Debugging spring GDB + Windows

[velteyn]

Hi all,
attempting to debug spring using gdb 7.5 + eclipse on windows here...

I have this problem: spring starts correctly and after Test Game (Empty mod + duck, null AI) the program finishes the init than exits !!!

If I launch spring.exe normally (no debug) anything works fine....

Can anyone explain to me this strange result ? :)

Thank you

GDB output

Code: Select all

[f=0000000] [Sound] Error: Unable to open audio file: sounds/multiselect.wav
[f=0000000] [Sound] Error: Unable to open audio file: FailedCommand
[f=0000000] [Sound] Error: CSound::GetSoundId: could not find sound: FailedCommand
[f=0000000] [unit_script.lua] Error: Loading gadget: Lua unit script framework  <unit_script.lua>
Quit (expect signal SIGINT when the program is resumed)

note: I had to exclude SIGSEGV from gdb init because I obtained many SIGSEGV stops , but I think that the code handles them !

[jK]

note: I had to exclude SIGSEGV from gdb init because I obtained many SIGSEGV stops , but I think that the code handles them !

no, when gdb is running you want that gdb handles _all_ errors

[velteyn]

Can this error:

Error: Loading gadget: Lua unit script framework <unit_script.lua>

make the program to exit ?

[gajop]

Most likely no.
Gadget/widget errors rarely cause termination of Spring itself.

[velteyn]

I that case I think that is a Windows GDB 7.5 problem... I'll try to get gdb 7.6 and see if something changes. Its strange , program exits unexpectedly and with no error !!! When I run the program alone , anything works fine !!

[velteyn]

QUESTION: Is there any angel here available to put up a remote session on my PC to try to help me and see what's wrong ? I sincerely want to start to learn the engine and give some contribution.. but I'm stuck on initial setup...

[velteyn]

I have another clue :
aftter finalizing STEP I have

614,500 =thread-group-exited,id="i1",exit-code="01777777777770000000005"
614,515 *stopped,reason="signal-received",signal-name="SIGSEGV",signal-meaning="Segmentation fault",\
reason="signal-received",signal-name="SIGSEGV",signal-meaning="Segmentation fault",reason="exited",e\
xit-code="030000000005"
614,515 (gdb)

[velteyn]

cause of segmentation fault is this :


0x636050d8 in omp_in_final() from c:\MinGW32\bin\libgomp-1.dll

Process exited with code 030000000005

I think I can stop here because I'm not able go further..

[abma]

the whole infolog.txt of such a crash would be really usefull...

[velteyn]

you can find it in as file attach...

I doubt it can be of some use....

I'm starting to study libgomp .. I think the problem is here !!

[abma]

[f=0000000] Warning: [Watchdog] Hang detection triggered for Spring 94.1 (OMP).

-> increase HangTimeout or disable watchdog by setting it to -1

not sure if it helps, but at least it will reduce possible causes for a crash with gdb attached to the process.

[velteyn]

Watchdog disabled

the result is the same as above,
I attach also a screenshot of gdb session with the SIGSEGV.

It only happens in gdb debug...

[abma]

"Continuing" <- that looks wrong for me, imo the cause of the crash is before omp_in_final

whats the output of "bt full" before "continue" is called?

don't skip SIGSEGV... they are there for reason.

edit: you should investigate the very FIRST SIGSEV, any crash/problem after it is likely a side-effect of it.

[velteyn]

here it is

Its obscure to me ....

[velteyn]

Into the file attach you can find the very first SIGSEGV but is absolutely BEFORE spring starts to log something !!

After i skip it the program works normally until I start a map than hangs after the Finalizing step !

Cannot understand.... I start to think that Windows is not the right environment for a spring developer !!

[abma]

"dll-hell"...

which version of mingw did you install? whats the output of gcc -v? are you using current mingwlibs?

(and yes, its much more harder to setup a build env on windows than on linux...)

[velteyn]

Im using, gcc 4.7.2 , but I had a try also with the official mingw 4.7.0 as described on the windows build instructions. The result was the same !

mingvlibs are the latest available!!

[abma]

mingw 4.7.0 as described on the windows build instructions

an easy workaround could be, to remove omp from your mingw-install. imo your mingw installation is broken. i suggest to uninstall completely mingw and reinstall it without installing two instances of gcc into it.

please only use "MingW Stable", any thing else is known to fail. there is a reason why there is a screenshot for that...

that could be the cause of your problem, afaik omp is bound to the compiler used, if you use a different version of omp, then its very likely that it breaks.

[velteyn]

ah ok, but it would be the fifth try.. I always completely removed C:\MINGW32 dir erasing all the content before re-installing...

Now I search how to remove content from mingw install..

I'll try to follow this road !!

[abma]

i'm currently compiling spring on win32 on my own, but compiling on windows is horrible slow, so will take a bit...