def test_old_ldlogger(self):
"""
Test log file parsing escape behaviour with pre-2017 Q2 LD-LOGGER.
"""
logfile = os.path.join(self.__test_files, "ldlogger-old.json")
# LD-LOGGER before http://github.com/Ericsson/codechecker/pull/631
# used an escape mechanism that, when parsed by the log parser via
# shlex, made CodeChecker parse arguments with multiword string
# literals in them be considered as "file" (instead of compile option),
# eventually ignored by the command builder, thus lessening analysis
# accuracy, as defines were lost.
#
# Logfile contains "-DVARIABLE="some value"".
#
# There is no good way to back-and-forth convert in log_parser or
# option_parser, so here we aim for a non-failing stalemate of the
# define being considered a file and ignored, for now.
build_action = log_parser.parse_log(logfile, ParseLogOptions())[0]
results = option_parser.parse_options(build_action.original_command)
self.assertEqual(' '.join(results.files),
r'"-DVARIABLE="some value"" /tmp/a.cpp')
self.assertEqual(len(build_action.analyzer_options), 0)
def test_compile_with_include_paths(self):
"""
sysroot should be detected as compiler option because it is needed
for the analyzer too to search for headers.
"""
source_files = ["main.cpp", "test.cpp"]
compiler_options = [
"-std=c++11", "-include/include/myheader.h",
"-include /include/myheader2.h", "--include",
"/include/myheader3.h", "--sysroot", "/home/sysroot",
"--sysroot=/home/sysroot3", "-isysroot /home/isysroot",
"-isysroot/home/isysroot2", "-I/home/test", "-I /home/test2",
"-idirafter /dirafter1", "-idirafter/dirafter2"
]
linker_options = ["-L/home/test_lib", "-lm"]
build_cmd = "g++ -o myapp " + \
' '.join(compiler_options) + ' ' + \
' '.join(linker_options) + ' ' + \
' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
co_no_space = []
for c in compiler_options:
co_no_space.append(c.replace(" ", ""))
print(set(co_no_space))
print(set(res.compile_opts))
self.assertTrue(set(source_files) == set(res.files))
self.assertTrue(set(co_no_space) == set(res.compile_opts))
self.assertTrue(set(linker_options) == set(res.link_opts))
self.assertEqual(ActionType.COMPILE, res.action)
def test_link_only_multiple_files(self):
"""
Should be link if only object files are in the command.
"""
build_cmd = "g++ -o fubar foo.o main.o bar.o -lm"
res = option_parser.parse_options(build_cmd)
print(res)
self.assertEquals(ActionType.LINK, res.action)
def test_ignore_flags(self):
ignore = [
"-Werror", "-MT hello", "-M", "-fsyntax-only",
"-mfloat-gprs=double", "-mfloat-gprs=yes"
]
build_cmd = "g++ {} main.cpp".format(' '.join(ignore))
res = option_parser.parse_options(build_cmd)
self.assertEqual(res.compile_opts, ["-fsyntax-only"])
def test_ignore_flags(self):
ignore = [
"-Werror", "-MT hello", "-M", "-fsyntax-only",
"-mfloat-gprs=double", "-mfloat-gprs=yes", "-mabi=spe",
"-mabi=eabi", '-Xclang', '-mllvm', '-Xclang',
'-instcombine-lower-dbg-declare=0'
]
build_cmd = "g++ {} main.cpp".format(' '.join(ignore))
res = option_parser.parse_options(build_cmd)
self.assertEqual(res.compile_opts, ["-fsyntax-only"])
def test_compile_onefile(self):
"""
Test the compiler command of one file.
"""
source_files = ["main.cpp"]
build_cmd = "g++ -c " + ' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertEquals(ActionType.COMPILE, res.action)
def test_build_multiplefiles(self):
"""
Test the build command of multiple files.
"""
source_files = ["lib.cpp", "main.cpp"]
build_cmd = "g++ -o main " + ' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertEquals(ActionType.COMPILE, res.action)
def test_build_onefile(self):
"""
Test the build command of a simple file.
