cpython/Lib/test/regrtest.py

333 lines
11 KiB
Python
Executable File

#! /usr/bin/env python
"""Regression test.
This will find all modules whose name is "test_*" in the test
directory, and run them. Various command line options provide
additional facilities.
Command line options:
-v: verbose -- run tests in verbose mode with output to stdout
-q: quiet -- don't print anything except if a test fails
-g: generate -- write the output file for a test instead of comparing it
-x: exclude -- arguments are tests to *exclude*
-s: single -- run only a single test (see below)
-r: random -- randomize test execution order
-l: findleaks -- if GC is available detect tests that leak memory
--have-resources -- run tests that require large resources (time/space)
If non-option arguments are present, they are names for tests to run,
unless -x is given, in which case they are names for tests not to run.
If no test names are given, all tests are run.
-v is incompatible with -g and does not compare test output files.
-s means to run only a single test and exit. This is useful when doing memory
analysis on the Python interpreter (which tend to consume to many resources to
run the full regression test non-stop). The file /tmp/pynexttest is read to
find the next test to run. If this file is missing, the first test_*.py file
in testdir or on the command line is used. (actually tempfile.gettempdir() is
used instead of /tmp).
"""
import sys
import os
import getopt
import traceback
import random
import test_support
def main(tests=None, testdir=None, verbose=0, quiet=0, generate=0,
exclude=0, single=0, randomize=0, findleaks=0,
use_large_resources=0):
"""Execute a test suite.
This also parses command-line options and modifies its behavior
accordingly.
tests -- a list of strings containing test names (optional)
testdir -- the directory in which to look for tests (optional)
Users other than the Python test suite will certainly want to
specify testdir; if it's omitted, the directory containing the
Python test suite is searched for.
If the tests argument is omitted, the tests listed on the
command-line will be used. If that's empty, too, then all *.py
files beginning with test_ will be used.
The other seven default arguments (verbose, quiet, generate, exclude,
single, randomize, and findleaks) allow programmers calling main()
directly to set the values that would normally be set by flags on the
command line.
"""
try:
opts, args = getopt.getopt(sys.argv[1:], 'vgqxsrl', ['have-resources'])
except getopt.error, msg:
print msg
print __doc__
return 2
for o, a in opts:
if o == '-v': verbose = verbose+1
if o == '-q': quiet = 1; verbose = 0
if o == '-g': generate = 1
if o == '-x': exclude = 1
if o == '-s': single = 1
if o == '-r': randomize = 1
if o == '-l': findleaks = 1
if o == '--have-resources': use_large_resources = 1
if generate and verbose:
print "-g and -v don't go together!"
return 2
good = []
bad = []
skipped = []
if findleaks:
try:
import gc
except ImportError:
print 'No GC available, disabling findleaks.'
findleaks = 0
else:
# Uncomment the line below to report garbage that is not
# freeable by reference counting alone. By default only
# garbage that is not collectable by the GC is reported.
#gc.set_debug(gc.DEBUG_SAVEALL)
found_garbage = []
if single:
from tempfile import gettempdir
filename = os.path.join(gettempdir(), 'pynexttest')
try:
fp = open(filename, 'r')
next = fp.read().strip()
tests = [next]
fp.close()
except IOError:
pass
for i in range(len(args)):
# Strip trailing ".py" from arguments
if args[i][-3:] == '.py':
args[i] = args[i][:-3]
stdtests = STDTESTS[:]
nottests = NOTTESTS[:]
if exclude:
for arg in args:
if arg in stdtests:
stdtests.remove(arg)
nottests[:0] = args
args = []
tests = tests or args or findtests(testdir, stdtests, nottests)
if single:
tests = tests[:1]
if randomize:
random.shuffle(tests)
test_support.verbose = verbose # Tell tests to be moderately quiet
test_support.use_large_resources = use_large_resources
save_modules = sys.modules.keys()
for test in tests:
if not quiet:
print test
ok = runtest(test, generate, verbose, quiet, testdir)
if ok > 0:
good.append(test)
elif ok == 0:
bad.append(test)
else:
skipped.append(test)
if findleaks:
gc.collect()
if gc.garbage:
print "Warning: test created", len(gc.garbage),
print "uncollectable object(s)."
