Bug 87717 – Unresponsive WebProcesses can be mistaken for crashes.

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Stephanie Lewis]

WebKitTestRunner reports #CRASHED in two places. The first place, webProcessDidCrash, always writes out a crash report (eventually).
The second spot is when we are resetting the WebProcess.  See https://bugs.webkit.org/show_bug.cgi?id=52757

If it becomes unresponsive the WebProcess is killed, which for all intents and purposes looks like a crash.  However, this makes the WebKit2 tests look very flaky, and creates confusion because no crash logs can be found for the unresponsive processes.

The first step is separating the two types of crashes, and giving an error message when the process was just unresponsive, not crashing.

[Stephanie Lewis]

Created attachment 144497 [details]
patch

[WebKit Review Bot]

Attachment 144497 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Tools/ChangeLog', u'Tools/Scripts/webkitpy..." exit_code: 1
Tools/Scripts/webkitpy/layout_tests/port/webkit_unittest.py:346:  trailing whitespace  [pep8/W291] [5]
Total errors found: 1 in 4 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Dirk Pranke]

Comment on attachment 144497 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=144497&action=review

The patch looks fine. I'm not sure how you're defining "unresponsive" in WebKit2, but does it make sense to map these things onto TIMEOUTs rather than CRASHes?

> Tools/Scripts/webkitpy/layout_tests/port/webkit.py:588
> +            crash_log += '  Process failed to become responsive before timing out.'

See my comment in bug 87714 about possibly changing the interface to _get_crash_log() so that you don't have to do the .startswith() check here.

[Dirk Pranke]

Oh, and fix the style nit, obviously :).

[Stephanie Lewis]

committed http://trac.webkit.org/projects/webkit/changeset/119018