jbailey | Bah. | 12:01 |
---|---|---|
jbailey | I just retested locally, and it doesn't fail there. | 12:02 |
=== jbailey sets up a minimal chroot. | ||
lamont__ | jbailey: minimal chroot? (maybe missing build-deps?) | 12:33 |
jbailey | Right, I usually test in a chroot I use for development. | 12:34 |
jbailey | I don't see how it built before if it was missing a build-dep, but I've slimmed down the chroot for this build. | 12:35 |
jbailey | I think I need to hack glibc (and maybe cdbs) so that if a build fails at configure time that config.log is spit to stdout. | 12:36 |
jbailey | Mm, missing build-dep on the bit the provides libcgcc_s_64 | 12:45 |
jbailey | erm | 12:46 |
jbailey | or not. | 12:46 |
jbailey | That's provided by gcc-3.4 supposedly. | 12:46 |
jbailey | doko: Still around? | 01:15 |
doko | yep | 01:28 |
jbailey | doko: I figured it out. collect2 was acting strange, but I hadn't realised that gcc would change its arguments to collect2 based on what was available. | 01:29 |
jbailey | lamont__: In Debian (and in Ubuntu if you know...) is libc6-dev-sparc64 usually already in the base chroots / is it part of build essential? | 01:29 |
jbailey | lamont__: I'd like to have ppc64 be consistant with current practice for sparc/s390 | 01:30 |
lamont__ | jbailey: nfc | 01:30 |
lamont__ | for it to be there (properly), it needs to either be essential, or a Depend: of build-essential | 01:30 |
doko | only for sparc, not s390 | 01:33 |
BenM | Hey doko, tseng sent me to you | 01:34 |
doko | BenM: i | 01:35 |
doko | BenM: hi | 01:35 |
BenM | mono is failing its regression test on gcc4 | 01:35 |
BenM | and tseng thought you might be able to help out | 01:36 |
doko | jbailey, lamont__: b-e lists the 64bit dev only for sparc, not s390 | 01:36 |
doko | BenM: I do? hmm. all of them? | 01:36 |
BenM | well, one specific test case gets a segfault | 01:37 |
BenM | we are pretty sure something is getting miscompiled | 01:37 |
BenM | http://primates.ximian.com/~bmaurer/mono-1.1.8.tar.gz | 01:38 |
BenM | its reproducable on that | 01:38 |
BenM | you just have to configure; make; make check | 01:38 |
doko | what does happen, if you reduce the optimization level? any warnings? | 01:38 |
BenM | it happens both on -O2 and -O0 | 01:39 |
BenM | no warnings for the file other than pointer signedness | 01:39 |
doko | ahh, you know the file. does gcc-3.4 work? | 01:39 |
BenM | yes | 01:41 |
BenM | well, by that i mean "the file where the backtrace says the segfault happens" | 01:41 |
BenM | there isn't anything suspcious in the entire build log though | 01:42 |
jbailey | lamont__: Any objection to it being added to b-e? | 01:44 |
doko | so, if you compile the testcase with 3.4, the testcase works? or a file from the distribution? | 01:50 |
lamont__ | jbailey: that gets back to the question of whether or not it is build-essential... | 01:51 |
BenM | well, the test case is c# :-) | 01:51 |
lamont__ | jbailey: for which, I'd recommend a discussion with the build-essential maintainer | 01:51 |
lamont__ | == keybuk | 01:51 |
BenM | the issue is, if I compile mono using gcc4, it fails | 01:52 |
BenM | if i compile mono with gcc!=4, it works fine | 01:52 |
=== BenM is now known as bhome | ||
jbailey | lamont: I couldn't remember your email address, so I didn't cc: you, but I've emailed scott. If it turns out that it gets added to b-e, we just need to respin the build, otherwise I'll need to add the build-dep | 02:47 |
