http/tests/misc/last-modified-parsing.html always fails on Windows EWS. Looks like a timezone issue (off by 8 hours): --- /home/buildbot/worker/Windows-EWS/build/layout-test-results/http/tests/misc/last-modified-parsing-expected.txt +++ /home/buildbot/worker/Windows-EWS/build/layout-test-results/http/tests/misc/last-modified-parsing-actual.txt @@ -3,34 +3,34 @@ -------- Frame: '<!--frame1-->' -------- -Fri, 21 Nov 2008 01:03:33 GMT +Fri, 21 Nov 2008 09:03:33 GMT -------- Frame: '<!--frame2-->' -------- -Fri, 21 Nov 2008 01:03:33 GMT +Fri, 21 Nov 2008 09:03:33 GMT -------- Frame: '<!--frame3-->' -------- -Fri, 21 Nov 1997 01:03:33 GMT +Fri, 21 Nov 1997 09:03:33 GMT -------- Frame: '<!--frame4-->' -------- -Fri, 21 Nov 2008 01:03:33 GMT +Fri, 21 Nov 2008 09:03:33 GMT -------- Frame: '<!--frame5-->' -------- -Thu, 21 Feb 2008 01:03:33 GMT +Thu, 21 Feb 2008 09:03:33 GMT -------- Frame: '<!--frame6-->' -------- -Sun, 03 Feb 2008 01:03:33 GMT +Sun, 03 Feb 2008 09:03:33 GMT -------- Frame: '<!--frame7-->' -------- -Mon, 03 Mar 2008 01:03:33 GMT +Mon, 03 Mar 2008 09:03:33 GMT
Committed r258892: <https://trac.webkit.org/changeset/258892/webkit> This should speed up the Windows EWS bot.
It's been failing for a while: <https://results.webkit.org/?suite=layout-tests&test=http%2Ftests%2Fmisc%2Flast-modified-parsing.html&platform=win> Looks like it regressed between r258625 (working) and r258626 (failing). <https://trac.webkit.org/changeset/258626/webkit> I wonder how a change to build scripts could have caused this regression? [GTK][WPE] Migrate to Flatpak-based dev SDK <https://bugs.webkit.org/show_bug.cgi?id=205658>
Is this bot running on a ... linux platform? :)
(In reply to Philippe Normand from comment #3) > Is this bot running on a ... linux platform? :) Of course not. I'm just reporting what I noticed in the results: <https://results.webkit.org/?suite=layout-tests&test=http%2Ftests%2Fmisc%2Flast-modified-parsing.html&platform=win> Note that there is shared Python and Perl code used by the Windows platform in build/test scripts, too, which were modified as part of the patch. (I did not review the patch to see if code shared by the Windows port was modified, though.)
As commented in the other bug, can you check if the TZ env var has any effect on your bots? I suspect this is the source of the problem, because if it's set on your bots, it is now white-listed into the test environment.
(In reply to Philippe Normand from comment #5) > As commented in the other bug, can you check if the TZ env var has any > effect on your bots? I suspect this is the source of the problem, because if > it's set on your bots, it is now white-listed into the test environment. Thanks Phillippe! I've added some folks to the Cc list that can check this.
There are two problems. run-webkit-tests set 'TZ' env var to 'US/Pacific' (Bug 186612). However, Windows' std::localtime doesn't understand it, and treats it GMT. 'TZ' env var should be 'PST8PDT' for Windows. _tzset | Microsoft Docs https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/tzset Date.parse() doesn't take 'TZ' env var into account. It always assumes the local timezone.