https://bugs.kde.org/show_bug.cgi?id=363753
Yichao Yu <yyc1...@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |UPSTREAM
--- Comment #13 from Yichao Yu <yyc1...@gmail.com> ---
> Is this the same bug for which Beojan sent a patch (created by a Debian
> package maintainer)?
Yes, but the debian patch is wrong.
> I think, the bug "added" in commit 3d8622c
> https://quickgit.kde.org/?p=qtcurve.git&a=commitdiff&h=3d8622c419a32033e36e940e8cb09b591ad93e29
(The "bug" was added**). It's not a qtcurve bug, it's a QtDBus one and should
be fixed there. It just happens that in some configuration the QtDBus is
unloaded before QtCurve and after in other cases. The commit should only be
reverted if
1. Qt guarantees that any callback registered will not be called again
automatically (i.e. without unregistering in the destructor) when the plugin is
unloaded (noted that the destructor of the Style itself is never called AFAICT)
and/or, 2. the plugin is never unloaded (i.e. never unmap'd)
> The patch by the Debian maintainer(s) does something else; it removes the
> cleanup mechanism. I haven't checked when that mechanism was introduced (or
> exactly why) but it does seem to introduce complexity that is probably
> supposed to be unnecessary.
The reason is that `Style` is never destructed so doing clean up in the
destructor doesn't work. It can't be removed if this is not fixed in Qt, after
which the Debian patch won't work anymore.
Close as upstream bug for now. Feel free to request for reopening/opening a new
one if the bug persists after the Qt bug is fixed (see PR linked above) or
there's consensus from upstream how the global destruction order can be handled
in another way (possibly by **not** doing destruction, including the two bullet
points I mentioned above.)
--
You are receiving this mail because:
You are watching all bug changes.