# move the uncollectable objects somewhere so we don't see
# them again
found_garbage.extend(gc.garbage)
del gc.garbage[:]
# Unload the newly imported modules (best effort finalization)
for module in sys.modules.keys():
if module not in save_modules and module.startswith("test."):
test_support.unload(module)
if good and not quiet:
if not bad and not skipped and len(good) > 1:
print "All",
print count(len(good), "test"), "OK."
if verbose:
print "CAUTION: stdout isn't compared in verbose mode: a test"
print "that passes in verbose mode may fail without it."
if bad:
print count(len(bad), "test"), "failed:",
print " ".join(bad)
if skipped and not quiet:
print count(len(skipped), "test"), "skipped:",
print " ".join(skipped)
if single:
alltests = findtests(testdir, stdtests, nottests)
for i in range(len(alltests)):
if tests[0] == alltests[i]:
if i == len(alltests) - 1:
os.unlink(filename)
else:
fp = open(filename, 'w')
fp.write(alltests[i+1] + '\n')
fp.close()
break
else:
os.unlink(filename)
return len(bad) > 0
STDTESTS = [
'test_grammar',
'test_opcodes',
'test_operations',
'test_builtin',
'test_exceptions',
'test_types',
]
NOTTESTS = [
'test_support',
'test_b1',
'test_b2',
]
def findtests(testdir=None, stdtests=STDTESTS, nottests=NOTTESTS):
"""Return a list of all applicable test modules."""
if not testdir: testdir = findtestdir()
names = os.listdir(testdir)
tests = []
for name in names:
if name[:5] == "test_" and name[-3:] == ".py":
modname = name[:-3]
if modname not in stdtests and modname not in nottests:
tests.append(modname)
tests.sort()
return stdtests + tests
def runtest(test, generate, verbose, quiet, testdir = None):
"""Run a single test.
test -- the name of the test
generate -- if true, generate output, instead of running the test
and comparing it to a previously created output file
verbose -- if true, print more messages
quiet -- if true, don't print 'skipped' messages (probably redundant)
testdir -- test directory
"""
test_support.unload(test)
if not testdir: testdir = findtestdir()
outputdir = os.path.join(testdir, "output")
outputfile = os.path.join(outputdir, test)
try:
if generate:
cfp = open(outputfile, "w")
elif verbose:
cfp = sys.stdout
else:
cfp = Compare(outputfile)
except IOError:
cfp = None
print "Warning: can't open", outputfile
try:
save_stdout = sys.stdout
try:
if cfp:
sys.stdout = cfp
print test # Output file starts with test name
__import__(test, globals(), locals(), [])
if cfp and not (generate or verbose):
cfp.close()
finally:
sys.stdout = save_stdout
except (ImportError, test_support.TestSkipped), msg:
if not quiet:
print "test", test,
print "skipped -- ", msg
return -1
except KeyboardInterrupt:
raise
except test_support.TestFailed, msg:
print "test", test, "failed --", msg
return 0
except:
type, value = sys.exc_info()[:2]
print "test", test, "crashed --", str(type) + ":", value
if verbose:
traceback.print_exc(file=sys.stdout)
return 0
else:
return 1
def findtestdir():
if __name__ == '__main__':
file = sys.argv[0]
else:
file = __file__
testdir = os.path.dirname(file) or os.curdir
return testdir
def count(n, word):
if n == 1:
return "%d %s" % (n, word)
else:
return "%d %ss" % (n, word)
class Compare:
def __init__(self, filename):
self.fp = open(filename, 'r')
self.stuffthatmatched = []
def write(self, data):
expected = self.fp.read(len(data))
if data == expected:
self.stuffthatmatched.append(expected)
else:
# This Compare instance is spoofing stdout, so we need to write
# to stderr instead.
from sys import stderr as e
print >> e, "The actual stdout doesn't match the expected stdout."
if self.stuffthatmatched:
print >> e, "This much did match (between asterisk lines):"
print >> e, "*" * 70
good = "".join(self.stuffthatmatched)
e.write(good)
if not good.endswith("\n"):
e.write("\n")
print >> e, "*" * 70
print >> e, "Then ..."
else:
print >> e, "The first write to stdout clashed:"
# Note that the prompts are the same length in next two lines.
# This is so what we expected and what we got line up.
print >> e, "We expected (repr):", `expected`
print >> e, "But instead we got:", `data`
raise test_support.TestFailed('Writing: ' + `data`+
', expected: ' + `expected`)
def writelines(self, listoflines):
map(self.write, listoflines)
def flush(self):
pass
def close(self):
leftover = self.fp.read()
if leftover:
raise test_support.TestFailed('Tail of expected stdout unseen: ' +
`leftover`)
self.fp.close()
def isatty(self):
return 0
if __name__ == '__main__':
sys.exit(main())