"""
source_files = ["main.cpp"]
build_cmd = "g++ -o main " + ' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertTrue(ActionType.COMPILE, res.action)
self.assertEquals(0, len(res.compile_opts))
def test_preprocess_onefile(self):
"""
Test the preprocess command of one file.
"""
source_files = ["main.c"]
build_cmd = "gcc -E " + ' '.join(source_files)
print(build_cmd)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertEqual(ActionType.PREPROCESS, res.action)
def test_compiler_toolchain(self):
"""
Test if compiler toolchain is parsed and forwarded properly.
"""
source_files = ["main.cpp"]
compiler_options = ["--gcc-toolchain=/home/user/mygcctoolchain"]
build_cmd = "g++ -c " + \
' '.join(compiler_options) + ' ' + \
' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(compiler_options) == set(res.compile_opts))
self.assertEqual(ActionType.COMPILE, res.action)
def test_compile_optimized(self):
"""
Test if the compilation arguments is
detected correctly from the command line.
"""
source_files = ["main.cpp"]
compiler_options = ["-O3"]
build_cmd = "g++ -c " + \
' '.join(compiler_options) + ' ' + \
' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(compiler_options) == set(res.compile_opts))
self.assertEqual(ActionType.COMPILE, res.action)
def test_compile_arch(self):
"""
Test if the compilation architecture is
detected correctly from the command line.
"""
source_files = ["main.c"]
arch = 'x86_64'
build_cmd = "gcc -c -arch " + arch + " " + ' '.join(source_files)
print(build_cmd)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertEqual(arch, res.arch)
self.assertEqual(ActionType.COMPILE, res.action)
def test_compile_lang(self):
"""
Test if the compilation language is
detected correctly from the command line.
"""
source_files = ["main.c"]
lang = 'c'
build_cmd = "gcc -c -x " + lang + " " + ' '.join(source_files)
print(build_cmd)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertEqual(lang, res.lang)
self.assertEqual(ActionType.COMPILE, res.action)
def test_link_with_include_paths(self):
"""
Should be link if only object files are in the command.
"""
object_files = ["foo.o", "main.o", "bar.o"]
compiler_options = [
"--sysroot", "/home/sysroot", "-isysroot/home/isysroot",
"-I/home/test"
]
linker_options = ["-L/home/test_lib", "-lm"]
build_cmd = "g++ -o fubar " + \
' '.join(compiler_options) + ' ' + \
' '.join(linker_options) + ' ' + \
' '.join(object_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(object_files) == set(res.files))
self.assertTrue(set(compiler_options) == set(res.compile_opts))
self.assertTrue(set(linker_options) == set(res.link_opts))
self.assertEqual(ActionType.LINK, res.action)
def test_compile_with_include_paths(self):
"""
sysroot should be detected as compiler option because it is needed
for the analyzer too to search for headers.
"""
source_files = ["main.cpp", "test.cpp"]
compiler_options = ["-sysroot", "/home/sysroot",
"-isysroot", "/home/isysroot",
"-I/home/test"]
linker_options = ["-L/home/test_lib", "-lm"]
build_cmd = "g++ -o myapp " + \
' '.join(compiler_options) + ' ' + \
' '.join(linker_options) + ' ' + \
' '.join(source_files)
res = option_parser.parse_options(build_cmd)
print(res)
self.assertTrue(set(source_files) == set(res.files))
self.assertTrue(set(compiler_options) == set(res.compile_opts))
self.assertTrue(set(linker_options) == set(res.link_opts))
self.assertEqual(ActionType.COMPILE, res.action)
def parse_compile_commands_json(logfile, add_compiler_defaults=False):
import json
LOG.debug('parse_compile_commands_json: ' + str(add_compiler_defaults))
actions = []
filtered_build_actions = {}
logfile.seek(0)
data = json.load(logfile)
compiler_defines = {}
compiler_includes = {}
counter = 0
for entry in data:
sourcefile = entry['file']
lang = option_parser.get_language(sourcefile[sourcefile.rfind('.'):])
if not lang:
continue
action = build_action.BuildAction(counter)
command = entry['command']
results = option_parser.parse_options(command)
action.original_command = command
action.analyzer_options = results.compile_opts
action.lang = results.lang
action.target = results.arch
# store the compiler built in include paths
# and defines
if add_compiler_defaults and results.compiler:
if not (results.compiler in compiler_defines):
compiler_defines[results.compiler] = \
get_compiler_defines(results.compiler)
compiler_includes[results.compiler] = \
get_compiler_includes(results.compiler)
action.compiler_defines = compiler_defines[results.compiler]
action.compiler_includes = compiler_includes[results.compiler]
if results.action == option_parser.ActionType.COMPILE or \
results.action == option_parser.ActionType.LINK:
