Bug 112381
| Summary: | [chromium] Intermittent timeouts of html5lib/generated/ tests on Release bots | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Kenneth Russell <kbr> |
| Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
| Status: | RESOLVED WONTFIX | ||
| Severity: | Normal | CC: | alancutter, eric, rafaelw, schenney, vsevik |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Bug Depends on: | 110876 | ||
| Bug Blocks: | |||
Kenneth Russell
https://bugs.webkit.org/show_bug.cgi?id=110876 was filed for timeouts of tests under html5lib/generated/ in Debug mode. There are still intermittent timeouts of these tests on webkit.org's Chromium Linux Release (Tests) bot. They seem to have started somewhere between http://trac.webkit.org/changeset/145575 and http://trac.webkit.org/changeset/145591 , but there aren't any changes in that range which would have affected these tests. The last build that ran completely cleanly was
http://build.webkit.org/builders/Chromium%20Linux%20Release%20%28Tests%29/builds/45898
and the first build that started reporting only these timeouts reliably was
http://build.webkit.org/builders/Chromium%20Linux%20Release%20%28Tests%29/builds/45912
html5lib/generated/run-tests16-data.html is the test reported most often but others show up as well including html5lib/generated/run-tests1-data.html and html5lib/generated/run-template-data.html .
Initially I'll mark run-tests16-data.html as slow.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Eric Seidel (no email)
Only the -data ones should be timing out. If the -write ones are, I'm concerned.
The -data timeouts are due to https://code.google.com/p/v8/issues/detail?id=2567
Kenneth Russell
Thanks, that's helpful information. So far I've only marked html5lib/generated/run-tests16-data.html as timing out on Linux Release. If any -write tests start timing out I'll tell you.
Eric Seidel (no email)
tests16 contains 191 sub-tests, which means 191 iframes created. In https://code.google.com/p/v8/issues/detail?id=2567 I noted that booting a v8 context takes about 1/70th a second in a Release build on my z620. So 191 / 70= 2.7, which is dangerously close to the 3s timeout. I can easily imagine this test (and other -data variants with many subtests) timing out on machines slower than a z620. :)
Kenneth Russell
The timeouts are happening on the Chromium Win Release (Tests) bot also. Adjusting synopsis and test expectations.
http://build.webkit.org/builders/Chromium%20Win%20Release%20%28Tests%29/builds/35145