aboutsummaryrefslogtreecommitdiffstats
path: root/testrunner.py
diff options
context:
space:
mode:
authorChristian Tismer <tismer@stackless.com>2016-08-10 16:05:35 +0200
committerChristian Tismer <tismer@stackless.com>2016-09-22 06:59:16 +0000
commit0102ef266868cde2406cba1c36d39551af1b1b94 (patch)
tree105bdc4c9d8761582a27964ac9726409b32757f7 /testrunner.py
parentce928d7a3c36e8685db57cf7568476a6be28d94c (diff)
support pyside CI testing with filtering
PySide is currently in an unfinished state. Therefore, running all the tests will never return a success. To make this more useful, this script is controlled by a blacklist, in order to keep record of the known-to-fail tests. On success: zero error status On Failure: nonzero (raises ValueError) The test parser uses a blacklist file that is compatible to standard Qt blacklist files. macOS, Ubuntu and windows (thanks to fkleint) are supported. The blacklist has a feature where new configurations are learned. It first removes old versions of the same tests, before it inserts the new ones. Blacklisting is now optional. We support now py2, py3, qt5, qt5.6 etc. Some consideration about collapsing the data are added. The parser has now a sub-command "test" that is mandatory. Support for moving the build_dir to another location without loosing the tests so far. Implement BPASS. Support CTEST_OUTPUT_ON_FAILURE Change-Id: If66d3cf1733a3b0c8a47fe9b8c6aec8ea430c699 Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Diffstat (limited to 'testrunner.py')
-rw-r--r--testrunner.py719
1 files changed, 719 insertions, 0 deletions
diff --git a/testrunner.py b/testrunner.py
new file mode 100644
index 00000000..279cc231
--- /dev/null
+++ b/testrunner.py
@@ -0,0 +1,719 @@
+# This file is part of PySide: Python for Qt
+#
+# Copyright (C) 2016 The Qt Company Ltd.
+#
+# Contact: http://www.qt.io/licensing/
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Lesser General Public License
+# version 2 as published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# Lesser General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public
+# License along with this program; if not, write to the Free Software
+# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+# 02110-1301 USA
+
+from __future__ import print_function
+
+"""
+testrunner
+
+Provide an interface to the pyside tests.
+
+- find the latest build dir.
+ This is found by the build_history in setup.py,
+ near the end of pyside_build.run()
+
+- run 'make test' and record the output
+ (not ready)
+
+- compare the result list with the current blacklist
+
+- return the correct error status
+ (zero if expected includes observed, else 1)
+
+Recommended build process:
+There is no need to install the project.
+Building the project with something like
+
+ python setup.py build --build-tests --qmake=<qmakepath> --ignore-git --debug
+
+is sufficient. The tests are run by changing into the latest build dir and there
+into pyside2, then 'make test'.
+
+"""
+
+import os
+import sys
+import re
+import subprocess
+import zipfile
+import argparse
+
+from six import PY3
+from subprocess import PIPE
+if PY3:
+ from subprocess import TimeoutExpired
+ from io import StringIO
+else:
+ class SubprocessError(Exception): pass
+ # this is a fake, just to keep the source compatible.
+ # timeout support is in python 3.3 and above.
+ class TimeoutExpired(SubprocessError): pass
+ from StringIO import StringIO
+from collections import namedtuple
+
+# Change the cwd to our source dir
+try:
+ this_file = __file__
+except NameError:
+ this_file = sys.argv[0]
+this_file = os.path.abspath(this_file)
+if os.path.dirname(this_file):
+ os.chdir(os.path.dirname(this_file))
+script_dir = os.getcwd()
+
+LogEntry = namedtuple("LogEntry", ["log_dir", "build_dir"])
+
+
+class BuildLog(object):
+ """
+ This class is a convenience wrapper around a list of log entries.
+
+ The list of entries is sorted by date and checked for consistency.
+ For simplicity and readability, the log entries are named tuples.
