gecko-dev/config/tests/unit-expandlibs.py

432 строки
19 KiB
Python
Исходник Обычный вид История

import subprocess
import unittest
import sys
import os
import imp
from tempfile import mkdtemp
from shutil import rmtree
import mozunit
from UserString import UserString
# Create a controlled configuration for use by expandlibs
config_win = {
'AR': 'lib',
'AR_EXTRACT': '',
'DLL_PREFIX': '',
'LIB_PREFIX': '',
'OBJ_SUFFIX': '.obj',
'LIB_SUFFIX': '.lib',
'DLL_SUFFIX': '.dll',
'IMPORT_LIB_SUFFIX': '.lib',
'LIBS_DESC_SUFFIX': '.desc',
'EXPAND_LIBS_LIST_STYLE': 'list',
}
config_unix = {
'AR': 'ar',
'AR_EXTRACT': 'ar -x',
'DLL_PREFIX': 'lib',
'LIB_PREFIX': 'lib',
'OBJ_SUFFIX': '.o',
'LIB_SUFFIX': '.a',
'DLL_SUFFIX': '.so',
'IMPORT_LIB_SUFFIX': '',
'LIBS_DESC_SUFFIX': '.desc',
'EXPAND_LIBS_LIST_STYLE': 'linkerscript',
}
config = sys.modules['expandlibs_config'] = imp.new_module('expandlibs_config')
from expandlibs import LibDescriptor, ExpandArgs, relativize
from expandlibs_gen import generate
from expandlibs_exec import ExpandArgsMore, SectionFinder
def Lib(name):
return config.LIB_PREFIX + name + config.LIB_SUFFIX
def Obj(name):
return name + config.OBJ_SUFFIX
def Dll(name):
return config.DLL_PREFIX + name + config.DLL_SUFFIX
def ImportLib(name):
if not len(config.IMPORT_LIB_SUFFIX): return Dll(name)
return config.LIB_PREFIX + name + config.IMPORT_LIB_SUFFIX
class TestRelativize(unittest.TestCase):
def test_relativize(self):
'''Test relativize()'''
os_path_exists = os.path.exists
def exists(path):
return True
os.path.exists = exists
self.assertEqual(relativize(os.path.abspath(os.curdir)), os.curdir)
self.assertEqual(relativize(os.path.abspath(os.pardir)), os.pardir)
self.assertEqual(relativize(os.path.join(os.curdir, 'a')), 'a')
self.assertEqual(relativize(os.path.join(os.path.abspath(os.curdir), 'a')), 'a')
# relativize is expected to return the absolute path if it is shorter
self.assertEqual(relativize(os.sep), os.sep)
os.path.exists = os.path.exists
class TestLibDescriptor(unittest.TestCase):
def test_serialize(self):
'''Test LibDescriptor's serialization'''
desc = LibDescriptor()
desc[LibDescriptor.KEYS[0]] = ['a', 'b']
self.assertEqual(str(desc), "{0} = a b".format(LibDescriptor.KEYS[0]))
desc['unsupported-key'] = ['a']
self.assertEqual(str(desc), "{0} = a b".format(LibDescriptor.KEYS[0]))
desc[LibDescriptor.KEYS[1]] = ['c', 'd', 'e']
self.assertEqual(str(desc),
"{0} = a b\n{1} = c d e"
.format(LibDescriptor.KEYS[0], LibDescriptor.KEYS[1]))
desc[LibDescriptor.KEYS[0]] = []
self.assertEqual(str(desc), "{0} = c d e".format(LibDescriptor.KEYS[1]))
def test_read(self):
'''Test LibDescriptor's initialization'''
desc_list = ["# Comment",
"{0} = a b".format(LibDescriptor.KEYS[1]),
"", # Empty line
"foo = bar", # Should be discarded
"{0} = c d e".format(LibDescriptor.KEYS[0])]
desc = LibDescriptor(desc_list)
self.assertEqual(desc[LibDescriptor.KEYS[1]], ['a', 'b'])
self.assertEqual(desc[LibDescriptor.KEYS[0]], ['c', 'd', 'e'])
self.assertEqual(False, 'foo' in desc)
def wrap_method(conf, wrapped_method):
'''Wrapper used to call a test with a specific configuration'''
def _method(self):
for key in conf:
setattr(config, key, conf[key])
self.init()
try:
wrapped_method(self)
except:
raise
finally:
self.cleanup()
return _method
class ReplicateTests(type):
'''Replicates tests for unix and windows variants'''
def __new__(cls, clsName, bases, dict):
for name in [key for key in dict if key.startswith('test_')]:
dict[name + '_unix'] = wrap_method(config_unix, dict[name])
dict[name + '_unix'].__doc__ = dict[name].__doc__ + ' (unix)'
dict[name + '_win'] = wrap_method(config_win, dict[name])
dict[name + '_win'].__doc__ = dict[name].__doc__ + ' (win)'
del dict[name]
return type.__new__(cls, clsName, bases, dict)
class TestCaseWithTmpDir(unittest.TestCase):
__metaclass__ = ReplicateTests
def init(self):
self.tmpdir = os.path.abspath(mkdtemp(dir=os.curdir))
def cleanup(self):
rmtree(self.tmpdir)
def touch(self, files):
for f in files:
