */

memset((char *)re + sizeof(pcre2_real_code) - 8, 0, 8);
re->memctl = ccontext->memctl;
re->tables = tables;
re->executable_jit = NULL;
memset(re->start_bitmap, 0, 32 * sizeof(uint8_t));
re->blocksize = re_blocksize;
re->magic_number = MAGIC_NUMBER;
re->compile_options = options;
re->overall_options = cb.external_options;
re->extra_options = ccontext->extra_options;
re->flags = PCRE2_CODE_UNIT_WIDTH/8 | cb.external_flags | setflags;
re->limit_heap = limit_heap;
re->limit_match = limit_match;
re->limit_depth = limit_depth;
re->first_codeunit = 0;
re->last_codeunit = 0;
re->bsr_convention = bsr;
re->newline_convention = newline;
re->max_lookbehind = 0;
re->minlength = 0;
re->top_bracket = 0;
re->top_backref = 0;
re->name_entry_size = cb.name_entry_size;
re->name_count = cb.names_found;

/* The basic block is immediately followed by the name table, and the compiled
code follows after that. */

codestart = (PCRE2_SPTR)((uint8_t *)re + sizeof(pcre2_real_code)) +
  re->name_entry_size * re->name_count;

/* Update the compile data block for the actual compile. The starting points of
the name/number translation table and of the code are passed around in the
compile data block. The start/end pattern and initial options are already set
from the pre-compile phase, as is the name_entry_size field. */

cb.parens_depth = 0;
cb.assert_depth = 0;
cb.lastcapture = 0;
cb.name_table = (PCRE2_UCHAR *)((uint8_t *)re + sizeof(pcre2_real_code));
cb.start_code = codestart;
cb.req_varyopt = 0;
cb.had_accept = FALSE;
cb.had_pruneorskip = FALSE;
cb.open_caps = NULL;

/* If any named groups were found, create the name/number table from the list
created in the pre-pass. */

if (cb.names_found > 0)
  {
  named_group *ng = cb.named_groups;
  for (i = 0; i < cb.names_found; i++, ng++)
    add_name_to_table(&cb, ng->name, ng->length, ng->number, i);
  }

/* Set up a starting, non-extracting bracket, then compile the expression. On
error, errorcode will be set non-zero, so we don't need to look at the result
of the function here. */

pptr = cb.parsed_pattern;
code = (PCRE2_UCHAR *)codestart;
*code = OP_BRA;
regexrc = compile_regex(re->overall_options, &code, &pptr, &errorcode, 0,
  &firstcu, &firstcuflags, &reqcu, &reqcuflags, NULL, &cb, NULL);
if (regexrc < 0) re->flags |= PCRE2_MATCH_EMPTY;
re->top_bracket = cb.bracount;
re->top_backref = cb.top_backref;
re->max_lookbehind = cb.max_lookbehind;

if (cb.had_accept)
  {
  reqcu = 0;              /* Must disable after (*ACCEPT) */
  reqcuflags = REQ_NONE;
  }

/* Fill in the final opcode and check for disastrous overflow. If no overflow,
but the estimated length exceeds the really used length, adjust the value of
re->blocksize, and if valgrind support is configured, mark the extra allocated
memory as unaddressable, so that any out-of-bound reads can be detected. */

*code++ = OP_END;
usedlength = code - codestart;
if (usedlength > length) errorcode = ERR23; else
  {
  re->blocksize -= CU2BYTES(length - usedlength);
#ifdef SUPPORT_VALGRIND
  VALGRIND_MAKE_MEM_NOACCESS(code, CU2BYTES(length - usedlength));
#endif
  }

/* Scan the pattern for recursion/subroutine calls and convert the group
numbers into offsets. Maintain a small cache so that repeated groups containing
recursions are efficiently handled. */

