DBus thread shutdown issue on Windows

[David Faure]

[Adjusting CC list]

On mercredi 10 octobre 2018 11:55:01 CEST Ben Cooksley wrote:
> On Thu, Oct 4, 2018 at 3:15 PM Thiago Macieira
> 
> <thiago.macieira at intel.com> wrote:
> > On Wednesday, 3 October 2018 18:12:08 PDT David Faure wrote:
> > > On another note, if Qt knows this part of Windows is broken, why is it
> > > even trying to spin down a thread when they know it's just going to
> > > shoot themselves in the foot?
> > 
> > Because C++. If the object exists, it will be destroyed properly.
> > 
> > The code also predates my finding of the issue on Windows. Nothing I've
> > tried so far has managed to fix the issue. I don't remember if simply
> > leaking the QThread objects works. I resigned my maintainership over my
> > inability to find da solution, so someone else would step up and come up
> > with new ideas.
> > 
> > It's been two and a half years and we're still waiting.
> 
> I see. As it seems only some parts of DBus trigger this (because
> KDevelop and KNotification tests both cause DBus to start and they
> don't get stuck like this) could we try disabling kssld on Windows?
> (to my knowledge it's only there to help with certificate stores, and
> those should be looked after by Windows infrastructure not by us on
> Windows systems...)

I believe it also stores "accept this certificate for this session",
which makes sense on Windows as well....

Though not in unittests, I guess, so disabling it use on CI sounds like a 
possibility. Not sure exactly how to do that though (no time to investigate 
right now).

-- 
David Faure, faure at kde.org, http://www.davidfaure.fr
Working on KDE Frameworks 5