Bug 90091
| Summary: | [Qt] REGRESSION(r121348): It made 500+ tests flakey | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Csaba Osztrogonác <ossy> |
| Component: | New Bugs | Assignee: | Nobody <webkit-unassigned> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | Critical | CC: | abecsi, achicu, hausmann, ossy, simon.fraser |
| Priority: | P1 | Keywords: | Qt, QtTriaged |
| Version: | 528+ (Nightly build) | ||
| Hardware: | All | ||
| OS: | All | ||
| Bug Depends on: | |||
| Bug Blocks: | 89475 | ||
Csaba Osztrogonác
http://trac.webkit.org/changeset/121349 made 500+ tests flakey on Qt bots
and it made the bots exit early, so the Qt bot is useless because of this bug.
http://build.webkit.org/builders/Qt%20Linux%20Release/builds/48817
http://build.webkit.sed.hu/builders/x86-64%20Linux%20Qt%20Release/builds/39264
Could you check it, please?
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Csaba Osztrogonác
András, have you got any idea what caused this annoying sideeffect bug?
Simon, what should we do in this case? Skip 500+ test? Rollout the patch which cause it? Maybe it is innocent and there are a serious bug in QtWebKit somewhere ... Or leave the bots broken?
Csaba Osztrogonác
I forgot to mention that it is Qt-WK1 related bug. It is valid
on Qt 4.8 and Qt 5.0 WebKit, but everything works on Qt 5.0 WK2.
Simon Hausmann
(In reply to comment #1)
> András, have you got any idea what caused this annoying sideeffect bug?
>
> Simon, what should we do in this case? Skip 500+ test? Rollout the patch which cause it? Maybe it is innocent and there are a serious bug in QtWebKit somewhere ... Or leave the bots broken?
My first impression is as well that the patch is innocent. Unfortunately I'm without a build machine for the next couple of days, otherwise I'd give this a try myself, but here's what I'd do:
1) Try to reproduce the failures locally
2) See if reverting just that patch _does_ fix it (I doubt it)
3) Bisect down to the change that caused it
or
4) Find out that it's some weird miscompilation or unclean build issue that this particular patch just uncovered by accident.
Csaba Osztrogonác
r121349 is innocent, http://trac.webkit.org/changeset/121348 is the culprit.
With r121347 everything works, and with r121348 tests are flakey.
I tried reverting r121348 on ToT, and it solved the problem for me.
Now I'm trying to find what caused the problem. My first idea is that
the new css3/filters/blur-filter-page-scroll.html does something wrong.
Csaba Osztrogonác
Oh, I got it. AcceleratedCompositingEnabled and MockScrollBarsEnabled isn't reset after each test in Qt's DRT. :-/ What a pathetic bug!
Here is the bug report for MockScrollBarsEnabled - https://bugs.webkit.org/show_bug.cgi?id=87680
Csaba Osztrogonác
*** This bug has been marked as a duplicate of bug 87680 ***