#define RSCAN_CACHE_SIZE 8

if (errorcode == 0 && cb.had_recurse)
  {
  PCRE2_UCHAR *rcode;
  PCRE2_SPTR rgroup;
  unsigned int ccount = 0;
  int start = RSCAN_CACHE_SIZE;
  recurse_cache rc[RSCAN_CACHE_SIZE];

  for (rcode = (PCRE2_UCHAR *)find_recurse(codestart, utf);
       rcode != NULL;
       rcode = (PCRE2_UCHAR *)find_recurse(rcode + 1 + LINK_SIZE, utf))
    {
    int p, groupnumber;

    groupnumber = (int)GET(rcode, 1);
    if (groupnumber == 0) rgroup = codestart; else
      {
      PCRE2_SPTR search_from = codestart;
      rgroup = NULL;
      for (i = 0, p = start; i < ccount; i++, p = (p + 1) & 7)
        {
        if (groupnumber == rc[p].groupnumber)
          {
          rgroup = rc[p].group;
          break;
          }

        /* Group n+1 must always start to the right of group n, so we can save
        search time below when the new group number is greater than any of the
        previously found groups. */

        if (groupnumber > rc[p].groupnumber) search_from = rc[p].group;
        }

      if (rgroup == NULL)
        {
        rgroup = PRIV(find_bracket)(search_from, utf, groupnumber);
        if (rgroup == NULL)
          {
          errorcode = ERR53;
          break;
          }
        if (--start < 0) start = RSCAN_CACHE_SIZE - 1;
        rc[start].groupnumber = groupnumber;
        rc[start].group = rgroup;
        if (ccount < RSCAN_CACHE_SIZE) ccount++;
        }
      }

    PUT(rcode, 1, rgroup - codestart);
    }
  }

/* In rare debugging situations we sometimes need to look at the compiled code
at this stage. */

#ifdef DEBUG_CALL_PRINTINT
pcre2_printint(re, stderr, TRUE);
fprintf(stderr, "Length=%lu Used=%lu\n", length, usedlength);
#endif

/* Unless disabled, check whether any single character iterators can be
auto-possessified. The function overwrites the appropriate opcode values, so
the type of the pointer must be cast. NOTE: the intermediate variable "temp" is
used in this code because at least one compiler gives a warning about loss of
"const" attribute if the cast (PCRE2_UCHAR *)codestart is used directly in the
function call. */

if (errorcode == 0 && (re->overall_options & PCRE2_NO_AUTO_POSSESS) == 0)
  {
  PCRE2_UCHAR *temp = (PCRE2_UCHAR *)codestart;
  if (PRIV(auto_possessify)(temp, utf, &cb) != 0) errorcode = ERR80;
  }

/* Failed to compile, or error while post-processing. */

if (errorcode != 0) goto HAD_CB_ERROR;

/* Successful compile. If the anchored option was not passed, set it if
we can determine that the pattern is anchored by virtue of ^ characters or \A
or anything else, such as starting with non-atomic .* when DOTALL is set and
there are no occurrences of *PRUNE or *SKIP (though there is an option to
disable this case). */

if ((re->overall_options & PCRE2_ANCHORED) == 0 &&
     is_anchored(codestart, 0, &cb, 0, FALSE))
  re->overall_options |= PCRE2_ANCHORED;

/* Set up the first code unit or startline flag, the required code unit, and
then study the pattern. This code need not be obeyed if PCRE2_NO_START_OPTIMIZE
is set, as the data it would create will not be used. Note that a first code
unit (but not the startline flag) is useful for anchored patterns because it
can still give a quick "no match" and also avoid searching for a last code
unit. */

if ((re->overall_options & PCRE2_NO_START_OPTIMIZE) == 0)
  {
  /* If we do not have a first code unit, see if there is one that is asserted
  (these are not saved during the compile because they can cause conflicts with
  actual literals that follow). */

  if (firstcuflags < 0)
    firstcu = find_firstassertedcu(codestart, &firstcuflags, 0);

  /* Save the data for a first code unit. */

  if (firstcuflags >= 0)
    {
    re->first_codeunit = firstcu;
    re->flags |= PCRE2_FIRSTSET;

