summaryrefslogtreecommitdiffstats
path: root/src/gui/text/qtextengine.cpp
Commit message (Collapse)AuthorAgeFilesLines
* Fix the font engines leakingKonstantin Ritt2013-03-201-5/+2
| | | | | | | | | | | | | | | 1. when there were some engines with ref > 1 in the cache, prior to calling QFontDatabase::{add,remove}ApplicationFont()/removeAllApplicationFonts() (QFontCache::clear() has never decreased engine's cache_count); 2. when the QFontEngineData's engine is not in cache i.e. the Box or Test font engine (~QFontEngineData() didn't free engines it keeps). Instead of using the font engine's (external) "cache_count" counter, QFontCache now references a given font engine every time it is inserted to the cache and dereferences exactly that number of times in clear(). Change-Id: I87677ebd24c1f4a81a53526f2e726e596b043c61 Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Move QTextEngine implementation details from QTextLayoutKonstantin Ritt2013-03-191-29/+71
| | | | | | | | | | | | | | | | As of now, we'll have everything related to the additional formats handling just in a single place. Make specialData private to prevent accessing it from outside. This helped in tracking-down several related issues: - in format(const QScriptItem *), the resolvedFormatIndices can not be empty at that point, so the code path is dead; - in resolveAdditionalFormats(), testing if formats has not been indexed yet is not needed since they are indexed just in the setter; - in useRawFont mode, hasFormats() didn't check if QTextEngine really has some formats, which potentially leads to formatting artifacts. Change-Id: Id4b912888fd5a1fa83f01007170134b6386e2879 Reviewed-by: Lars Knoll <lars.knoll@digia.com> Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
* QTextEngine capitalization doesn't work with QRawFontAllan Sandfeld Jensen2013-03-141-2/+3
| | | | | | | | | The capitalization settings is attempted read via specialData->addFormats, but the QTextCharFormat there is reset when added, and should instead be accessed through specialData->addFormatIndices. Change-Id: I190c419f07c3f7e803ca1d44059e8f538216b9ab Reviewed-by: Pierre Rossi <pierre.rossi@gmail.com>
* Move Unicode script itemization code from text engine to UnicodeToolsKonstantin Ritt2013-03-141-6/+10
| | | | | | | | | | | This is still the same trivial implementation with the only difference in that that it properly handles surrogate pairs and combining marks. This temporarily makes QTextEngine::itemize() insignificatly slower due to using intermediate buffer, until refactoring is done. Change-Id: I7987d6306b0b5cdb21b837968e292dd70abfe223 Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
* QTextEngine::itemize(): Minor optimizationKonstantin Ritt2013-03-131-1/+1
| | | | | | | | | Avoid using script_to_hbscript() in generateScriptItems()'s loop. This is absolutely unnecessary due to hbscript_to_script(script_to_hbscript(..)) in QTextEngine::itemize(). Change-Id: I44e710d010f8e57043a6c74dd80f3164e95e18e6 Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Avoid using HB types in QFontEngine APIKonstantin Ritt2013-03-131-2/+2
| | | | | | | | | | | This affects HB_Font and HB_Face. As of now, the Hurfbuzz API usage is concentrated in qfontengine(|_ft).cpp and qtextengine.cpp, thus it is a lot easier to switch to Hurfbuzz-NG. Change-Id: Ie06959efd5d6080fe44c407d9f5de0a07dd1c210 Reviewed-by: Lars Knoll <lars.knoll@digia.com> Reviewed-by: Josh Faust <jfaust@suitabletech.com>
* Replace HB types used in API with Qt analogsKonstantin Ritt2013-03-121-24/+21
| | | | | | | | | | hb_uint32 <-> quint32 HB_Glyph <-> glyph_t HB_Fixed <-> QFixed, HB_FixedPoint <-> QFixedPoint HB_GlyphAttributes <-> QGlyphAttributes Change-Id: I4338ec4ce9a9c20ea591606ab10e6ef4f19931b8 Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Merge remote-tracking branch 'origin/stable' into devFrederik Gladhorn2013-01-221-1/+1
|\ | | | | | | | | | | | | | | | | | | Conflicts: src/corelib/io/qsavefile_p.h src/corelib/tools/qregularexpression.cpp src/gui/util/qvalidator.cpp src/gui/util/qvalidator.h Change-Id: I58fdf0358bd86e2fad5d9ad0556f3d3f1f535825
| * Update copyright year in Digia's license headersSergio Ahumada2013-01-181-1/+1
| | | | | | | | | | Change-Id: Ic804938fc352291d011800d21e549c10acac66fb Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* | Merge branch 'stable' into devFrederik Gladhorn2013-01-071-1/+3
|\| | | | | | | | | | | | | Conflicts: tests/auto/widgets/graphicsview/qgraphicsscene/tst_qgraphicsscene.cpp Change-Id: I6b8d505fc22f052c307ca27f58f7d16f98965f47