+
+ """
+ def __init__(self, script_dir=script_dir):
+ history_dir = os.path.join(script_dir, 'build_history')
+ build_history = []
+ for timestamp in os.listdir(history_dir):
+ log_dir = os.path.join(history_dir, timestamp)
+ if not os.path.isdir(log_dir):
+ continue
+ fpath = os.path.join(log_dir, 'build_dir.txt')
+ if not os.path.exists(fpath):
+ print("Warning: %s not found, skipped" % fpath)
+ continue
+ with open(fpath) as f:
+ build_dir = f.read().strip()
+ if not os.path.exists(build_dir):
+ rel_dir, low_part = os.path.split(build_dir)
+ rel_dir, two_part = os.path.split(rel_dir)
+ if two_part == "pyside_build":
+ build_dir = os.path.abspath(os.path.join(two_part, low_part))
+ if os.path.exists(build_dir):
+ print("Note: build_dir was probably moved.")
+ else:
+ print("Warning: missing build dir %s" % build_dir)
+ continue
+ entry = LogEntry(log_dir, build_dir)
+ build_history.append(entry)
+ # we take the latest build for now.
+ build_history.sort()
+ self.history = build_history
+
+ @property
+ def last_build(self):
+ if(self.history):
+ return self.history[-1]
+ else:
+ return None
+
+ @property
+ def classifiers(self):
+ if not self.last_build:
+ raise ValueError('+++ No last build with the configuration found!')
+ res = [sys.platform]
+ # the rest must be guessed from the given filename
+ path = self.last_build.build_dir
+ base = os.path.basename(path)
+ res.extend(base.split('-'))
+ # add the python version py2 and py3
+ # also add the keys qt5 and qt5.6 etc.
+ for entry in res:
+ if entry.startswith("py"):
+ key = entry[:3]
+ if key not in res:
+ res.append(key)
+ if entry.startswith("qt"):
+ key = entry[:3]
+ if key not in res:
+ res.append(key)
+ key = entry[:5]
+ if key not in res:
+ res.append(key)
+ # this will become more difficult when the version has two digits
+ return res
+
+
+class TestRunner(object):
+ def __init__(self, log_entry):
+ self.log_entry = log_entry
+ built_path = log_entry.build_dir
+ self.test_dir = os.path.join(built_path, "pyside2")
+ log_dir = log_entry.log_dir
+ self.logfile = os.path.join(log_dir, "testrun.log")
+ os.environ['CTEST_OUTPUT_ON_FAILURE'] = '1'
+ self._setup()
+
+ def _setup(self):
+ if sys.platform == 'win32':
+ # Windows: Helper implementing 'which' command using 'where.exe'
+ def winWhich(binary):
+ cmd = ['where.exe', binary]
+ stdOut = subprocess.Popen(cmd, stdout=subprocess.PIPE).stdout
+ result = stdOut.readlines()
+ stdOut.close()
+ if len(result) > 0:
+ return re.compile('\\s+').sub(' ', result[0].decode('utf-8'))
+ return None
+
+ self.makeCommand = 'nmake'
+ qmakeSpec = os.environ.get('QMAKESPEC')
+ if qmakeSpec is not None and 'g++' in qmakeSpec:
+ self.makeCommand = 'mingw32-make'
+ # Can 'tee' be found in the environment (MSYS-git installation with usr/bin in path)?
+ self.teeCommand = winWhich('tee.exe')
+ if self.teeCommand is None:
+ git = winWhich('git.exe')
+ if 'cmd' in git:
+ # Check for a MSYS-git installation with 'cmd' in the path and grab 'tee' from usr/bin
+ index = git.index('cmd')
+ self.teeCommand = git[0:index] + 'usr\\bin\\tee.exe'
+ else:
+ self.makeCommand = 'make'
+ self.teeCommand = 'tee'
+
+ def run(self, timeout = 300):
+ """
+ perform a test run in a given build. The build can be stopped by a
+ keyboard interrupt for testing this script. Also, a timeout can
+ be used.
