Red Hat Bugzilla – Bug 77567
gcc 3.2 segfaults during compilation of mplayer
Last modified: 2007-04-18 12:48:19 EDT
From Bugzilla Helper:
User-Agent: Opera/6.1 (Linux 2.4.19 i686; U) [en]
Description of problem:
Download the CVS snapshot of ffmpeg(www.ffmpeg.org) & compile it
Download the CVS snapshot of Mplayer(mplayer.sourceforge.net)
Compile mplayer-> error when linking with libavcodec(from ffmpeg)
in mplayer tree
./configure (Do not make clean)
At interface.c (mplayer) :
gcc -c -O4 -march=athlon-tbird -mcpu=athlon-tbird -pipe -ffast-math -fomit-frame-pointer -D_REENTRANT -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -fomit-frame-pointer -fexpensive-optimizations -fschedule-insns2 -Wall -malign-double -I. -I../loader -I./wm -I./skin -I/usr/include/gtk-1.2 -I/usr/include/glib-1.2 -I/usr/lib/glib/include -I/usr/X11R6/include -I/usr/include/gtk-1.2 -I/usr/include/glib-1.2 -I/usr/lib/glib/include -I/usr/X11R6/include -DDEBUG -o interface.o interface.c
interface.c: In function `guiGetEvent':
interface.c:682: unrecognizable insn:
(insn 3688 3427 742 (set (reg/f:SI 671)
(plus:SI (address:SI (const_int 2 [0x2]))
(const_int -20 [0xffffffec]))) -1 (nil)
interface.c:682: Internal compiler error in extract_insn, at recog.c:2148
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://bugzilla.redhat.com/bugzilla/> for instructions.
make: *** [interface.o] Error 1
make: Leaving directory `/pub/compile/MPlayer-20021108/Gui'
make: *** [Gui/libgui.a] Error 2
Version-Release number of selected component (if applicable):
Steps to Reproduce:
Expected Results: To work??! :-)
i have 2.4.19 kernel,Nvidia ge force 2 Gts+accelerated drivers.
No oops or other kernel erros were present .
The distribution is not updated /mdofied in any way
Please attach preprocessed source (interface.i; hint: add -save-temps
to the above command line).
I'll send preprocessed source if will be able to reproduce the bug
Same problem here. The strange thing is as kornos said, it is not easily
reproducible because when gcc segfaults, if you restart it, the compile process
continues without a glitch (as if gcc didn't really have a problem with that
specific source file)...
Same thing when I compile a kernel v2.5...
Motherboard problems can cause errors in GCC since GCC stresses out the
resources on your system. Are you sure your system is up to spec (no freezes,
no overclocking, etc.)? Is your memory name brand or house brand?
mplayer compile wfm on an Athlon box
$ rpm -q gcc
If it is not reproduceable, then it means you either have faulty hardware,
or you run out of memory.
I'm very aware of the effects of faulty hardware on compilation with gcc. I was "blessed" with many of these on my k6-2/350 with via mb.
My system is ok.
It;s not overclocked or screwed in any way.
I have 512mb of Ram and 256 of swap.
On my k6-2/250 if i wanted to compile the kernel i would have to type make bzImage many times before it could finish it.
Many gcc segfaults were appearing.
This is not the case now.
gcc NEVER segfaulted while compiling the kernel.
I have this system for approx a year and this kind of gcc fault
apeeared 2 or 3 times.
Please do not CLOSE this bug.
I'll try to reproduce it , but for now i'm pretty busy at work.
Probably next week i'll take another lokok at it.
You may also find you can only reproduce it if you have the nvidia binary only
3d stuff loaded. Thats worth checking just to be sure
ooh...cool..Alan Cox replied.....
First of all , let me tell you theat i'm a big fan of your
"look'n'feel" (hair,style,etc). :-)
You're totally right.
NVdriver module is a BIG pain in the ass.
It segfaults the X server , freezes the kernel,etc.
I think it's the only module that freezes the kernel with no oops
or other message.
Until now i've used the version 3123 ,which was quite acceptable.
however the last version 4194 slows X very much(especially gnome-terminal with a jpg as background).
However i couldn't imagine that it could affect gcc.
Now i understand why the compilation in X have so many problems
compared to the text terminal compilation.
I have the same pblm on 2 differents machine.
I try to compile the RH 8.0 kernel provided on a AMDk6-2/350 with 128 Mb RAM
(128 SWAP) not overclocked
and under a ATHLON 1700 XP with 256 Mb of ram (and 256 SWAP)not overclocked
(buy to compaq)
under the K6 gcc segfault a lot 'd of time and it's not reproductible !
same with the athlon.
I try to remove the -O2 option, and it made "less" segfaults...
The compaq have a nvidia drivers not loaded (but nvidia hw) because I prefered
make the compilation myself.
I'm ready to help (ie make compilation with bug report) on both machine.
I'll try to do a silly thing : until make bzImage; do date; done
on my 350. the kernel was exploded after 5 minutes.
hope this help.
So I tried the following :
I downgrade to my old RH 7.2 with gcc-2.95-98
and I dowload the gnu version of gcc-3.2.1 (the latest version on gnu.org)
I make it on my 350 (without any pblm) and try to make some compilation (bochs
to be exact)
No Segfault where made.
I add that with the RH gcc under the 8.0 I have tried to make windowmaker, with
a lot's of segfault with a nicely message saying that make a report on
Hope this help !
PS : I tried to compile gcc-3.2.1 under my athlon xp when my girls friends
leave me it :)
As Alan Cox suggested in another bug,
I'have test the memory of my 2 computers with memtest86
without any errors.
So it does'nt seems to be a memory error ...
Hope this help.
(I Haven't enougth time to make more test under the Athlon XP)
Preprocessed file never got attached, and the entire comment log points
to buggy kernel module.