| * Fix possible crash in QTextDocumentEskil Abrahamsen Blomfeldt2013-01-031-1/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | With trailing spaces in some cases, we would not get the "no justification at end of paragraph" special case, and continue in the code, getting the unexpected case where line_length becomes < 0 which would lead to memory corruption because we were writing outside our buffers. I added an assert to catch this type of bug earlier, and I added the trailing spaces to the test for the end of the paragraph. The test case added is one example which would crash. Task-number: QTBUG-27354 Change-Id: Id720a6fa55dbc709ce04dd5321e55687bf960d75 Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* | Update Qt internals to use QChar::ScriptKonstantin Ritt2012-12-211-15/+16
|/ | | | | | | | | | | | | | ...and remove the outdated QUnicodeTables::Script enum. QFontEngineData now has one extra slot that never used (engines[QChar::Script_Inherited]). engines[QChar::Script_Unknown], if accessed, would be set with a Box engine instance, and could be used as a minor optimization some time later. In order to preserve the existing behavior, we map all scripts up to Latin to Common. Change-Id: Ide4182a0f8447b4bf25713ecc3fe8097b8fed040 Reviewed-by: Pierre Rossi <pierre.rossi@gmail.com> Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
* QTextEngine: Support SMP code points when case changingKonstantin Ritt2012-10-251-4/+16
| | | | | | | Semi-related to QTBUG-17337 Change-Id: I6b42c0f7e588bbeab27bf410fcdfa1a6f80e4ac2 Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
* Revert hacks in text rendering code pathEskil Abrahamsen Blomfeldt2012-10-251-3/+12
| | | | | | | | | | | | | | There are a lot of hacks here and there in Qt trying to align the text in a correct way which caused regressions to appear once the default coordinate system changed. We need to remove these hacks to get a more consistent and maintainable base. This also fixes the regression introduced by changing the aliased coordinate system. Task-number: QTBUG-27667 Change-Id: I620db2ca23b7ff6c912f3a51e86e7e36bbef81f0 Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
* Get rid of unused memberKonstantin Ritt2012-10-161-2/+0
| | | | | | Change-Id: Ic6d207cacf02f2e6837f9f35020be7728ec60706 Reviewed-by: Lars Knoll <lars.knoll@digia.com> Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
* Use constFind()/constEnd() for const_iterator-sKonstantin Ritt2012-10-131-2/+2
| | | | | | | to make the strict iterators happy. Change-Id: Ief4ec309b815f18dc4b2017d4f34c063db510c31 Reviewed-by: Pierre Rossi <pierre.rossi@gmail.com>
* Update/fix QTextBoundaryFinder simple usage cases in qtbaseKonstantin Ritt2012-10-131-10/+10
| | | | | | Change-Id: I4d3000558bce86e2de3c32247915868ba18fc8b7 Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com> Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Revert "Handle additional format ranges when itemizing."Pierre Rossi2012-10-101-40/+23
| | | | | | | | | | | This reverts commit 101d04681f4ceb7410681eae684534a206a9d90a. That change seems to have introduced a few regressions, and Creator hits an assertion that it introduced because it assumes the additionalFormats consists of well-formed ranges (QTCREATORBUG-7995). Change-Id: Ic4ae761e6e7f6df8a6b5ca565ceb250647420c15 Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
* Handle additional format ranges when itemizing.Pierre Rossi2012-10-081-23/+40
| | | | | | | | | | | | | | This is useful when the additional formats are used on a text layout using a raw font. It can also come in handy for input methods operating on a QTextDocument. We now consider all format range edges to generate the associated items. The capitalization can be overridden via the additionnal formats mechanism. Adds an autotest that checks that this works with font capitalization. Change-Id: I782d2c48d05b0dfbad480a9ca77198465292b358 Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
* compile fix for MSVC 2008 and std::upper_boundThomas Hartmann2012-10-041-1/+4
| | | | | | | | | | | | qUpperBound was replaced by std::upper_bound. Unfortunately the STL of MSVC 2008 enforces the definition of the operator in both directions. Change-Id: I3e0f775c23e43332d106e0847d3611e488da6c06 Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com> Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
* Change copyrights from Nokia to DigiaIikka Eklund2012-09-221-24/+24
