[mythtv] Developing RPI2 MythTV package for ArchLinux ARM

Lawrence Rust lvr at softsystem.co.uk
Thu Jan 7 10:04:38 UTC 2016

On Tue, 2016-01-05 at 22:12 +0100, warpme wrote:
> > Wiadomość napisana przez Lawrence Rust <lvr at softsystem.co.uk> w dniu
> 04.01.2016, o godz. 16:45:
> > 
> > Apologies if I was unclear, but I meant that you should remove the
> > setuid() call on line 1799, not disable line 1789. i.e.
> > 
> > -if (setuid(getuid()) != 0)
> > +if (0 && setuid(getuid()) != 0)
> Lawrence,
> With this mod I finally was able to start FE - bun only under GDB and
> X11 SSH forwarding.
> Starting FE without GDB always gives me segfault.
> I believe GDB allows me to overcome qFatal and move forward. 
> Pls look on fe-under-remote-x11.log.zip
> I think this issue is probably related to Qt5.5.1 setuid thing and
> I’ll be happy if You can look on this issue.
> Running only under X11forwarding is probably related to EGL issues in
> my enviroment. 
> Pls look on fe-under-direct-x11.txt.zip
> I’m wonder Your opinion here.
> (Probably I have to rebuild Qt5 for OpenMAX….)


The fe-under-direct-x11.txt log shows the SEGV to be occurring from Qt
code in qt/plugins/xcbglintegrations/libqxcb-egl-integration.so.  I'm
guessing but it looks like Qt 5.5.1 and the RPi EGL libraries are not
happy neighbours.  I believe, from similar experience with Qt 5.4, that
the EGL initialisation has a race condition.  I added this patch:

8ffdd74 UI: Avoid a race condition and SEGV in EGLFS during
MythMainWindow creation

to workaround a SEGV during startup.  I believe you are suffering
similar problems.  The remote ssh version gets much further probably
because of a difference in timing during startup

It's difficult to be more explicit because the backtrace is lacking
detail due to missing debug symbols.  I would recommend installing the
Qt debug symbols together with those for libc and glib.

Once you have a filename & line number in Qt then it's a matter of
trawling the Qt code for the problem.  This is likely to need a patch to
Qt to fix.  It may be easier to revert to Qt5.4 if that's possible?

Lawrence Rust

More information about the mythtv-dev mailing list