open(f, 'w').close()
def tmpfile(self, *args):
return os.path.join(self.tmpdir, *args)
class TestExpandLibsGen(TestCaseWithTmpDir):
def test_generate(self):
'''Test library descriptor generation'''
files = [self.tmpfile(f) for f in
[Lib('a'), Obj('b'), Lib('c'), Obj('d'), Obj('e'), Lib('f')]]
self.touch(files[:-1])
self.touch([files[-1] + config.LIBS_DESC_SUFFIX])
desc = generate(files)
self.assertEqual(desc['OBJS'], [self.tmpfile(Obj(s)) for s in ['b', 'd', 'e']])
self.assertEqual(desc['LIBS'], [self.tmpfile(Lib(s)) for s in ['a', 'c', 'f']])
self.assertRaises(Exception, generate, files + [self.tmpfile(Obj('z'))])
self.assertRaises(Exception, generate, files + [self.tmpfile(Lib('y'))])
class TestExpandInit(TestCaseWithTmpDir):
def init(self):
''' Initializes test environment for library expansion tests'''
super(TestExpandInit, self).init()
# Create 2 fake libraries, each containing 3 objects, and the second
# including the first one and another library.
os.mkdir(self.tmpfile('libx'))
os.mkdir(self.tmpfile('liby'))
self.libx_files = [self.tmpfile('libx', Obj(f)) for f in ['g', 'h', 'i']]
self.liby_files = [self.tmpfile('liby', Obj(f)) for f in ['j', 'k', 'l']] + [self.tmpfile('liby', Lib('z'))]
self.touch(self.libx_files + self.liby_files)
with open(self.tmpfile('libx', Lib('x') + config.LIBS_DESC_SUFFIX), 'w') as f:
f.write(str(generate(self.libx_files)))
with open(self.tmpfile('liby', Lib('y') + config.LIBS_DESC_SUFFIX), 'w') as f:
f.write(str(generate(self.liby_files + [self.tmpfile('libx', Lib('x'))])))
# Create various objects and libraries
self.arg_files = [self.tmpfile(f) for f in [Lib('a'), Obj('b'), Obj('c'), Lib('d'), Obj('e')]]
# We always give library names (LIB_PREFIX/SUFFIX), even for
# dynamic/import libraries
self.files = self.arg_files + [self.tmpfile(ImportLib('f'))]
self.arg_files += [self.tmpfile(Lib('f'))]
self.touch(self.files)
def assertRelEqual(self, args1, args2):
self.assertEqual(args1, [relativize(a) for a in args2])
class TestExpandArgs(TestExpandInit):
def test_expand(self):
'''Test library expansion'''
# Expanding arguments means libraries with a descriptor are expanded
# with the descriptor content, and import libraries are used when
# a library doesn't exist
args = ExpandArgs(['foo', '-bar'] + self.arg_files + [self.tmpfile('liby', Lib('y'))])
self.assertRelEqual(args, ['foo', '-bar'] + self.files + self.liby_files + self.libx_files)
Bug 1077151 - Always use expandlibs descriptors when they exist. r=mshal Currently, when there is both an expandlibs descriptor and an actual static library, expandlibs picks the static library. This has the side effect that if there are object files in the static library that aren't directly used, they're dropped when linking, even when they export symbols that would be exported in the final linked binary. In most cases in the code base, files are not dropped that way. The most notable counter-example is xpcomglue, where actually not dropping files leads to link failure because of missing symbols those files reference (yes, that would tend to say the glue is broken in some way). On the opposite side, there is mozglue, which does have both a descriptor and a static library (the latter being necessary for the SDK), and that linking as a static library drops files that shouldn't be dropped (like jemalloc). We're currently relying on -Wl,--whole-archive for those files not to be dropped, but that won't really be possible without much hassle in a world where mozglue dependencies live in moz.build land. Switching expandlibs to use descriptors when they exist, even when there is a static library (so, the opposite of the current behavior) allows to drop -Wl,--whole-archive and prepare for a better future. However, as mentioned, xpcomglue does still require to be linked through the static library, so we need to make it a static library only. To achieve that, we make NO_EXPAND_LIBS now actually mean no expandlibs and use that to build the various different xpcomglues.