| | | | | | | | Change copyrights and license headers from Nokia to Digia Change-Id: If1cc974286d29fd01ec6c19dd4719a67f4c3f00e Reviewed-by: Lars Knoll <lars.knoll@digia.com> Reviewed-by: Sergio Ahumada <sergio.ahumada@digia.com>
* A step out from Harfbuzz (reduce dependency)Konstantin Ritt2012-09-221-32/+28
| | | | | | | | | | | | | | | | | Introduce QCharAttributes and use it instead of HB_CharAttributes everywhere in Qt (in Harfbuzz, the HB_CharAttributes is only used in the text segmentation algorithm which has been moved from HB to Qt (well, most of it)). Rename some members to better reflect their meaning, remember to keep HB_CharAttributes in sync with QCharAttributes. Also replace HB_ScriptItem with a (temporary) QUnicodeTools::ScriptItem struct that will be replaced with a more efficient/friendly solution a bit later. The soft hyphen and the mandatory break detection has been factored out of the default text breaking algorithm to a higher level in order to refactor the QCharAttributes bitfields and to optimize the implementation for the common case. Change-Id: Ieb365623ae954430f1c8b2dfcd65c82973143eec Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Remove qUpperBound usages from qtbaseGiuseppe D'Angelo2012-09-191-3/+4
| | | | | | | | | Replace them with std::upper_bound; this allows for deprecation of qUpperBound. Change-Id: Idef01d2228b9a70eee3d52931d7aedb5bb6ba902 Reviewed-by: Marc Mutz <marc.mutz@kdab.com> Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
* Fix crash when QTextLayout::setAdditionalFormats is called with ranges with ↵Olivier Goffart2012-09-071-2/+4
| | | | | | | | | | negative length That can be reproduced in Qt Creator when deleting code. Task-number: QTBUG-27140 Change-Id: Ida7177612653f10e5d866bf9a422c71c632f4eb3 Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
* Optimize QTextLayout/QTextEngine usage outside of QTextDocument.Milian Wolff2012-09-051-14/+18
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | When QTextLayout is used in a QTextDocument, many code paths use special caches and thus greatly outperform the raw QTextLayout version that operates directly on a QString. This patch brings some of these optimizations also to the raw version. We now also use a QFormatCollection in such cases and enable the functionality of QTextEngine::indexAdditionalFormats() and QTextEngine::resolveAdditionalFormats(). Thanks to that, we can greatly speed up QTextEngine::format(), which now uses an amort O(1) hash table lookup instead of a O(N) linear search. The added benchmark shows a gain in the order of one magnitude: ./tst_bench_QText formattedLayout:long-many before applying the patch: 378.19 msecs per iteration (total: 37,820, iterations: 100) after applying the patch: 25.80 msecs per iteration (total: 2,580, iterations: 100) Note: This change is source-incompatible for applications using the private QTextEngine API. Task-number: QTBUG-8389 Change-Id: Ifcf7a8902a394428979ea06a6d955f886ee739c7 Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com> Reviewed-by: Olivier Goffart <ogoffart@woboq.com> Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Speed up QTextEngine::setBoundaryOlivier Goffart2012-08-231-6/+14
| | | | | | | | | | | | | | Before, setBoundary was O(N), now it is O(log N) assuming the boundaries are added in order. (splitItem might still be linear because of the call to QVector::insert) This speeds up a lot QTextEngine::addRequiredBoundaries when there is a lot of addFormats. (For example if there is a very long line with syntax highlighting) Task-number: QTBUG-8389 Change-Id: I10c9fca72343f46db24b1d4f519f7747188d4009 Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Speed up QTextEngine::resolveAdditionalFormatsOlivier Goffart2012-08-231-2/+65
| | | | | | | | | Don't call the linear QTextEngine::format in the loop. Instead, keep track of the current formats by indexing their start and end position. Task-number: QTBUG-8389 Change-Id: I89c18b804111edfab6254442cbee33da39d1a273 Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Move ShaperFlags enum from QTextEngine to QFontEngineKonstantin Ritt2012-07-101-3/+3
| | | | | | | | These flags are specific to font engine(s) and has nothing to do with the text engine or the text layout. Change-Id: I4bb793c3c634b3cf0ae0a8a8c23b946fad5874b6 Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
* Add QChar::SoftHyphen enum valueKonstantin Ritt2012-07-031-1/+1
| | | | | | | | | Just like for the QChar::ByteOrderMark, `ch == QChar::SoftHyphen` is much more readable than `ch == 0x00ad // (soft-hyphen)`, etc. Change-Id: I9c85f14cfd979037d35103c3259a435fd729b869 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
