[webkit-qt] [week40] updating qt5?
oszi at inf.u-szeged.hu
Fri Oct 5 09:36:50 PDT 2012
Andras Becsi írta:
> On 5 October 2012 15:55, Osztrogonac Csaba <oszi at inf.u-szeged.hu> wrote:
>> Hi all,
>> I updated the Qt5 on the bots to 749b83f337ec82eff305ed8ac85473ad984c84fa .
>> You can find the build script here:https://github.com/ossy-szeged/qt5-tools
>> @rgabor: Could you update the ARM bots, please?
>> @azbest_hu: Could you update the Windows bot, please?
>> Unfortunately tst_qmltests_WebView WK2/API test still crashes, so we
>> don't know _anything_ about the ~100 testcases in it for a month ago.
>> ( https://bugs.webkit.org/show_bug.cgi?id=95968 )
> As I already indicated this crash is not reproducible outside the
> buildbot environment, I can run all the API tests locally and
> currently all succeed.
> The crash is most probably related to the buildbot environment.
> Try removing $HOME/.local/share/tst_qmltests_WebView
Of course I tried many thing before filing the bug and blaming the newer Qt5.
It is unrelated to the bot environment at all. You can easily reproduce this
crash on the reference platform with and without XVFB:
- Ubuntu 11.10 32/64 bit (default install) pps
- sedkit-env-qtwebkit metapackage from https://launchpad.net/~u-szeged/+archive/sedkit
- Qt5 build with this script: https://github.com/ossy-szeged/qt5-tools
(And we get same crashes on Debian Squeeze too.)
I think you use different distro (Arch), you have different packages with different
versions and you build Qt in a different way. (Un)fortunately Qt's and WebKit's
build system is too clever and switch on/off features automatically depending
on configure tests. Maybe we use totally different configs because of this.
The bug is still valid on the platform I mentioned. If you can't reproduce
the bug on your platform, it doesn't mean if it is invalid everywhere.
PS. It is a good news if all tests pass for you. But it would be great
if they pass on the bots to be able catch new regressions in time.
More information about the webkit-qt