2014-10-04 05:33:00 +04:00
# When a library exists at the same time as a descriptor, we still use
# the descriptor.
self.touch([self.tmpfile('libx', Lib('x'))])
args = ExpandArgs(['foo', '-bar'] + self.arg_files + [self.tmpfile('liby', Lib('y'))])
Bug 1077151 - Always use expandlibs descriptors when they exist. r=mshal Currently, when there is both an expandlibs descriptor and an actual static library, expandlibs picks the static library. This has the side effect that if there are object files in the static library that aren't directly used, they're dropped when linking, even when they export symbols that would be exported in the final linked binary. In most cases in the code base, files are not dropped that way. The most notable counter-example is xpcomglue, where actually not dropping files leads to link failure because of missing symbols those files reference (yes, that would tend to say the glue is broken in some way). On the opposite side, there is mozglue, which does have both a descriptor and a static library (the latter being necessary for the SDK), and that linking as a static library drops files that shouldn't be dropped (like jemalloc). We're currently relying on -Wl,--whole-archive for those files not to be dropped, but that won't really be possible without much hassle in a world where mozglue dependencies live in moz.build land. Switching expandlibs to use descriptors when they exist, even when there is a static library (so, the opposite of the current behavior) allows to drop -Wl,--whole-archive and prepare for a better future. However, as mentioned, xpcomglue does still require to be linked through the static library, so we need to make it a static library only. To achieve that, we make NO_EXPAND_LIBS now actually mean no expandlibs and use that to build the various different xpcomglues.
2014-10-04 05:33:00 +04:00
self.assertRelEqual(args, ['foo', '-bar'] + self.files + self.liby_files + self.libx_files)
self.touch([self.tmpfile('liby', Lib('y'))])
args = ExpandArgs(['foo', '-bar'] + self.arg_files + [self.tmpfile('liby', Lib('y'))])
Bug 1077151 - Always use expandlibs descriptors when they exist. r=mshal Currently, when there is both an expandlibs descriptor and an actual static library, expandlibs picks the static library. This has the side effect that if there are object files in the static library that aren't directly used, they're dropped when linking, even when they export symbols that would be exported in the final linked binary. In most cases in the code base, files are not dropped that way. The most notable counter-example is xpcomglue, where actually not dropping files leads to link failure because of missing symbols those files reference (yes, that would tend to say the glue is broken in some way). On the opposite side, there is mozglue, which does have both a descriptor and a static library (the latter being necessary for the SDK), and that linking as a static library drops files that shouldn't be dropped (like jemalloc). We're currently relying on -Wl,--whole-archive for those files not to be dropped, but that won't really be possible without much hassle in a world where mozglue dependencies live in moz.build land. Switching expandlibs to use descriptors when they exist, even when there is a static library (so, the opposite of the current behavior) allows to drop -Wl,--whole-archive and prepare for a better future. However, as mentioned, xpcomglue does still require to be linked through the static library, so we need to make it a static library only. To achieve that, we make NO_EXPAND_LIBS now actually mean no expandlibs and use that to build the various different xpcomglues.
