JunitTest_bean_complex - [qt5,kf5] failed

Jan-Marek Glogowski glogow at fbihome.de
Thu Sep 12 21:52:22 UTC 2019


Am 12.09.19 um 22:55 schrieb gmail Vladimir Koković:
> Following the suggestion to comment line 44 in vcl/qt5/Qt5Clipboard.cxx, now
> there is something new that is unsolvable to me.

There was some additional problematic output even before that Qt5Clipboards
assertion: "QObject::setParent: Cannot set parent, new parent is in a different
thread"

So LO is doing something I didn't expect when I added the assert, which
definitely conflicts with the current assumptions of the implementation. I'll
see, if I find some time to look into this, but probably not before the end of
next week. But it looks very much that some additional stuff should be run in
the main thread. If you want to debug and fix that, look for RunInMainThread
calls sites in the qt5 code as an example. And think about opening a bug report
in Bugzilla.

> JunitTest_bean_complex logs are in logs.tar.gz, in the attachment file.

office.bin:
/mnt/sdd1/home/src/libreoffice-vk/libreoffice/vcl/qt5/Qt5Frame.cxx:431: void
Qt5Frame::SetDefaultSize(): Assertion `!m_bDefaultSize' failed.

After reading the code, that assertion is quite probably a wrong. Interesting,
that this didn't crash the video overlay, when we developed it, but maybe I
added the assertion later, when I fixed the VCL geometry handling inside the qt5
plugin.

And then there is your previous mail:

> Thanks for the tip, with SAL_USE_VCLPLUGIN=gtk3 DocumentLoader passes!
>
> But, JunitTest_bean_complex does not pass with any plugin, as can be seen
> in logs.tar.gz, in the attachment file!

If you don't want to fix the qt5 stuff, there seems to be plenty of other stuff
broken in the formerly disabled unit test. I'm actually not sure this could ever
run without native windows, if the test relies on overlay plugin windows from
Java. I actually didn't look into it, so that is just a guess.

HTH

Jan-Marek


More information about the LibreOffice mailing list