+ """
+
+ if sys.platform == "win32":
+ cmd = (self.makeCommand, 'test')
+ tee_cmd = (self.teeCommand, self.logfile)
+ print("running", cmd, 'in', self.test_dir, ',\n logging to', self.logfile, 'using ', tee_cmd)
+ make = subprocess.Popen(cmd, cwd=self.test_dir, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+ tee = subprocess.Popen(tee_cmd, cwd=self.test_dir, stdin=make.stdout, shell=True)
+ else:
+ cmd = (self.makeCommand, 'test', '-C', "%s" % self.test_dir)
+ tee_cmd = (self.teeCommand, self.logfile)
+ print("running", cmd, 'in', self.test_dir, ',\n logging to', self.logfile, 'using ', tee_cmd)
+ make = subprocess.Popen(cmd, stdout=subprocess.PIPE)
+ tee = subprocess.Popen(tee_cmd, stdin=make.stdout)
+ make.stdout.close()
+ try:
+ if PY3:
+ output = tee.communicate(timeout=timeout)[0]
+ else:
+ output = tee.communicate()[0]
+ except (TimeoutExpired, KeyboardInterrupt):
+ print()
+ print("aborted")
+ tee.kill()
+ make.kill()
+ outs, errs = tee.communicate()
+ finally:
+ print("End of the test run")
+ tee.wait()
+
+
+_EXAMPLE = """
+Example output:
+
+ip1 n sharp mod_name code tim
+-----------------------------------------------------------------------------------------
+114/391 Test #114: QtCore_qfileinfo_test ........................... Passed 0.10 sec
+ Start 115: QtCore_qfile_test
+115/391 Test #115: QtCore_qfile_test ...............................***Failed 0.11 sec
+ Start 116: QtCore_qflags_test
+
+We will only look for the dotted lines and calculate everything from that.
+The summary statistics at the end will be ignored. That allows us to test
+this functionality with short timeout values.
+
+Note the field "mod_name". I had split this before, but it is necessary
+to use the combination as the key, because the test names are not unique.
+"""
+
+# validation of our pattern:
+
+_TEST_PAT = r"""
+ ^ # start
+ \s* # any whitespace ==: WS
+ ([0-9]+)/([0-9]+) # ip1 "/" n
+ \s+ # some WS
+ Test # "Test"
+ \s+ # some WS
+ \# # sharp symbol "#"
+ ([0-9]+) # sharp
+ : # colon symbol ':'
+ \s+ # some WS
+ (\w+) # mod_name
+ .*? # whatever (non greedy)
+ ( #
+ (Passed) # either "Passed", None
+ | #
+ \*\*\*(\w+) # or None, "Something"
+ ) # code
+ \s+ # some WS
+ ([0-9]+\.[0-9]+) # tim
+ \s+ # some WS
+ sec # "sec"
+ \s* # any WS
+ $ # end
+ """
+assert re.match(_TEST_PAT, _EXAMPLE.splitlines()[5], re.VERBOSE)
+assert len(re.match(_TEST_PAT, _EXAMPLE.splitlines()[5], re.VERBOSE).groups()) == 8
+assert len(re.match(_TEST_PAT, _EXAMPLE.splitlines()[7], re.VERBOSE).groups()) == 8
+
+TestResult = namedtuple("TestResult", ["idx", "mod_name", "passed",
+ "code", "time"])
+
+
+class TestParser(object):
+ def __init__(self, test_log):
+ self._result = _parse_tests(test_log)
+
+ @property
+ def result(self):
+ return self._result
+
+ def __len__(self):
+ return len(self._result)
+
+ def iter_blacklist(self, blacklist):
+ bl = blacklist
+ for line in self._result:
+ mod_name = line.mod_name
+ passed = line.passed
+ match = bl.find_matching_line(line)
+ if not passed:
+ if match:
+ res = "BFAIL"
+ else:
+ res = "FAIL"
+ else:
+ if match:
+ res = "BPASS"
+ else:
+ res = "PASS"
+ yield mod_name, res
+
+
+class BlackList(object):
+ def __init__(self, blname):
+ if blname == None:
+ f = StringIO()
+ self.raw_data = []
+ else:
+ with open(blname) as f:
+ self.raw_data = f.readlines()
+ # keep all lines, but see what is not relevant
+ lines = self.raw_data[:]
+
+ def filtered_line(line):
+ if '#' in line:
+ line = line[0:line.index('#')]
+ return line.split()
+
+ # now put every bracketed line in a test
+ # and use subsequent identifiers for a match
+ def is_test(fline):
+ return fline and fline[0].startswith("[")
+
+ self.tests = {}
+
+ if not lines:
+ # nothing supplied
+ return
+
+ self.index = {}
+ for idx, line in enumerate(lines):
+ fline = filtered_line(line)
+ if not fline:
+ continue
+ if is_test(fline):
+ break
+ # we have a global section
+ name = ''
+ self.tests[name] = []
+ for idx, line in enumerate(lines):
+ fline = filtered_line(line)
+ if is_test(fline):
+ # a new name
+ name = decorate(fline[0][1:-1])
+ self.tests[name] = []
+ self.index[name] = idx
+ elif fline:
+ # a known name with a new entry
+ self.tests[name].append(fline)
+
+ def find_matching_line(self, test):
+ """
+ Take a test result.
