Also, move the --timeout=900 flag from the recipe to the mb.py script. That way it is passed as an arg to gtest-parallel and not to the test executable.
BUG=webrtc:7568
NOTRY=True
Review-Url: https://codereview.webrtc.org/2862803002
Cr-Commit-Position: refs/heads/master@{#18015}
It makes the logs hard to read, and swarming doesn't output colors anyway.
Also:
- Fix a bug where we tried to access output_dir when combining the logs even if the flag was not set.
- Update a comment explaining why we just 'eat' the --isolated-script-test-chartjson-output flag.
NOTRY=True
BUG=webrtc:7524
Review-Url: https://codereview.webrtc.org/2843263005
Cr-Commit-Position: refs/heads/master@{#17973}
Due to recent Chrome infra changes in
https://chromium-review.googlesource.com/c/472290/
tests running on swarming are now assumed to emit JSON results
or will be marked as failing. This requires us to use our
gtest-parallel wrapper for all our Swarming tests
(or implement the --isolated-script-test-output flag, which
normally only is implemented by the Chromium test launcher).
The low_bandwidth_audio_test can actually run in parallel,
so just change that.
The webrtc_nonparallel_tests cannot, so this CL changes MB
to pass --workers=1 flag to gtest-parallel, which makes the
tests run in sequence. This adds a little confusion but the root
problem is really that our gtest-parallel script [1] does a lot more
than just running the tests in parallel these days, so it should
probably be renamed.
Also make sure gtest-parallel-wrapper.py [2] consumes the
--isolated-script-test-chartjson-output flag (unused) so we don't
pass it on to the test executable.
[1]: https://chromium.googlesource.com/external/github.com/google/gtest-parallel/+/master/gtest-parallel
[2]: https://chromium.googlesource.com/external/webrtc/+/master/tools-webrtc/gtest-parallel-wrapper.py
BUG=709988
TBR=ehmaldonado@webrtc.org
NOTRY=True
Review-Url: https://codereview.webrtc.org/2806373002
Cr-Commit-Position: refs/heads/master@{#17646}
