=== Starting logfile irclogs/ubuntu-toolchain.log === ubuntulog [i=ubuntulo@ubuntu/bot/ubuntulog] has joined #ubuntu-toolchain === doko_ [n=doko@dslb-088-073-098-216.pools.arcor-ip.net] has joined #ubuntu-toolchain [10:20] score [10:20] gcc in edgy miscompiles silo [10:20] both 4.0 and 4.1 [10:21] doko_: so.. what do you need to debug this one? [10:27] does it work with -O0 or -O1? [10:35] getting there.. it takes ages to reboot the NIagara [10:35] it seems to be niagara specifc the problem [10:35] the netra boots fine [10:35] but the code is exactly the same [10:36] we are building with -Os [10:36] want me to switch to -O0 ? [10:37] Try 0, 1, and 2. [10:37] Could try 2 first, since the difference between 2 and s aren't many, and the problematic optimisations in that variance tend to be known. [10:38] gcc-4.1 -m32 -fno-stack-protector -O2 -Os -Wall -I. -I../include -fomit-frame-pointer -fno-strict-aliasing -DSMALL_RELOC=0x280000 -DLARGE_RELOC=0x380000 bin2h.c -o bin2h [10:38] which one wins here? [10:38] The last one, I think. Not positive, though. [10:40] ok.. got rid of -Os [10:40] (base)adconrad@cthulhu:~$ gcc -O2 -o hello2 hello.c [10:40] (base)adconrad@cthulhu:~$ gcc -Os -o hellos hello.c [10:40] (base)adconrad@cthulhu:~$ gcc -O2 -Os -o hello2s hello.c [10:40] -rwxr-xr-x 1 adconrad adconrad 7786 2006-09-15 18:40 hello2 [10:40] -rwxr-xr-x 1 adconrad adconrad 7754 2006-09-15 18:40 hello2s [10:40] -rwxr-xr-x 1 adconrad adconrad 7754 2006-09-15 18:40 hellos [10:41] So, the last one wins. [10:42] no.. i win.. i removed -Os :) [10:42] ;) [10:45] -O2 -> no win === doko__ [n=doko@dslb-088-073-098-216.pools.arcor-ip.net] has joined #ubuntu-toolchain [10:54] doko_: what was the last thing you read? [10:57] just my sugestion to build with -O0 and/or -O1 [11:02] -01 booting now [11:02] or better.. trying no [11:02] +w [11:03] -O1 -> no win [11:18] -O0 -> no win [11:18] so what's next? [11:21] Assume it's not miscompilation, but rather incorrect code that was relying on an old GCC bug? [11:22] do I understand it right, that the "hello" testcase is self-contained? [11:27] Hrm? hello.c was just a printf. [11:27] infinity: on both gcc-4.0 and 4.1? unlikely [11:27] infinity: and it works on one sparc but not another [11:27] note that i am building the same code as in dapper and it was working on both [11:28] so there is no code change in the middle [11:28] only the compiler [11:28] and that part is pure asm [11:28] written by davem ... [11:28] If it's pure asm, then perhaps you're barking up the wrong tree here and you have a binutils bug. [11:29] hmm that's something to try [11:29] as soon as this mofo boots up [11:45] no luck downgrading binutils === rodarvus [n=rodarvus@ubuntu/member/rodarvus] has joined #ubuntu-toolchain [12:48] so what is next guys? === rodarvus [n=rodarvus@ubuntu/member/rodarvus] has joined #ubuntu-toolchain === Keybuk [n=scott@quest.netsplit.com] has joined #ubuntu-toolchain === rodarvus [n=rodarvus@ubuntu/member/rodarvus] has joined #ubuntu-toolchain [05:05] doko_: ping? [05:13] what's up? [05:15] did you get my e-mail? [05:16] (about potential global toolchain settings for Edgy+1) [05:19] we usually prepare the toolchain before we open the archive for the next release. If you do have tested patches for new things, then you should have them prepared now. everything else should go into edgy+2 [05:19] Edgy+1 is not open yet is it? [05:20] The patches I was primarily looking at are actually in Binutils CVS now [05:21] (since some time around Jul 10 actually) [05:23] http://sourceware.org/ml/libc-alpha/2006-06/msg00095.html <-- there's the announcement and patches [05:28] we usually don't use unreleased versions. If you would like to see some feature in edgy, please provide the patches as debdiffs, tested for more than one platform [05:28] Edgy+1, not edgy. Any idea where I can find the version planned for use to check if it made it down from CVS yet? [05:32] I was hoping to discuss with you, infinity, and mdz at some point.. when's a good time? You're probably all busy as hell preparing for edgy release (next month?) [05:47] ok, it's in 2.17.50.0.3 at ftp://ftp.kernel.org/pub/linux/devel/binutils/binutils-2.17.50.0.3.tar.bz2 .... It seems it's not in a released glibc (glibc has a patch that makes it use the DT_GNU_HASH during dynamic linking) [05:47] but i don't know if Ubuntu uses the .50 releases or just the straight 2.16 2.17 etc line [05:48] it looks like it's in .50.0.1 too [05:49] either way Jelnek doesn't stick his changes in the ChangeLog because he's a brat :| [05:51] the .50 releases are unsupported versions release by H.J. Lu. Again, if you do want to see it, post the debdiff's. [05:51] Alright, yeah, I wasn't sure. [05:53] that means i have to figure out how to make a debdiff... maintainer's guide probably has that [05:53] <-- noob [05:54] I have other things to do here, I'll come back to this in a couple days. === doko_ [n=doko@dslb-088-073-098-216.pools.arcor-ip.net] has joined #ubuntu-toolchain === doko__ [n=doko@dslb-088-073-098-216.pools.arcor-ip.net] has joined #ubuntu-toolchain === doko_ [n=doko@dslb-088-073-104-229.pools.arcor-ip.net] has joined #ubuntu-toolchain