+ Find a line in the according blacklist file where all keys of the line are found.
+ If line not found, do nothing.
+ if line found and test passed, it is a BPASS.
+ If line found and test failed, it is a BFAIL.
+ """
+ passed = test.passed
+ classifiers = set(builds.classifiers)
+
+ if "" in self.tests:
+ # this is a global section
+ for line in self.tests[""]:
+ keys = set(line)
+ if keys <= classifiers:
+ # found a match!
+ return line
+ mod_name = test.mod_name
+ if mod_name not in self.tests and decorate(mod_name) not in self.tests:
+ return None
+ if mod_name in self.tests:
+ thing = mod_name
+ else:
+ thing = decorate(mod_name)
+ for line in self.tests[thing]:
+ keys = set(line)
+ if keys <= classifiers:
+ # found a match!
+ return line
+ else:
+ return None # noting found
+
+
+"""
+Simplified blacklist file
+-------------------------
+
+A comment reaches from '#' to the end of line.
+The file starts with an optional global section.
+A test is started with a [square-bracketed] section name.
+A line matches if all keys in the line are found.
+If a line matches, the corresponding test is marked BFAIL or BPASS depending if the test passed or
+not.
+
+Known keys are:
+
+darwin
+win32
+linux2
+...
+
+qt5.6.1
+qt5.6.2
+...
+
+py3
+py2
+
+32bit
+64bit
+
+debug
+release
+"""
+
+"""
+Data Folding v2
+===============
+
+In the first layout of data folding, we distinguished complete domains
+like "debug/release" and incomplete domains like "ubuntu/win32" which
+can be extended to any number.
+
+This version is simpler. We do a first pass over all data and collect
+all data. Therefore, incomplete domains do not exist. The definition
+of the current members of the domain goes into a special comment at
+the beginning of the file.
+
+
+Compressing a blacklist
+-----------------------
+
+When we have many samples of data, it is very likely to get very similar
+entries. The redundancy is quite high, and we would like to compress
+data without loosing information.
+
+Consider the following data set:
+
+[some::sample_test]
+ darwin qt5.6.1 py3 64bit debug
+ darwin qt5.6.1 py3 64bit release
+ darwin qt5.6.1 py2 64bit debug
+ darwin qt5.6.1 py2 64bit release
+ win32 qt5.6.1 py3 64bit debug
+ win32 qt5.6.1 py3 64bit release
+ win32 qt5.6.1 py2 64bit debug
+ win32 qt5.6.1 py2 64bit release
+
+The keys "debug" and "release" build the complete set of keys in their
+domain. When sorting the lines, we can identify all similar entries which
+are only different by the keys "debug" and "release".
+
+[some::sample_test]
+ darwin qt5.6.1 py3 64bit
+ darwin qt5.6.1 py2 64bit
+ win32 qt5.6.1 py3 64bit
+ win32 qt5.6.1 py2 64bit
+
+We can do the same for "py3" and "py2", because we have again the complete
+set of possible keys available:
+
+[some::sample_test]
+ darwin qt5.6.1 64bit
+ win32 qt5.6.1 64bit
+
+The operating system has the current keys "darwin" and "win32".
+They are kept in a special commend, and we get:
+
+# COMPRESSION: darwin win32
+[some::sample_test]
+ qt5.6.1 64bit
+
+
+Expanding a blacklist
+---------------------
+
+All of the above steps are completely reversible.
