[Code Coverage] Add support for component build + update documentation.
Bug: 831939
Change-Id: I2804796894045dc69aa8309aa3ab34e6c80b38d7
Reviewed-on: https://chromium-review.googlesource.com/1010464
Commit-Queue: Max Moroz <[email protected]>
Reviewed-by: Yuke Liao <[email protected]>
Reviewed-by: Abhishek Arya <[email protected]>
Cr-Commit-Position: refs/heads/master@{#553211}diff --git a/build/config/coverage/coverage.gni b/build/config/coverage/coverage.gni
index ee4734cf..61716dd 100644
--- a/build/config/coverage/coverage.gni
+++ b/build/config/coverage/coverage.gni
@@ -5,21 +5,9 @@
import("//build/toolchain/toolchain.gni")
declare_args() {
- # Enable Clang's Source-based Code Coverage. This should not be used with
- # is_component_build=true.
- # TODO(metzman): Add an assertion that will prevent is_component_build=true
- # from being passed as an argument to gn, but will not force toolchains that
- # set is_component_build=true to set it to false.
+ # Enable Clang's Source-based Code Coverage.
use_clang_coverage = false
}
assert(!use_clang_coverage || is_clang,
"Clang Source-based Code Coverage requires clang.")
-
-assert(
- !(use_clang_coverage && is_component_build),
- "Clang Source-based Code Coverage requires \"is_component_build=false\" " +
- "flag because: There will be no coverage info for libraries in " +
- "component builds and \"is_component_build\" is set to true by " +
- "\"is_debug\" unless it is explicitly set to false.$0x0ATry adding " +
- "\"is_component_build=false\" to your GN args.")
diff --git a/docs/code_coverage.md b/docs/code_coverage.md
index bbc7e8b8..0245e80 100644
--- a/docs/code_coverage.md
+++ b/docs/code_coverage.md
@@ -8,6 +8,8 @@
* [Step 2 Create Raw Profiles](#step-2-create-raw-profiles)
* [Step 3 Create Indexed Profile](#step-3-create-indexed-profile)
* [Step 4 Create Coverage Reports](#step-4-create-coverage-reports)
+- [Contacts](#contacts)
+- [FAQ](#faq)
Chromium uses Clang source-based code coverage, this [documentation] explains
how to use Clang’s source-based coverage features in general.
@@ -135,12 +137,24 @@
For more information on how to use llvm-cov, please refer to the [guide].
-## Reporting problems
+## Contacts
+
+### Reporting problems
For any breakage report and feature requests, please [file a bug].
-## Mail list
+### Mailing list
For questions and general discussions, please join [chrome-code-coverage group].
+## FAQ
+
+### Can I use `is_component_build=true` for code coverage build?
+
+Yes, code coverage instrumentation works with both component and non-component
+builds. Component build is usually faster to compile, but can be up to several
+times slower to run with code coverage instrumentation. For more information,
+see [crbug.com/831939](https://crbug.com/831939).
+
+
[documentation]: https://clang.llvm.org/docs/SourceBasedCodeCoverage.html
[coverage script]: https://cs.chromium.org/chromium/src/tools/code_coverage/coverage.py
[code coverage report directory view]: images/code_coverage_directory_view.png
diff --git a/tools/code_coverage/coverage.py b/tools/code_coverage/coverage.py
index a6a1c65..6d8b6e4 100755
--- a/tools/code_coverage/coverage.py
+++ b/tools/code_coverage/coverage.py
@@ -313,6 +313,50 @@
html_file.write(html_header + html_table + html_footer)
+def _GetSharedLibraries(binary_paths):
+ """Returns set of shared libraries used by specified binaries."""
+ libraries = set()
+ cmd = []
+ shared_library_re = None
+
+ if sys.platform.startswith('linux'):
+ cmd.extend(['ldd'])
+ shared_library_re = re.compile(
+ r'.*\.so\s=>\s(.*' + BUILD_DIR + '.*\.so)\s.*')
+ elif sys.platform.startswith('darwin'):
+ cmd.extend(['otool', '-L'])
+ shared_library_re = re.compile(r'\s+(@rpath/.*\.dylib)\s.*')
+ else:
+ assert False, ('Cannot detect shared libraries used by the given targets.')
+
+ assert shared_library_re is not None
+
+ cmd.extend(binary_paths)
+ output = subprocess.check_output(cmd)
+
+ for line in output.splitlines():
+ m = shared_library_re.match(line)
+ if not m:
+ continue
+
+ shared_library_path = m.group(1)
+ if sys.platform.startswith('darwin'):
+ # otool outputs "@rpath" macro instead of the dirname of the given binary.
+ shared_library_path = shared_library_path.replace('@rpath', BUILD_DIR)
+
+ assert os.path.exists(shared_library_path), ('Shared library "%s" used by '
+ 'the given target(s) does not '
+ 'exist.' % shared_library_path)
+ with open(shared_library_path) as f:
+ data = f.read()
+
+ # Do not add non-instrumented libraries. Otherwise, llvm-cov errors outs.
+ if '__llvm_cov' in data:
+ libraries.add(shared_library_path)
+
+ return list(libraries)
+
+
def _GetHostPlatform():
"""Returns the host platform.
@@ -823,23 +867,25 @@
"""Runs a single command and generates a profraw data file."""
# Per Clang "Source-based Code Coverage" doc:
#
- # "%p" expands out to the process ID.
+ # "%p" expands out to the process ID. It's not used by this scripts due to:
+ # 1) If a target program spawns too many processess, it may exhaust all disk
+ # space available. For example, unit_tests writes thousands of .profraw
+ # files each of size 1GB+.
+ # 2) If a target binary uses shared libraries, coverage profile data for them
+ # will be missing, resulting in incomplete coverage reports.
#
# "%Nm" expands out to the instrumented binary's signature. When this pattern
# is specified, the runtime creates a pool of N raw profiles which are used
# for on-line profile merging. The runtime takes care of selecting a raw
# profile from the pool, locking it, and updating it before the program exits.
- # If N is not specified (i.e the pattern is "%m"), it's assumed that N = 1.
# N must be between 1 and 9. The merge pool specifier can only occur once per
# filename pattern.
#
- # "%p" is used when tests run in single process, however, it can't be used for
- # multi-process because each process produces an intermediate dump, which may
- # consume hundreds of gigabytes of disk space.
+ # "%1m" is used when tests run in single process, such as fuzz targets.
#
- # For "%Nm", 4 is chosen because it creates some level of parallelism, but
- # it's not too big to consume too much computing resource or disk space.
- profile_pattern_string = '%p' if _IsFuzzerTarget(target) else '%4m'
+ # For other cases, "%4m" is chosen as it creates some level of parallelism,
+ # but it's not too big to consume too much computing resource or disk space.
+ profile_pattern_string = '%1m' if _IsFuzzerTarget(target) else '%4m'
expected_profraw_file_name = os.extsep.join(
[target, profile_pattern_string, PROFRAW_FILE_EXTENSION])
expected_profraw_file_path = os.path.join(OUTPUT_DIR,
@@ -1262,6 +1308,7 @@
logging.info('Generating code coverage report in html (this can take a while '
'depending on size of target!)')
+ binary_paths.extend(_GetSharedLibraries(binary_paths))
per_file_coverage_summary = _GeneratePerFileCoverageSummary(
binary_paths, profdata_file_path, absolute_filter_paths,
args.ignore_filename_regex)