* Fix capitalization with newline bugYann Bodson2012-07-031-1/+1
| | | | | | | | | We only want to change the capitalization if the QScriptAnalysis flag was Lowercase, Uppercase or SmallCaps. Task-number: QTBUG-17485 Change-Id: Icbecb09b06a9153866ae81d592b3f6779c2dafb5 Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* shapeTextWithHarfbuzz() minor optimizationKonstantin Ritt2012-07-031-5/+7
| | | | | | | | | | | If the used font engine is of type QFontEngine::Multi, then we already have both the "font engine runs" and the stripped glyph indicies for the kerning -> just call the actual font engine's doKerning() method. If the font engine is not of type QFontEngine::Multi, then we could safely avoid the glyph indicies extra manipulations. Change-Id: Ia8899e6d7beecfd391b050683ac41c21465336e4 Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
* Add const & to foreach 'iterators'Albert Astals Cid2012-06-271-1/+1
| | | | | | | Change-Id: I8c0600dfd919f45d14a0011f2da9b9fe0b9a0df3 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Marc Mutz <marc.mutz@kdab.com> Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
* Remove all references to XLFD fontsLars Knoll2012-06-271-6/+4
| | | | | | | | XLFD fonts are no longer supported in Qt 5. Change-Id: I83400dab417c933d5cd956c0d168c45b9d79dab7 Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com> Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
* Eliminate a warning in unsigned integer comparisonJiang Jiang2012-05-311-1/+1
| | | | | | | | The length we cast is from int QTextEngine::length() so it's safe to do so. Change-Id: I60fdbcb5a96c99b767093102a30e97951ef5b6ea Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* Fix text decoration width roundingJiang Jiang2012-05-301-1/+1
| | | | | | | | | | | | | | width passed in qreal should be set with setWidthF instead of setWidth, otherwise the fractional part is lost. Also remove the raster paint engine hack that has been removed in Qt 4.8, it's no longer needed. This fixed the drawUnderline and drawRightToLeft auto tests in tst_qglyphrun. Change-Id: Iace46fc27ff156c07fb4d08a145269fad36e9fdb Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
* Introduce QUnicodeToolsKonstantin Ritt2012-05-301-4/+4
| | | | | | | | | | | | | Add QUnicodeTools namespace and rename qGetCharAttributes to initCharAttributes; Make it possible to disable tailoring globally by overriding qt_initcharattributes_default_algorithm_only value (useful for i.e. running the specification conformance tests); This is mostly a preparation step for the upcoming patches. Change-Id: I783879fd17b63b52d7983e25dad5b820f0515e7f Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* QTextEngine: minor clean-upKonstantin Ritt2012-05-221-9/+14
| | | | | | Change-Id: I7e21d5cf0130b412f8053b35abdd3420ed006c1c Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com> Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* add some more widely-used QChar::SpecialCharacter enum valuesKonstantin Ritt2012-05-171-2/+2
| | | | | | Change-Id: Iad58f4366ba6cd6da29a268c56c8a4bc4cf0329c Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Fix cursorToX for right to left text with trailing whitespace.Andrew den Exter2012-05-171-11/+3
| | | | | | | | | | | | | | | | QTextLine::cursorToX returned the line width for cursor positions outside the width of a wrapped right to left line because the leading space width was always calculated as 0. Returning a non-zero width for the leading space does cause problems for other uses of QTextEngine::alignLine() though as the textAdvance already doesn't include the leading/trailing space so subtracting it there double accounts for it. Task-number: QTBUG-24801 Change-Id: I56cbb139814c32813bebb49de8c045b29154a958 Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
* move the default text breaking algorithm impl from HarfBuzz to QtKonstantin Ritt2012-05-101-0/+1
| | | | | | | | | | | there are several reasons to do this: * text breaking is not a shaper's job; * since the text breaking rules are bound to a specific Unicode version, updating Qt's internal unicode data would require updating the data in HB as well; * makes porting to HurfBuzz-NG some easier Change-Id: I0bbf8e8a343bc074696f4ddf2ae4e7fa32a61629 Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Don't ignore default font's font capitalization in QTextDocumentEskil Abrahamsen Blomfeldt2012-05-101-4/+6
