Sync to r72870 and run "new-run-webkit-tests --lint-test-files". It should give errors that there are 3 duplicate lines on mac. It doesn't give any errors, but if you actually run the tests, it does. This means the presubmit for test_expectations isn't working.
When you say "it doesn't give any errors", what are you referring to? for me, it does print messages to stderr, but the return code is 0. check-webkit-style reports the error as well. Is that what you see/saw?
Created attachment 75241 [details] Patch
Per conversation w/ Ojan, it sounds like it was user error with new-run-webkit-tests and either user error with webkit-patch land-cowboy or a bug. For the former, he didn't specify --platform chromium-mac, so the command linted the platform/mac (webkit) version of the file, which was fine. Also, he ran land-cowboy, which skips the check-webkit-style presubmit check, which should have reported the error correctly (it does for me when I try to reproduce what he saw). That said, I'll use this bug as an excuse to clean up the --lint-test-files code and logging. The log output should be cleaner, we no longer use the 'print' statement, and the exception raising is a bit more straightforward. Also, I removed the 'suppress_errors' keyword param to the constructor, which appears to be unused (I don't remember why it was there in the first place). Also, in reviewing the patch you'll notice that platform/chromium was parsing the expectations files with is_lint_mode=True, and I've changed it to False. True was almost certainly the wrong value, and the new error propagation was causing things to fail in a way that the old code wouldn't (in the face of non-fatal errors).
Sounds like the bug title no longer reflects what this bug is about. It would be best to fix it.
good idea. done.
hmm ... this seems to have slipped through the cracks. Can someone please review this?
Committed r73654: <http://trac.webkit.org/changeset/73654>