+
+
+Alternate implementation
+------------------------
+
+Instead of using a special comment, I am currently in favor of
+the following:
+
+The global section gets the complete set of variables, like so
+
+# Globals
+ darwin win32 linux2
+ qt5.6.1 qt5.6.2
+ py3 py2
+ 32bit 64bit
+ debug release
+[some::sample_test]
+ qt5.6.1 64bit
+
+This approach has the advantage that it does not depend on comments.
+The lines in the global section can always added without any conflict,
+because these test results are impossible. Therefore, we list all our
+keys without adding anything that could influence a test.
+It makes also sense to have everything explicitly listed here.
+"""
+
+def _parse_tests(test_log):
+ """
+ Create a TestResult object for every entry.
+ """
+ result = []
+ if isinstance(test_log, StringIO):
+ lines = test_log.readlines()
+ elif test_log is not None and os.path.exists(test_log):
+ with open(test_log) as f:
+ lines = f.readlines()
+ else:
+ lines = []
+ pat = _TEST_PAT
+ for line in lines:
+ match = re.match(pat, line, re.VERBOSE)
+ if match:
+ idx, n, sharp, mod_name, much_stuff, code1, code2, tim = tup = match.groups()
+ # either code1 or code2 is None
+ code = code1 or code2
+ if idx != sharp:
+ raise ValueError("inconsistent log lines or program error: %s" % tup)
+ idx, n, code, tim = int(idx), int(n), code.lower(), float(tim)
+ res = TestResult(idx, mod_name, code == "passed", code, tim)
+ result.append(res)
+ return result
+
+def decorate(mod_name):
+ """
+ Write the combination of "modulename_funcname"
+ in the Qt-like form "modulename::funcname"
+ """
+ if "_" not in mod_name:
+ return mod_name
+ if "::" in mod_name:
+ return mod_name
+ name, rest = mod_name.split("_", 1)
+ return name + "::" + rest
+
+def create_read_write(filename):
+ if os.path.isfile(filename):
+ # existing file, open for read and write
+ return open(filename, 'r+')
+ elif os.path.exists(filename):
+ # a directory?
+ raise argparse.ArgumentTypeError(None, "invalid file argument: %s" % filename)
+ else:
+ try:
+ return open(filename, 'w')
+ except IOError:
+ raise argparse.ArgumentError(None, "cannot create file: %s" % filename)
+
+def learn_blacklist(fname, result, latest_build):
+ with open(fname, "r+") as f:
+ _remove_from_blacklist(f.name)
+ _add_to_blacklist(f.name, result)
+ _update_header(f.name, latest_build)
+
+def _remove_from_blacklist(old_blname):
+ # get rid of existing classifiers
+ classifierset = set(builds.classifiers)
+
+ # for every line, remove comments and see if the current set if an exact
+ # match. We will touch only exact matches.
+ def filtered_line(line):
+ if '#' in line:
+ line = line[0:line.index('#')]
+ return line.split()
+
+ with open(old_blname, "r") as f:
+ lines = f.readlines()
+ deletions = []
+ for idx, line in enumerate(lines):
+ fline = filtered_line(line)
+ if not fline:
+ continue
+ if '[' in fline[0]:
+ # a heading line
+ continue
+ if set(fline) == classifierset:
+ deletions.append(idx)
+ while deletions:
+ delete = deletions.pop()
+ del lines[delete]
+ # remove all blank lines, but keep comments
+ for idx, line in reversed(list(enumerate(lines))):
+ if not line.split():
+ del lines[idx]
+ # remove all consecutive sections, but keep comments
+ for idx, line in reversed(list(enumerate(lines))):
+ fline = line.split()
+ if fline and fline[0].startswith("["):
+ if idx+1 == len(lines):
+ # remove section at the end
+ del lines[idx]
+ continue
+ gline = lines[idx+1].split()
+ if gline and gline[0].startswith("["):
+ # next section right after this, remove this
+ del lines[idx]
+ with open(old_blname, "w") as f:
+ f.writelines(lines)
