[CRASH] segfault in ImageLoadThread::run

[brian@…]

Using 0.24-fixes from today, I am getting a pretty consistent segfault trying to start viewing a particular program.

The stack trace of the active thread in gdb when it segfaults:

#0  0x0a5e2e60 in ?? ()
#1  0x0053f0b3 in ImageLoadThread::run (this=0xb3cf790) at mythuiimage.cpp:104
#2  0x01a6625c in QThreadPoolThread::run (this=0x9bffa58) at concurrent/qthreadpool.cpp:106
#3  0x01a7132e in QThreadPrivate::start (arg=0x9bffa58) at thread/qthread_unix.cpp:248
#4  0x018e596e in start_thread (arg=0xa9bdfb70) at pthread_create.c:300
#5  0x01e72a4e in clone () at ../sysdeps/unix/sysv/linux/i386/clone.S:130

I will attach the full threaded backtrace.

[brian@…]

threaded back trace of crash

[Github]

Initialize the MythUIImage filename correctly

Fixes #9634. Fixes #9708. Refs #9547.

If the filename does not start with "/", "​http://", "​https://" or "​ftp://", then the actual filename loaded is uninitialized. It can be any random crap on the stack. I'm honestly surprised this worked at all :)

This may fix things for #9547 as well, and should be tested with that scenario too.

Branch: master Changeset: 2c7735e7f80b5d2737d2bc8c4e9144f3e42da2e7

[beirdo]

It's worth trying the re-fixed code, but the above commit got reverted.

[beirdo]

Is this still happening in the particular test case that was cited in the ticket?

[stuartm]

See also #9388

[beirdo]

See also #9547

[beirdo]

I'm removing the milestone as it is not clear that this will be fixed with no input from affected users. When we get more input, it can be set at that time.

[beirdo]

We need either a test case that allows us to reproduce this, or to have someone who is affected to work with us. This is a serious bug, but we are not being able to move forward on it. To the original poster: Is this still happening? Have you tried master (0.25-pre)?

[brian@…]

I have not seen it in a while. I am also on master now though.