FS#18553 - [gkrellm] 2.3.4-1 gives "illegal instruction"

Attached to Project: Community Packages
Opened by Jack (ostroffjh) - Thursday, 04 March 2010, 00:58 GMT
Last edited by Gavin Bisesi (Daenyth) - Saturday, 20 March 2010, 21:06 GMT
Task Type Bug Report
Category Packages
Status Closed
Assigned To Gavin Bisesi (Daenyth)
Architecture i686
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

gkrellm 2.3.4-1 gives "illegal instruction" and quits before any window opens. I"m on i686 (AMD Athlon XP)

Steps to reproduce:
install and try to run.

I downloaded 2.3.4 from upstream and compiled, and it runs fine.

strace shows it fails with
--- SIGILL (Illegal instruction) @ 0 (0) ---
right after
writev(3, [{"\216\5\7\0(\0@\3\16\1\0\0\0\6@\0 \0\0\0'\0@\3\0\0\0\0\227\4\5\0"..., 16360}, {NULL, 0}, {"", 0}], 3) = 16360
read(3, 0x87371a0, 4096) = -1 EAGAIN (Resource temporarily unavailable)
on a sockekt opened with
socket(PF_FILE, SOCK_STREAM, 0) = 3
connect(3, {sa_family=AF_FILE, path=@"/tmp/.X11-unix/X0"}, 20) = 0
getpeername(3, {sa_family=AF_FILE, path=@"/tmp/.X11-unix/X0"...}, [20]) = 0

There is a forum thread on this:
http://bbs.archlinux.org/viewtopic.php?pid=713216

Is there anything else I can provide?
This task depends upon

Closed by  Gavin Bisesi (Daenyth)
Saturday, 20 March 2010, 21:06 GMT
Reason for closing:  Fixed
Additional comments about closing:  Rebuilt as -2
Comment by Gavin Bisesi (Daenyth) - Thursday, 04 March 2010, 02:24 GMT
The same version that you have runs fine here on i686. Can you try rebuilding from abs with options=('!strip') and getting a backtrace if the issue happens again? I have a suspicion that the bug is actually in something that gkrellm is linked against, not the software itself.. I'm not sure.
Comment by Gavin Bisesi (Daenyth) - Thursday, 04 March 2010, 02:27 GMT
I wonder if this is related to  FS#18514  actually -- Can you attach /proc/cpuinfo?
Comment by Jack (ostroffjh) - Thursday, 04 March 2010, 03:42 GMT
I'll be glad to rebuild, but the problem hasn't happened with my build from the upstream download, and happens always with my pacman version. However, I'm quite willing to believe the problem is in a build or link dep. If building from abs means from the upstream download, I don't think it will help. If it's a variant of the packaged version - I'll be glad to try - just point me to some instructions. A problem with i686 vs amc_64 binaries is certainly possible, but "file /usr/bin/gkrellm" shows a 32 bit binary (as opposed to my other machine (gentoo, which shows a 64 bit binary)).

/proc/cpuinfo: (being careful to get it from the arch machine, which is not where I'm posting this.....)

processor : 0
vendor_id : AuthenticAMD
cpu family : 6
model : 10
model name : AMD Athlon(TM) XP
stepping : 0
cpu MHz : 1150.086
cache size : 512 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow up
bogomips : 2301.01
clflush size : 32
cache_alignment : 32
address sizes : 34 bits physical, 32 bits virtual
power management: ts
Comment by Gavin Bisesi (Daenyth) - Thursday, 04 March 2010, 03:50 GMT
Hm, ok.

To build from abs, pacman -S abs, abs community/gkrellm, then cp /var/abs/community/gkrellm/ elsewhere and makepkg -i
Comment by Jack (ostroffjh) - Thursday, 04 March 2010, 14:45 GMT
Built from abs, it works fine. (I didn't use the !strip, but it doesn't fail....) The only error I noticed during compile was "No package 'libntlm' found. Package libntlm was not found in the pkg-config search path."

As far as I can tell (looking at my Gentoo box) libntlm is part of Cyrus SASL, which is installed, but it does not seem to have any libntlm. Don't know if this is a problem or just a red herring (irrelevant).
Comment by Gavin Bisesi (Daenyth) - Thursday, 04 March 2010, 15:03 GMT
I really don't know what's wrong. I guess I can upload a recompiled version, but it does work for me. Perhaps the previous builder had some strange makeflags set.
Comment by Allan McRae (Allan) - Sunday, 07 March 2010, 13:32 GMT
It looks like everyone having issues has an AMD processor. Perhaps some processor specific optimisation slipped in.
Comment by Gavin Bisesi (Daenyth) - Sunday, 07 March 2010, 13:45 GMT
I have an AMD also, and it works. I can try uploading a rebuild of it, but I'd like too know why this is breaking before I do anything..
Comment by Allan McRae (Allan) - Sunday, 07 March 2010, 14:16 GMT
I just did a local rebuild and can not see any processor specific optimisation being applied. I'd just rebuild and chalk it up to ghosts in the machine...
Comment by Gavin Bisesi (Daenyth) - Sunday, 07 March 2010, 14:24 GMT
I wonder if the previous packager (dragonlord) had some strange makeflags set

Loading...