+
+def _add_to_blacklist(old_blname, result):
+ # insert new classifiers
+ classifiers = " " + " ".join(builds.classifiers) + "\n"
+ insertions = []
+ additions = []
+ old_bl = BlackList(old_blname)
+ lines = old_bl.raw_data[:]
+ if lines and not lines[-1].endswith("\n"):
+ lines[-1] += "\n"
+ for test in result:
+ if test.passed:
+ continue
+ if test.mod_name in old_bl.tests:
+ # the test is already there, add to the first line
+ idx = old_bl.index[test.mod_name]
+ insertions.append(idx)
+ if decorate(test.mod_name) in old_bl.tests:
+ # the same, but the name was decorated
+ idx = old_bl.index[decorate(test.mod_name)]
+ insertions.append(idx)
+ else:
+ # the test is new, append it to the end
+ additions.append("[" + decorate(test.mod_name) + "]\n")
+ while insertions:
+ this = insertions.pop()
+ lines[this] += classifiers
+ for line in additions:
+ lines.append(line)
+ lines.append(classifiers)
+ # now write the data out
+ with open(old_blname, "r+") as f:
+ f.writelines(lines)
+
+def _update_header(old_blname, build_history):
+ with open(old_blname) as f:
+ lines = f.readlines()
+ classifierset = set(builds.classifiers)
+ for idx, line in reversed(list(enumerate(lines))):
+ fline = line.split()
+ if fline and fline[0].startswith('#'):
+ if set(fline) >= classifierset:
+ del lines[idx]
+
+ classifiers = " ".join(builds.classifiers)
+ path = latest_build.log_dir
+ base = os.path.basename(path)
+ test = '### test date = %s classifiers = %s\n' % (base, classifiers)
+ lines.insert(0, test)
+ with open(old_blname, "w") as f:
+ f.writelines(lines)
+
+
+if __name__ == '__main__':
+ # create the top-level parser
+ parser = argparse.ArgumentParser()
+ subparsers = parser.add_subparsers()
+
+ # create the parser for the "test" command
+ parser_test = subparsers.add_parser('test')
+ group = parser_test.add_mutually_exclusive_group(required=False)
+ group.add_argument("--blacklist", "-b", type=argparse.FileType('r'),
+ help="a Qt blacklist file")
+ group.add_argument("--learn", "-l", type=create_read_write,
+ help="add new entries to a blacklist file")
+ parser_test.add_argument("--skip", action='store_true',
+ help="skip the tests if they were run before")
+ args = parser.parse_args()
+
+ builds = BuildLog(script_dir)
+
+ latest_build = builds.last_build
+ if latest_build is None:
+ raise ValueError("you have never created a test build")
+ runner = TestRunner(latest_build)
+ if os.path.exists(runner.logfile) and args.skip:
+ print("Parsing existing log file:", runner.logfile)
+ else:
+ runner.run()
+ result = TestParser(runner.logfile)
+
+ if args.blacklist:
+ args.blacklist.close()
+ bl = BlackList(args.blacklist.name)
+ elif args.learn:
+ args.learn.close()
+ learn_blacklist(args.learn.name, result.result, latest_build)
+ bl = BlackList(args.learn.name)
+ else:
+ bl = BlackList(None)
+ print("********* Start testing of PySide *********")
+ print("Config: Using", " ".join(builds.classifiers))
+ pass_, skipped, fail, bfail, bpass = 0, 0, 0, 0, 0
+ for test, res in result.iter_blacklist(bl):
+ print("%-6s" % res, ":", decorate(test) + "()")
+ pass_ += 1 if res == "PASS" else 0
+ skipped += 1 if res == "SKIPPED" else 0 # not yet supported
+ fail += 1 if res == "FAIL" else 0
+ bfail += 1 if res == "BFAIL" else 0
+ bpass += 1 if res == "BPASS" else 0
+ print("Totals:", len(result), "tests.", pass_, "passed,", fail, "failed,", skipped, "skipped,", bfail, "blacklisted,", bpass, "bpasses")
+ print("********* Finished testing of PySide *********")
+ for test, res in result.iter_blacklist(bl):
+ if res == "FAIL":
+ raise ValueError("At least one failure was not blacklisted")