[Pkg-fglrx-devel] Bug#777675: qtcreator: hang and repeated SIGSEGV on exit

[Martin Kourim]

Hi,

thanks for your answer. The crash is in OpenGL driver but it doesn't 
necessarily mean that the bug is in the driver. The crash appears on 
program exit so it's possible that there is a memory corruption that 
manifests during cleanup. I don't observe such issue with any other 
Qt/QML/OpenGL application.

Second issue is that the programs hangs and generates CPU load on exit 
because it executes the SIGSEGV signal handler over and over again (it's 
obvious from strace output attached to original bug report).
When the signal handler returns, the code will continue at the point the 
signal occurred, re-executing the same instruction.  In this case, what 
is the reason for returning from SIGSEGV signal handler? Is this desired 
behavior?

Thanks,
Martin

On Thu, 12 Feb 2015 09:33:55 -0600 Adam Majer <adamm at zombino.com> wrote:
 > On Thu, Feb 12, 2015 at 10:11:28AM +0100, Martin Kourim wrote:
 > > So, does that mean that it doesn't matter if something doesn't work 
with
 > > proprietary driver?
 > >
 > > I can't use radeon driver due to it's instability on this 
particular card. I
 > > have no other choice than using this proprietary driver and I 
believe I'm
 > > not the only one in such situation.
 > > I see no reason why qtcreator should not work with this proprietary 
driver
 > > while everything else does work. I.e. it looks like a bug in
 > > qtcreator/Qt/QML, not the driver.
 >
 > The crash is in the OpenGL driver, not QtCreator. You can see that in
 > the backtrace from the debugger. libGL.so.1 is your driver.
 >
 >
 > > I already tested this on other machine with open source driver. I can't
 > > reproduce it there. Yes, it's graphics-driver-specific. What's the 
point of
 > > trying to reproduce it on radeon driver when I can reproduce it in 
100% of
 > > cases on fglrx-driver?
 > >
 > > I can try to find out if it's issue with qtcreator or with Qt/QML. Some
 > > pointers what to try would be helpful, I'm not familiar with Qt and 
QML.
 >
 > One workaround is to use software based OpenGL driver, instead of
 > hardware accelerated OpenGL.
 >
 > libgl1-mesa-swx11
 >
 > Other than this, we can't duplicate the problem or even fix it because
 > software responsible for the crash (your opengl driver) is non-free
 > and thus not possible to fix by anyone except by AMD. You should
 > report this bug on their bug reporting place (forums?).
 >
 > - Adam
 >
 > --
 > Adam Majer
 > adamm at zombino.com
 >
 >