2014-10-04 05:33:00 +04:00
self.assertRelEqual(args, ['foo', '-bar'] + self.files + self.liby_files + self.libx_files)
class TestExpandArgsMore(TestExpandInit):
def test_makelist(self):
'''Test grouping object files in lists'''
# ExpandArgsMore does the same as ExpandArgs
with ExpandArgsMore(['foo', '-bar'] + self.arg_files + [self.tmpfile('liby', Lib('y'))]) as args:
self.assertRelEqual(args, ['foo', '-bar'] + self.files + self.liby_files + self.libx_files)
# But also has an extra method replacing object files with a list
args.makelist()
# self.files has objects at #1, #2, #4
self.assertRelEqual(args[:3], ['foo', '-bar'] + self.files[:1])
self.assertRelEqual(args[4:], [self.files[3]] + self.files[5:] + [self.tmpfile('liby', Lib('z'))])
# Check the list file content
objs = [f for f in self.files + self.liby_files + self.libx_files if f.endswith(config.OBJ_SUFFIX)]
if config.EXPAND_LIBS_LIST_STYLE == "linkerscript":
self.assertNotEqual(args[3][0], '@')
filename = args[3]
content = ['INPUT("{0}")'.format(relativize(f)) for f in objs]
with open(filename, 'r') as f:
self.assertEqual([l.strip() for l in f.readlines() if len(l.strip())], content)
elif config.EXPAND_LIBS_LIST_STYLE == "list":
self.assertEqual(args[3][0], '@')
filename = args[3][1:]
content = objs
with open(filename, 'r') as f:
self.assertRelEqual([l.strip() for l in f.readlines() if len(l.strip())], content)
tmp = args.tmp
# Check that all temporary files are properly removed
self.assertEqual(True, all([not os.path.exists(f) for f in tmp]))
def test_extract(self):
'''Test library extraction'''
# Divert subprocess.call
subprocess_call = subprocess.call
subprocess_check_output = subprocess.check_output
def call(args, **kargs):
if config.AR == 'lib':
self.assertEqual(args[:2], [config.AR, '-NOLOGO'])
self.assertTrue(args[2].startswith('-EXTRACT:'))
extract = [args[2][len('-EXTRACT:'):]]
self.assertTrue(extract)
args = args[3:]
else:
# The command called is always AR_EXTRACT
ar_extract = config.AR_EXTRACT.split()
self.assertEqual(args[:len(ar_extract)], ar_extract)
args = args[len(ar_extract):]
# Remaining argument is always one library
self.assertEqual(len(args), 1)
arg = args[0]
self.assertEqual(os.path.splitext(arg)[1], config.LIB_SUFFIX)
# Simulate file extraction
lib = os.path.splitext(os.path.basename(arg))[0]
if config.AR != 'lib':
extract = [lib, lib + '2']
extract = [os.path.join(kargs['cwd'], f) for f in extract]
if config.AR != 'lib':
extract = [Obj(f) for f in extract]
if not lib in extracted:
extracted[lib] = []
extracted[lib].extend(extract)
self.touch(extract)
subprocess.call = call
def check_output(args, **kargs):
# The command called is always AR
ar = config.AR
self.assertEqual(args[0:3], [ar, '-NOLOGO', '-LIST'])
# Remaining argument is always one library
self.assertRelEqual([os.path.splitext(arg)[1] for arg in args[3:]],
[config.LIB_SUFFIX])
# Simulate LIB -NOLOGO -LIST
lib = os.path.splitext(os.path.basename(args[3]))[0]
return '%s\n%s\n' % (Obj(lib), Obj(lib + '2'))
subprocess.check_output = check_output
# ExpandArgsMore does the same as ExpandArgs
self.touch([self.tmpfile('liby', Lib('y'))])
for iteration in (1, 2):
with ExpandArgsMore(['foo', '-bar'] + self.arg_files + [self.tmpfile('liby', Lib('y'))]) as args:
Bug 1077151 - Always use expandlibs descriptors when they exist. r=mshal Currently, when there is both an expandlibs descriptor and an actual static library, expandlibs picks the static library. This has the side effect that if there are object files in the static library that aren't directly used, they're dropped when linking, even when they export symbols that would be exported in the final linked binary. In most cases in the code base, files are not dropped that way. The most notable counter-example is xpcomglue, where actually not dropping files leads to link failure because of missing symbols those files reference (yes, that would tend to say the glue is broken in some way). On the opposite side, there is mozglue, which does have both a descriptor and a static library (the latter being necessary for the SDK), and that linking as a static library drops files that shouldn't be dropped (like jemalloc). We're currently relying on -Wl,--whole-archive for those files not to be dropped, but that won't really be possible without much hassle in a world where mozglue dependencies live in moz.build land. Switching expandlibs to use descriptors when they exist, even when there is a static library (so, the opposite of the current behavior) allows to drop -Wl,--whole-archive and prepare for a better future. However, as mentioned, xpcomglue does still require to be linked through the static library, so we need to make it a static library only. To achieve that, we make NO_EXPAND_LIBS now actually mean no expandlibs and use that to build the various different xpcomglues.