| | | | | | | | | The consequence of this was e.g. that it would be impossible to set the capitalization of TextEdit elements (or Text elements with RichText format) in QML. Change-Id: Ie219fac7e4305efc59a4ac9da5b06c25ff3ac9ac Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
* QTextEngine: simplify the code a bitKonstantin Ritt2012-05-091-22/+7
| | | | | | | by using QVarLengthArray so we can avoid the manual clean-up code. Change-Id: I35e2f7150d777c1760f722553e6fe7a20f6ecc46 Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* Retain bidirectional marks when eliding textEskil Abrahamsen Blomfeldt2012-04-201-2/+44
| | | | | | | | | | | | When the text contains unicode markers for bidirectional text, these will not affect the width of the text, but they might affects its layout. Thus, we need to retain these markers in the elided text to avoid the layout of the string changing when its allocated width changes. Task-number: QTBUG-3768 Change-Id: I3866c32cd1dcd5fff2b8fa5b25ae89e551531a97 Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
* Fix memory leak when drawing text in non-common scriptEskil Abrahamsen Blomfeldt2012-04-181-3/+14
| | | | | | | | | | | | When the font engine cache was updated for new parameters, we would overwrite the cached engine, but not deref the old cached engine, thus it would never be deleted. This is manual cherry pick of 44c14470e5b28e45c19d6959b114e063cf3f4d83 from 4.8 branch. Change-Id: Ice2e555fe380dc9851ef490fa1a6928936ae4c35 Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
* Merge remote-tracking branch 'origin/master' into api_changesLars Knoll2012-04-161-1/+1
|\ | | | | | | | | | | | | | | | | | | | | | | | | | | Conflicts: configure src/corelib/io/qurl.cpp src/gui/kernel/qwindow.cpp src/tools/moc/generator.cpp src/widgets/kernel/qwidget_qpa.cpp src/widgets/styles/qstyle.h src/widgets/widgets/qtabbar.cpp tests/auto/corelib/codecs/utf8/tst_utf8.cpp Change-Id: Ia457228d6f684ec8184e13e8fcc9d25857b1751e
| * Fix crash with invalid pre-edit position.Andrew den Exter2012-04-101-1/+1
| | | | | | | | | | | | | | | | Ensure the script item position is within the bounds of the text block when adjusted to compensate for pre-edit text. Change-Id: I2c745bf08afc0d9bc3aba27c24c2a123af017dc6 Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* | Deprecate qMemCopy/qMemSet in favour of their stdlib equivilents.Robin Burchell2012-04-111-1/+1
|/ | | | | | | | | | | | | | | | | | Just like qMalloc/qRealloc/qFree, there is absolutely no reason to wrap these functions just to avoid an include, except to pay for it with worse runtime performance. On OS X, on byte sizes from 50 up to 1000, calling memset directly is 28-15% faster(!) than adding an additional call to qMemSet. The advantage on sizes above that is unmeasurable. For qMemCopy, the benefits are a little more modest: 16-7%. Change-Id: I98aa92bb765aea0448e3f20af42a039b369af0b3 Reviewed-by: Giuseppe D'Angelo <dangelog@gmail.com> Reviewed-by: John Brooks <john.brooks@dereferenced.net> Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
* Fix warnings about truncations in constants.Friedemann Kleint2012-04-031-1/+1
| | | | | Change-Id: I46872c5b2866454112092c1ec5efbfe15db5af33 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
* Line up underlines if fallback fonts are used (QTBUG-21832)Stefan Hundhammer2012-03-121-0/+99
| | | | | Change-Id: Icecc514f6c47c0576af8cabd39cdc0987f8d93fa Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>
* Make cache of opentype tables in Harfbuzz face lazyEskil Abrahamsen Blomfeldt2012-03-061-1/+1
| | | | | | | | | | | | | | | | | | | | The mechanism in fontconfig which determines if a certain character is available (FcCharSetHasChar()) may give false positives, in which case we would load and unload those fonts per every char for which FC gave us a false positive. This was a major performance regression. Specifically the false positives happened when looking at e.g. italic variants of certain multilingual fonts, since we only check the charset of the font family as a whole and not of the specific variant, which may only support a subset of the chars. To optimize this, we remove the deletion of the font engines after loading them, but also wait with loading the opentype tables until they are actually needed. This means that for the false positives, we will load the font, but the cached data for each unused font will be much smaller. Change-Id: Idfc794401a2080da5946bf65204eb947aeb635ed Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-11 00:30:44 +0000