Museeq gets segmentation fault on startup

[Silver <silver.salonen@…>]

I see it trying to connect to Museek, but then it gets segmentation fault. I'll add my backtrace.

[Silver <silver.salonen@…>]

When I killed my existing museekd process, I could run Museeq again.

[lbponey]

Don't understand why you got a segfault... maybe a qt bug? We'll see this later.

[Silver <silver.salonen@…>]

Are you sure you want to see this later? I'm still getting the segfault often :(

[lbponey]

Ok, let's try to fix it. Can you try r1095? I'm not sure it will fix the bug but anyway, it's safer like this.

[Silver <silver.salonen@…>]

It didn't crash now, but actually it didn't crash every time previously either. It crashed only sometimes.. I'll let you know if I see it again.

[Silver <silver.salonen@…>]

I think I've played enough with the feature now.. to my mind the sure thing to crash Museeq before the change was to have "ShutDown? Daemon on exit" unchecked and then logout from KDE. Then, after logging in again, Museeq couldn't be started before shutting down running museekd first. Now it survived this :)

[lbponey]

Great! So I'm closing this. Add a comment if you still have this problem later. Thanks

[Silver <silver.salonen@…>]

Heh, celebrated too early. I had it again.. I'll attach backtrace.

[lbponey]

ok. I'll have a look at it.

[lbponey]

Can you try replacing museeq/userlistview.cpp by the attached file? I'd like to see the last few lines of museeq output when it crashes.

[Silver <silver.salonen@…>]

OK, got it now.. the last few lines in gdb:

"Zumbi" 0
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"Zyklonlad" 1
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0
"[RKM]" 0

Program received signal SIGSEGV, Segmentation fault.

Do you want backtrace too?

[lbponey]

Ok. And [RKM] is in one of your user list (buddy/banned/ignored/...)? You can attach a backtrace too if it's different from the previous one.

[Silver <silver.salonen@…>]

Nope, I don't know anything about [RKM] - maybe he's uploading from me or smth.. because I have a long queue there.
New output (as can be seen from there, I can't get backtrace for some reason):

"Wischma Abda" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzo" 2
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0
"Wizzy_1967" 0

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x2a401100 (LWP 100420)]
0x292d0690 in open () from /lib/libc.so.7
(gdb) backtrace
#0  0x292d0690 in open () from /lib/libc.so.7
Error accessing memory address 0xbf9ffec0: Bad address.

[lbponey]

Can you test with test2 now? I need the same thing: the last few lines of output + backtrace if possible. thanks

[Silver <silver.salonen@…>]

OK, I'll need to wait for this to happen again now :)

[Silver <silver.salonen@…>]

Finally I got it again :)

[lbponey]

I couldn't find anything very helpful in the log... I will look at it later. Any new backtrace would be helpful I think. Anyway, 0.2 is now out so the target is now 0.3!

[belegdol]

Silver, you could improve your backtrace by using a build with enabled debug symbols. If you are building from sources, you need to add -g to CFLAGS/CXXFLAGS, I'm not sure if museek+'s build system has a easy-to-use switch to do that (I'm using the RPMs I maintain).

[Silver]

OK, I'll try. I'm using FreeBSD port that I maintain and I changed this in the port's Makefile:
CONFIGURE_ENV+= CPPFLAGS="-g ${CPPFLAGS}" LDFLAGS="${LDFLAGS}"

Does it seem OK?

[belegdol]

I think it's ok. Also make sure you are not stripping the binaries. In case of rpm they are stripped, but the symbols are then stored in -debuginfo rpm, which can be later installed and make gdb produce meaningful backtraces. I have no idea how it works in FreeBSD world, sorry.
I *think* that if you do this correctly, there will be no "no debugging symbols found" warnings in gdb when you attach it to museeq. Hope that helps.

[lbponey]

Just a little clarification: I think the proper way to have the -g flag is to use this cmake option:

-DCMAKE_BUILD_TYPE=Debug

(see: http://www.museek-plus.org/wiki/MuseekFromSources#Availableoptions)

[Silver <silver.salonen@…>]

Yes, the option is in use in FreeBSD port. I'll forget about -g then :)

[Silver <silver.salonen@…>]

Any progress on that?

[lbponey]

I'm sorry but I don't have the time to work on museek+ at the moment. The last backtrace has the same problem as the fifth one: there is a problem with debug symbols... Don't know how to make it more informative... Any help would be welcomed.

[Silver <silver.salonen@…>]

Well, that's weird, because I did set -DCMAKE_BUILD_TYPE=Debug, and my binary is not stripped either.

[belegdol]

Try installing the equivalents of debuginfo packages for qt and glibc, it should help a bit.

[Silver <silver.salonen@…>]

Yeah, well.. FreeBSD doesn't have those packages. I'll have to reinstall some QT4 packages with DEBUG enabled or smth - I guess I don't have to enable debugging for all of them, do I? Would qt4-gui suffice?

[lbponey]

I think there are not enough infos to fix it in 0.3...

[Keischa]

That's not just the best answer. It's the bsetest answer!