2014-10-04 05:33:00 +04:00
files = self.files + self.liby_files + self.libx_files
self.assertRelEqual(args, ['foo', '-bar'] + files)
extracted = {}
# ExpandArgsMore also has an extra method extracting static libraries
# when possible
args.extract()
# With AR_EXTRACT, it uses the descriptors when there are, and
# actually
# extracts the remaining libraries
extracted_args = []
for f in files:
if f.endswith(config.LIB_SUFFIX):
base = os.path.splitext(os.path.basename(f))[0]
# On the first iteration, we test the behavior of
# extracting archives that don't have a copy of their
# contents next to them, which is to use the file
# extracted from the archive in a temporary directory.
# On the second iteration, we test extracting archives
# that do have a copy of their contents next to them,
# in which case those contents are used instead of the
# temporarily extracted files.
if iteration == 1:
extracted_args.extend(sorted(extracted[base]))
else:
dirname = os.path.dirname(f[len(self.tmpdir)+1:])
if base.endswith('f'):
dirname = os.path.join(dirname, 'foo', 'bar')
extracted_args.extend([self.tmpfile(dirname, Obj(base)), self.tmpfile(dirname, Obj(base + '2'))])
else:
extracted_args.append(f)
self.assertRelEqual(args, ['foo', '-bar'] + extracted_args)
tmp = args.tmp
# Check that all temporary files are properly removed
self.assertEqual(True, all([not os.path.exists(f) for f in tmp]))
# Create archives contents next to them for the second iteration.
base = os.path.splitext(Lib('_'))[0]
self.touch(self.tmpfile(Obj(base.replace('_', suffix))) for suffix in ('a', 'a2', 'd', 'd2'))
try:
os.makedirs(self.tmpfile('foo', 'bar'))
except:
pass
self.touch(self.tmpfile('foo', 'bar', Obj(base.replace('_', suffix))) for suffix in ('f', 'f2'))
self.touch(self.tmpfile('liby', Obj(base.replace('_', suffix))) for suffix in ('z', 'z2'))
# Restore subprocess.call and subprocess.check_output
subprocess.call = subprocess_call
subprocess.check_output = subprocess_check_output
class FakeProcess(object):
def __init__(self, out, err = ''):
self.out = out
self.err = err
def communicate(self):
return (self.out, self.err)
OBJDUMPS = {
'foo.o': '''
00000000 g F .text\t00000001 foo
00000000 g F .text._Z6foobarv\t00000001 _Z6foobarv
00000000 g F .text.hello\t00000001 hello
00000000 g F .text._ZThn4_6foobarv\t00000001 _ZThn4_6foobarv
''',
'bar.o': '''
00000000 g F .text.hi\t00000001 hi
00000000 g F .text.hot._Z6barbazv\t00000001 .hidden _Z6barbazv
''',
}
PRINT_ICF = '''
ld: ICF folding section '.text.hello' in file 'foo.o'into '.text.hi' in file 'bar.o'
ld: ICF folding section '.foo' in file 'foo.o'into '.foo' in file 'bar.o'
'''
class SubprocessPopen(object):
def __init__(self, test):
self.test = test
def __call__(self, args, stdout = None, stderr = None):
self.test.assertEqual(stdout, subprocess.PIPE)
self.test.assertEqual(stderr, subprocess.PIPE)