action.skip = False
# TODO: check arch.
action.directory = entry['directory']
action.sources = sourcefile
# Filter out duplicate compilation commands.
unique_key = action.cmp_key
if filtered_build_actions.get(unique_key) is None:
filtered_build_actions[unique_key] = action
del action
counter += 1
for ba_hash, ba in filtered_build_actions.items():
actions.append(ba)
return actions
def parse_compile_commands_json(logfile, add_compiler_defaults=False):
import json
LOG.debug('parse_compile_commands_json: ' + str(add_compiler_defaults))
actions = []
filtered_build_actions = {}
logfile.seek(0)
data = json.load(logfile)
compiler_defines = {}
compiler_includes = {}
counter = 0
for entry in data:
sourcefile = entry['file']
if not os.path.isabs(sourcefile):
# Newest versions of intercept-build can create the 'file' in the
# JSON Compilation Database as a relative path.
sourcefile = os.path.join(os.path.abspath(entry['directory']),
sourcefile)
lang = option_parser.get_language(sourcefile[sourcefile.rfind('.'):])
if not lang:
continue
action = build_action.BuildAction(counter)
if 'command' in entry:
command = entry['command']
# Old versions of intercept-build (confirmed to those shipping
# with upstream clang-5.0) do escapes in another way:
# -DVARIABLE="a b" becomes -DVARIABLE=\"a b\" in the output.
# This would be messed up later on by options_parser, so need a
# fix here. (Should be removed once we are sure noone uses this
# intercept-build anymore!)
if r'\"' in command:
command = command.replace(r'\"', '"')
elif 'arguments' in entry:
# Newest versions of intercept-build create an argument vector
# instead of a command string.
command = ' '.join(entry['arguments'])
else:
raise KeyError("No valid 'command' or 'arguments' entry found!")
results = option_parser.parse_options(command)
action.original_command = command
action.analyzer_options = results.compile_opts
action.lang = results.lang
action.target = results.arch
# Store the compiler built in include paths and defines.
if add_compiler_defaults and results.compiler:
if not (results.compiler in compiler_defines):
# Fetch defaults from the compiler,
# make sure we use the correct architecture.
extra_opts = []
for regex in COMPILE_OPTS_FWD_TO_DEFAULTS_GETTER:
pattern = re.compile(regex)
for comp_opt in action.analyzer_options:
if re.match(pattern, comp_opt):
extra_opts.append(comp_opt)
compiler_defines[results.compiler] = \
get_compiler_defines(results.compiler, extra_opts)
compiler_includes[results.compiler] = \
get_compiler_includes(results.compiler, extra_opts)
action.compiler_defines = compiler_defines[results.compiler]
action.compiler_includes = compiler_includes[results.compiler]
if results.action == option_parser.ActionType.COMPILE or \
results.action == option_parser.ActionType.LINK:
action.skip = False
# TODO: Check arch.
action.directory = entry['directory']
