=== doko_ [n=doko@dslb-088-073-069-075.pools.arcor-ip.net] has joined #ubuntu-toolchain | ||
=== lamont [i=lamont@nat/hp/x-9bde1aa00c800769] has joined #ubuntu-toolchain | ||
=== Keybuk [n=scott@217.205.109.249] has joined #ubuntu-toolchain | ||
=== jbailey [n=jbailey@modemcable178.77-70-69.static.videotron.ca] has joined #ubuntu-toolchain | ||
fabbione | jbehey dude | 03:10 |
---|---|---|
fabbione | whops | 03:10 |
fabbione | jbailey: dude... | 03:10 |
jbailey | fabbione: yayaya | 03:10 |
jbailey | Fine, don't remember who I am. | 03:10 |
jbailey | See if my feelings are hurt... 8'{ *cRy* | 03:10 |
fabbione | (gdb) run | 03:10 |
fabbione | Starting program: /root/a.out | 03:10 |
fabbione | we are set | 03:10 |
fabbione | Program received signal SIGUSR1, User defined signal 1. | 03:10 |
fabbione | 0xf7e9536c in nanosleep () from /lib/v9v/libc.so.6 | 03:10 |
fabbione | (gdb) next | 03:10 |
fabbione | Single stepping until exit from function nanosleep, | 03:10 |
fabbione | which has no line number information. | 03:10 |
fabbione | Program terminated with signal SIGILL, Illegal instruction. | 03:10 |
fabbione | The program no longer exists. | 03:10 |
fabbione | (gdb) That was all, folks | 03:10 |
fabbione | this is bad.. | 03:10 |
fabbione | i can't figure why we are getting a SIGILL | 03:10 |
fabbione | and only on Niagara machine | 03:10 |
jbailey | SIGILL is a kernel or processor trap. | 03:11 |
fabbione | i am afraid the last gcc updates are generating wrong code | 03:11 |
fabbione | because it's not triggered on non Niagara machines | 03:11 |
jbailey | That or the kernel is dissallowing something it should allow. | 03:11 |
fabbione | and usually that's a sign of badly generated code | 03:11 |
fabbione | unlikely | 03:11 |
jbailey | Depends on how the protection bits all work on Sparc. | 03:11 |
fabbione | i have seen a lot the other way around | 03:11 |
jbailey | Is there a test failure in either LTP or glibc's testsuite for it? | 03:12 |
fabbione | i have a reduced case | 03:12 |
fabbione | and i am trying to reduce it even more | 03:12 |
jbailey | Even better. =) | 03:12 |
fabbione | no idea about LTP or glibc | 03:12 |
fabbione | jbailey: can you get access to fluorine? | 03:12 |
jbailey | Well, you're looking at a glibc function yes? | 03:12 |
fabbione | from cr3? | 03:12 |
jbailey | I can, but don't have time for 4 to 6 hours. | 03:13 |
fabbione | i want to show you the code | 03:13 |
fabbione | feh ok | 03:13 |
jbailey | Sorry, my morning is lined up with customer appointments. | 03:13 |
jbailey | Does your reduced testsuite still call nanosleep? | 03:13 |
jbailey | Or have you extracted that out? | 03:13 |
fabbione | i am pretty sure the problem is either gcc or binutils | 03:20 |
fabbione | declaring an int foo; makes the code explode | 03:20 |
fabbione | removing that line = works | 03:20 |
fabbione | and that so should not happen | 03:20 |
fabbione | root@fluorine:~# gcc -Wall -g -O0 main.c | 03:21 |
fabbione | root@fluorine:~# ./a.out | 03:21 |
fabbione | root@fluorine:~# gcc -Wall -g -O0 main.c | 03:21 |
fabbione | main.c: In function 'main': | 03:21 |
fabbione | main.c:36: warning: unused variable 'i' | 03:21 |
fabbione | root@fluorine:~# ./a.out | 03:21 |
fabbione | Illegal instruction | 03:21 |
fabbione | doko: ping? | 03:21 |
jbailey | Right, so once you've got that, do a disassemble and a stepi to find the offending instruction. | 03:23 |
fabbione | do you recall the gdb sintax to do that? | 03:25 |
jbailey | 'dis' I think. | 03:27 |
fabbione | no that's disable.. | 03:28 |
jbailey | Hmm, lemme look it up after this call. | 03:28 |
fabbione | sure thanks | 03:28 |
=== tonyyserver [n=anthony@ubuntu/member/tonyyarusso] has joined #ubuntu-toolchain | ||
jbailey | fabbione: disass | 03:41 |
jbailey | You didn't try, did you? =) | 03:41 |
jbailey | So break before the sigill. | 03:42 |
jbailey | step to find the offending C statement. | 03:43 |
jbailey | Do the disass there | 03:43 |
jbailey | then stepi to the sigill. | 03:43 |
fabbione | jbailey: ok | 03:47 |
fabbione | dis was claiming disable.. but i might have been at the wrong point | 03:48 |
fabbione | hoooowwwowowowoowow | 04:13 |
fabbione | this is a mess | 04:13 |
doko | fabbione: pong | 04:22 |
fabbione | doko: do you have access to fluorine? | 04:22 |
doko | no | 04:23 |
fabbione | the T1000 at the london DC from -support? | 04:23 |
fabbione | i was told you had acccess | 04:23 |
doko | ohh right, didn't use it yet | 04:30 |
fabbione | doko: see your email about that niagara test case | 04:39 |
fabbione | it's possible to reproduce it on fluorine | 04:39 |
fabbione | there is a feisty chroot in /src/chroots/feisty | 04:39 |
fabbione | and the sources are in ~/root | 04:39 |
doko | fabbione: could you check with gcc-4.2? | 04:40 |
doko | what are working gcc/binutils versions? | 04:40 |
fabbione | doko: i have been staring at that code for 10 hours now.. i can't look at it anylonger | 04:41 |
fabbione | doko: i did test the latest in feisty... and no.. i didn't check back yet | 04:41 |
fabbione | i thought it was a coding error | 04:41 |
fabbione | but i just arrived at toolchain not too long ago when removing the int i; | 04:41 |
fabbione | (unused) | 04:41 |
fabbione | because i was killing code to the minimum to reduce the test case | 04:42 |
fabbione | the original parted_server is 2200 lines :) | 04:42 |
fabbione | also.. flourine has 32 threads.. it takes nothing to build gcc with make -j :) | 04:42 |
=== __keybuk [n=scott@217.205.109.249] has joined #ubuntu-toolchain |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!