if args[0] == 'objdump':
self.test.assertEqual(args[1], '-t')
self.test.assertTrue(args[2] in OBJDUMPS)
return FakeProcess(OBJDUMPS[args[2]])
else:
return FakeProcess('', PRINT_ICF)
class TestSectionFinder(unittest.TestCase):
def test_getSections(self):
'''Test SectionFinder'''
# Divert subprocess.Popen
subprocess_popen = subprocess.Popen
subprocess.Popen = SubprocessPopen(self)
config.EXPAND_LIBS_ORDER_STYLE = 'linkerscript'
config.OBJ_SUFFIX = '.o'
config.LIB_SUFFIX = '.a'
finder = SectionFinder(['foo.o', 'bar.o'])
self.assertEqual(finder.getSections('foobar'), [])
self.assertEqual(finder.getSections('_Z6barbazv'), ['.text.hot._Z6barbazv'])
self.assertEqual(finder.getSections('_Z6foobarv'), ['.text._Z6foobarv', '.text._ZThn4_6foobarv'])
self.assertEqual(finder.getSections('_ZThn4_6foobarv'), ['.text._Z6foobarv', '.text._ZThn4_6foobarv'])
subprocess.Popen = subprocess_popen
class TestSymbolOrder(unittest.TestCase):
def test_getOrderedSections(self):
'''Test ExpandMoreArgs' _getOrderedSections'''
# Divert subprocess.Popen
subprocess_popen = subprocess.Popen
subprocess.Popen = SubprocessPopen(self)
config.EXPAND_LIBS_ORDER_STYLE = 'linkerscript'
config.OBJ_SUFFIX = '.o'
config.LIB_SUFFIX = '.a'
config.LD_PRINT_ICF_SECTIONS = ''
args = ExpandArgsMore(['foo', '-bar', 'bar.o', 'foo.o'])
self.assertEqual(args._getOrderedSections(['_Z6foobarv', '_Z6barbazv']), ['.text._Z6foobarv', '.text._ZThn4_6foobarv', '.text.hot._Z6barbazv'])
self.assertEqual(args._getOrderedSections(['_ZThn4_6foobarv', '_Z6barbazv']), ['.text._Z6foobarv', '.text._ZThn4_6foobarv', '.text.hot._Z6barbazv'])
subprocess.Popen = subprocess_popen
def test_getFoldedSections(self):
'''Test ExpandMoreArgs' _getFoldedSections'''
# Divert subprocess.Popen
subprocess_popen = subprocess.Popen
subprocess.Popen = SubprocessPopen(self)
config.LD_PRINT_ICF_SECTIONS = '-Wl,--print-icf-sections'
args = ExpandArgsMore(['foo', '-bar', 'bar.o', 'foo.o'])
self.assertEqual(args._getFoldedSections(), {'.text.hello': ['.text.hi'], '.text.hi': ['.text.hello']})
subprocess.Popen = subprocess_popen
def test_getOrderedSectionsWithICF(self):
'''Test ExpandMoreArgs' _getOrderedSections, with ICF'''
# Divert subprocess.Popen
subprocess_popen = subprocess.Popen
subprocess.Popen = SubprocessPopen(self)
config.EXPAND_LIBS_ORDER_STYLE = 'linkerscript'
config.OBJ_SUFFIX = '.o'
config.LIB_SUFFIX = '.a'
config.LD_PRINT_ICF_SECTIONS = '-Wl,--print-icf-sections'
args = ExpandArgsMore(['foo', '-bar', 'bar.o', 'foo.o'])
self.assertEqual(args._getOrderedSections(['hello', '_Z6barbazv']), ['.text.hello', '.text.hi', '.text.hot._Z6barbazv'])
self.assertEqual(args._getOrderedSections(['_ZThn4_6foobarv', 'hi', '_Z6barbazv']), ['.text._Z6foobarv', '.text._ZThn4_6foobarv', '.text.hi', '.text.hello', '.text.hot._Z6barbazv'])
subprocess.Popen = subprocess_popen
if __name__ == '__main__':
mozunit.main()