Bug 362897 – sys-devel/binutils-2.21 - "collect2: ld terminated with signal 11 [Segmentation fault]"

• Product: Gentoo Linux
• Status: RESOLVED OBSOLETE

[Description eu9gu4 2011-04-10 16:40:59 UTC]

Lately, an increasing number of ebuilds fail with

"collect2: ld terminated with signal 11 [Segmentation fault]"

The attached reports refer to system-config-printers-common, but all the listed packages fail with a similar error.

Reproducible: Sometimes

Steps to Reproduce:
1. emerge foo
2. emerge fails
Actual Results:  
The ebuilds listed below fail with the same error.

Expected Results:  
emerge successful

I am updating my system weekly and I run stable, with a few things unstable.
All this started to happen about 3 weeks ago.

A list of ebuilds that fail with this error below:

system-config-printers-common-1.3.2
cmake-2.8.4
poppler-0.14.5
gimp-2.6.11-r1
feh-1.12
midori-0.2.9-r1
rhythmbox-0.12.8-r1
gnomebaker-0.6.4
mail-notification-5.4-r4
seahorse-2.32.0
sensors-applet-2.2.7-r1
evolution-2.32.2
ffmpegthumbnailer-2.0.6
scribus-1.3.9 and 1.3.6

emerge --info attached
emerge -pqv attached
/var/log/portage attached
environment attached

Same thing happens regardless of using gcc-4.4.5 or 4.5.2, binutils 2.18 or 2.21.

[Comment 1 eu9gu4 2011-04-10 16:44:28 UTC]

Created attachment 269321 [details]
output of emerge -pqv

[Comment 2 eu9gu4 2011-04-10 16:45:12 UTC]

Created attachment 269325 [details]
portage log file

[Comment 3 eu9gu4 2011-04-10 16:45:39 UTC]

Created attachment 269327 [details]
environment

[Comment 4 eu9gu4 2011-04-10 16:46:21 UTC]

Created attachment 269329 [details]
output of emerge --info

[Comment 5 SpanKY 2011-04-11 00:10:53 UTC]

have you tried using memtest86+ ?

[Comment 6 eu9gu4 2011-04-11 02:05:03 UTC]

(In reply to comment #5)
> have you tried using memtest86+ ?

I have, after you asked. It passed all 8 gigs of RAM with no errors.

[Comment 7 SpanKY 2011-04-11 04:31:13 UTC]

does it fail in the same place every time ?  what if you use LC_ALL=C ?  what if you build binutils with CFLAGS=-O2 (no -march) ?

[Comment 8 eu9gu4 2011-04-11 11:04:23 UTC]

(In reply to comment #7)
> does it fail in the same place every time ?  what if you use LC_ALL=C ?  what
> if you build binutils with CFLAGS=-O2 (no -march) ?

i'll try this evening when I get back from work, but yes, with the current CFLAGS it always fails in the same place

[Comment 9 eu9gu4 2011-04-11 11:29:44 UTC]

(In reply to comment #7)
> does it fail in the same place every time ?  what if you use LC_ALL=C ?  what
> if you build binutils with CFLAGS=-O2 (no -march) ?

i actually tried that this morning: the result is the same.

maybe it's not a binutils problem, maybe it's something that has to do with those packages, but i have no idea what they might have in common.

yesteray i've built the kde 4.6.2: with the exception of system-config-printer-common, juk and kscd, everything else went just fine.

if it was binutils, it would fail much more often, imho.

[Comment 10 SpanKY 2011-04-11 18:52:30 UTC]

if you go into one of the build dirs and run `make` yourself, does it fail ?

[Comment 11 eu9gu4 2011-04-12 00:43:27 UTC]

(In reply to comment #10)
> if you go into one of the build dirs and run `make` yourself, does it fail ?

yes!

i wanted to make a new try and built glibc, binutils, udev, glib and cups using just "-O2" instead of "-march=native -O2", then system-config-printer-common. same outcome. i could try to build my whole system with "-O2", but it will probably take days and not sure if it helps.

i'm not sure about cmake and ffmpegthumbnailer, but i think that all the others link against glib.

thanks for your patience. don't know what to do anymore.

[Comment 12 SpanKY 2011-04-12 02:37:42 UTC]

rebuild just binutils with CFLAGS=-O2 and see if the printer pkg fails

please post /proc/cpuinfo as an attachment

[Comment 13 eu9gu4 2011-04-12 10:42:30 UTC]

(In reply to comment #12)
> rebuild just binutils with CFLAGS=-O2 and see if the printer pkg fails
> 
> please post /proc/cpuinfo as an attachment

my CFLAGS are now set to -O2. built binutils, got the same result.

my /proc/cpuinfo is attached.
please note that i use cpufrequtils to control the frequency of the processor with "ondemand" governor. just in case you would ask, i disabled the frequency control and unloaded the kernel module. same result.

[Comment 14 eu9gu4 2011-04-12 10:43:32 UTC]

Created attachment 269637 [details]
/proc/cpuinfo