    /* Handle caseless first code units. */

    if ((firstcuflags & REQ_CASELESS) != 0)
      {
      if (firstcu < 128 || (!utf && firstcu < 255))
        {
        if (cb.fcc[firstcu] != firstcu) re->flags |= PCRE2_FIRSTCASELESS;
        }

      /* The first code unit is > 128 in UTF mode, or > 255 otherwise. In
      8-bit UTF mode, codepoints in the range 128-255 are introductory code
      points and cannot have another case. In 16-bit and 32-bit modes, we can
      check wide characters when UTF (and therefore UCP) is supported. */

#if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 8
      else if (firstcu <= MAX_UTF_CODE_POINT &&
               UCD_OTHERCASE(firstcu) != firstcu)
        re->flags |= PCRE2_FIRSTCASELESS;
#endif
      }
    }

  /* When there is no first code unit, for non-anchored patterns, see if we can
  set the PCRE2_STARTLINE flag. This is helpful for multiline matches when all
  branches start with ^ and also when all branches start with non-atomic .* for
  non-DOTALL matches when *PRUNE and SKIP are not present. (There is an option
  that disables this case.) */

  else if ((re->overall_options & PCRE2_ANCHORED) == 0 &&
           is_startline(codestart, 0, &cb, 0, FALSE))
    re->flags |= PCRE2_STARTLINE;

  /* Handle the "required code unit", if one is set. In the case of an anchored
  pattern, do this only if it follows a variable length item in the pattern. */

  if (reqcuflags >= 0 &&
       ((re->overall_options & PCRE2_ANCHORED) == 0 ||
        (reqcuflags & REQ_VARY) != 0))
    {
    re->last_codeunit = reqcu;
    re->flags |= PCRE2_LASTSET;

    /* Handle caseless required code units as for first code units (above). */

    if ((reqcuflags & REQ_CASELESS) != 0)
      {
      if (reqcu < 128 || (!utf && reqcu < 255))
        {
        if (cb.fcc[reqcu] != reqcu) re->flags |= PCRE2_LASTCASELESS;
        }
#if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 8
      else if (reqcu <= MAX_UTF_CODE_POINT && UCD_OTHERCASE(reqcu) != reqcu)
        re->flags |= PCRE2_LASTCASELESS;
#endif
      }
    }

  /* Finally, study the compiled pattern to set up information such as a bitmap
  of starting code units and a minimum matching length. */

  if (PRIV(study)(re) != 0)
    {
    errorcode = ERR31;
    goto HAD_CB_ERROR;
    }
  }   /* End of start-of-match optimizations. */

/* Control ends up here in all cases. When running under valgrind, make a
pattern's terminating zero defined again. If memory was obtained for the parsed
version of the pattern, free it before returning. Also free the list of named
groups if a larger one had to be obtained, and likewise the group information
vector. */

EXIT:
#ifdef SUPPORT_VALGRIND
if (zero_terminated) VALGRIND_MAKE_MEM_DEFINED(pattern + patlen, CU2BYTES(1));
#endif
if (cb.parsed_pattern != stack_parsed_pattern)
  ccontext->memctl.free(cb.parsed_pattern, ccontext->memctl.memory_data);
if (cb.named_group_list_size > NAMED_GROUP_LIST_SIZE)
  ccontext->memctl.free((void *)cb.named_groups, ccontext->memctl.memory_data);
if (cb.groupinfo != stack_groupinfo)
  ccontext->memctl.free((void *)cb.groupinfo, ccontext->memctl.memory_data);
return re;    /* Will be NULL after an error */

/* Errors discovered in parse_regex() set the offset value in the compile
block. Errors discovered before it is called must compute it from the ptr
value. After parse_regex() is called, the offset in the compile block is set to
the end of the pattern, but certain errors in compile_regex() may reset it if
an offset is available in the parsed pattern. */

HAD_CB_ERROR:
ptr = pattern + cb.erroroffset;

HAD_EARLY_ERROR:
*erroroffset = ptr - pattern;

HAD_ERROR:
*errorptr = errorcode;
pcre2_code_free(re);
re = NULL;
goto EXIT;
}

/* End of pcre2_compile.c */