action.sources = sourcefile
# Filter out duplicate compilation commands.
unique_key = action.cmp_key
if filtered_build_actions.get(unique_key) is None:
filtered_build_actions[unique_key] = action
del action
counter += 1
for ba_hash, ba in filtered_build_actions.items():
actions.append(ba)
return actions
def parse_compile_commands_json(logfile, parseLogOptions):
"""
logfile: is a compile command json
"""
output_path = parseLogOptions.output_path
if output_path is not None:
remove_file_if_exists(
os.path.join(output_path, compiler_includes_dump_file))
remove_file_if_exists(
os.path.join(output_path, compiler_target_dump_file))
actions = []
filtered_build_actions = {}
data = json.load(logfile)
compiler_includes = {}
compiler_target = {}
counter = 0
for entry in data:
sourcefile = entry['file']
if not os.path.isabs(sourcefile):
# Newest versions of intercept-build can create the 'file' in the
# JSON Compilation Database as a relative path.
sourcefile = os.path.join(os.path.abspath(entry['directory']),
sourcefile)
lang = option_parser.get_language(sourcefile[sourcefile.rfind('.'):])
if not lang:
continue
action = build_action.BuildAction(counter)
if 'command' in entry:
command = entry['command']
# Old versions of intercept-build (confirmed to those shipping
# with upstream clang-5.0) do escapes in another way:
# -DVARIABLE="a b" becomes -DVARIABLE=\"a b\" in the output.
# This would be messed up later on by options_parser, so need a
# fix here. (Should be removed once we are sure noone uses this
# intercept-build anymore!)
if r'\"' in command:
command = command.replace(r'\"', '"')
elif 'arguments' in entry:
# Newest versions of intercept-build create an argument vector
# instead of a command string.
command = ' '.join(entry['arguments'])
else:
raise KeyError("No valid 'command' or 'arguments' entry found!")
results = option_parser.parse_options(command)
action.original_command = command
# If the original include directory could not be found
# in the filesystem, it is possible that it was provided
# relative to the working directory in the compile json.
compile_opts = results.compile_opts
for i, opt in enumerate(compile_opts):
if opt.startswith('-I'):
inc_dir = opt[2:].strip()
if not os.path.isdir(inc_dir):
compile_opts[i] = '-I' + \
os.path.join(entry['directory'], inc_dir)
action.analyzer_options = compile_opts
action.lang = results.lang
action.target = results.arch
action.output = results.output
add_compiler_defaults = True
# With gcc-toolchain a non default compiler toolchain can be set.
# Clang will search for include paths and libraries based on the
# gcc-toolchain parameter.
# Detecting extra include paths from the host compiler could
# conflict with this.
# For example if the compiler in the compile command is clang
# and gcc-toolchain is set we will get the include paths
# for clang and not for the compiler set in gcc-toolchain.
# This can cause missing headers during the analysis.
toolchain = gcc_toolchain.toolchain_in_args(action.analyzer_options)
if toolchain:
add_compiler_defaults = False
# Store the compiler built in include paths and defines.
if add_compiler_defaults and results.compiler:
if not (results.compiler in compiler_includes):
# Fetch defaults from the compiler,
# make sure we use the correct architecture.
extra_opts = []
for regex in COMPILE_OPTS_FWD_TO_DEFAULTS_GETTER:
pattern = re.compile(regex)
for comp_opt in action.analyzer_options:
if re.match(pattern, comp_opt):
extra_opts.append(comp_opt)
compiler_includes[results.compiler] = \
get_compiler_includes(parseLogOptions, results.compiler,
results.lang, results.compile_opts,
extra_opts)
if not (results.compiler in compiler_target):
compiler_target[results.compiler] = \
get_compiler_target(parseLogOptions, results.compiler)
action.compiler_includes = compiler_includes[results.compiler]
action.target = compiler_target[results.compiler]
if results.action != option_parser.ActionType.COMPILE:
action.skip = True
# TODO: Check arch.
action.directory = entry['directory']
action.sources = sourcefile
# Filter out duplicate compilation commands.
unique_key = action.cmp_key
if filtered_build_actions.get(unique_key) is None:
filtered_build_actions[unique_key] = action
del action
counter += 1
for _, ba in filtered_build_actions.items():
actions.append(ba)
return actions