Summary: | Dead-lock with --help on Qt 5.6 | ||
---|---|---|---|
Product: | [Plasma] Breeze | Reporter: | Martin Flöser <mgraesslin> |
Component: | QStyle | Assignee: | Hugo Pereira Da Costa <hugo.pereira.da.costa> |
Status: | RESOLVED WORKSFORME | ||
Severity: | grave | CC: | aleixpol, bhush94, cpigat242, heiko.becker, hugo.pereira.da.costa, kde |
Priority: | NOR | ||
Version: | 5.5.95 | ||
Target Milestone: | --- | ||
Platform: | Other | ||
OS: | Linux | ||
Latest Commit: | Version Fixed In: | ||
Sentry Crash Report: | |||
Attachments: | test patch (not to be committed) |
Description
Martin Flöser
2016-03-16 07:34:52 UTC
I noticed this problem before, but it vanishes for me if qtbase is built with this patch: https://codereview.qt-project.org/#/c/151496/ @Heiko Interesting. Indeed breeze does connect to some dbus signals in its constructor. @Martin, rather than patching Qt, maybe you can first try the following patch to breeze ? (which just comments the dbus connections) ? I cannot test here: do not have recent enough Qt5 (I have 5.5.1) Created attachment 97922 [details]
test patch (not to be committed)
yes, the patch solves the problem ok. So definitly related to the issue reported upstream. What should we do ? Close as upstream and wait for the patch to go in ? Find a workaround ? (no clue how) Comment out the code ? (in breeze and oxygen) ? The debus calls are necessary so that the widget style is updated "on fly" when the configuration is changes via system settings (or breeze-settings, or oxygen-settings) On one hand Locked application is somewhat a big deal, but on the other hand this only affects corner cases (--help) wdyt ? (In reply to Hugo Pereira Da Costa from comment #5) > On one hand Locked application is somewhat a big deal, but on the other hand > this only affects corner cases (--help) nah, it affects more. Every error handling case going directly to ::exit(). I assume it means that one cannot do a kwin_x11 --replace any more and similar things. What I wanted to try is whether an explicit disconnect in the dtor could help. (In reply to Martin Gräßlin from comment #6) > What I wanted to try is whether an explicit disconnect in the dtor could > help. nope, that just results in a crash. The Qt patch fixes the issue for me. This seems to be not true anymore. |