jbailey | lamont: Did you keep your ubuntu.com address? I remember that sabdfl was talking about them being available to members, so thought you might be testcase. =) | 02:48 |
lamont | jbailey: yes, still have ubuntu.com | 03:03 |
jbailey | Luvly | 03:03 |
=== BenM [~benm@c-24-128-49-153.hsd1.ma.comcast.net] has joined #ubuntu-toolchain | ||
BenM | doko, gotten a chance to try that tarball yet | 03:08 |
BenM | ? | 03:08 |
=== BenM is now known as bcook | ||
=== bcook is now known as BenM | ||
=== lamont__ [~lamont@rover3.mmjgroup.com] has joined #ubuntu-toolchain | ||
fabbione | morning | 06:18 |
fabbione | lamont: ping? | 06:26 |
lamont | ack | 06:26 |
fabbione | lamont: we will need to revert the gcc-3.4 b-d stuff... | 06:27 |
lamont | fabbione: having difficulty verifying that ipsec traffic is passing back through INPUT once unencapsulated. | 06:27 |
lamont | iz annoying | 06:27 |
lamont | fabbione: why? | 06:27 |
fabbione | unfortunatly we need the latest gcc and latesr k-p to build | 06:27 |
lamont | on all architectures? | 06:27 |
fabbione | 11753 | 06:27 |
fabbione | no only i386 | 06:27 |
lamont | because the relaxed check made it so that I could build... :-) | 06:27 |
fabbione | yeah i know | 06:29 |
lamont | ah, so make the build-dep be powerpc i386 | 06:29 |
fabbione | yeah that too | 06:30 |
fabbione | kernel-package (>= 8.135ubuntu4) [powerpc] , kernel-package (>= 8.132ubuntu2) [!powerpc] ??? | 06:32 |
fabbione | kernel-package is arch all | 06:32 |
fabbione | but we will need to upgrade that too :/ | 06:33 |
fabbione | gcc-3.4 (>= 3.4.4-0ubuntu6) [powerpc i386] , gcc-3.4 [!powerpc !i386] | 06:34 |
fabbione | does it look sane? | 06:34 |
fabbione | i can never remeber how to add more than arch to [] | 06:34 |
lamont | I believe that's it | 06:35 |
lamont | (comma is right out..) | 06:35 |
fabbione | yeah i did check with other packages too :) | 06:37 |
fabbione | lamont: what did you change in the configs? | 06:39 |
lamont | fabbione: I added the missing lines. | 06:41 |
lamont | R2[45] 00, SQUASHFS, HOSTAP | 06:41 |
fabbione | ah ok | 06:41 |
fabbione | it was just an update | 06:41 |
lamont | well, it was a 'fix hppa configs since they didn't get updated with the addition of the drivers'... but I shortened that... | 06:42 |
fabbione | humm | 06:42 |
fabbione | sorry.. i might have done that in a different way recently | 06:42 |
=== BenM is now known as bsleep | ||
=== chmj [~d3vic3@dumbledore.hbd.com] has joined #ubuntu-toolchain | ||
chmj | jbailey, thanks for the patch | 10:02 |
jbailey | chmj: Glad to show it to you. =) IT's just really nice to have all that information for tracking. | 11:41 |
jbailey | chmj: We found that we were starting to lose track of where things things came from and why we did them. =) | 11:41 |
chmj | oh ok, that format sure has enough info | 11:43 |
jbailey | doko: I thought sparc killed the wrapper a couple years ago on the grounds that it was unpredictable and that it sucked? | 11:45 |
jbailey | lamont, fabbione: ping re: builds of glibc on your archs. =) | 11:46 |
fabbione | jbailey: i am still building gcc-* | 11:46 |
fabbione | jbailey: once it's done i will do glibc | 11:46 |
jbailey | fabbione: Cool. | 11:47 |
fabbione | it's running the test-suite right now | 11:47 |
fabbione | it shouldn't take too long | 11:47 |
jbailey | Any idea how the hack is going to get the build-logs onto people so I can just see them myself? | 11:47 |
doko | jbailey: no, the thread on d-sparc | 11:47 |
doko | jbailey: so you upload a glibc with fixed b-d's? | 11:48 |
jbailey | doko: Ah, a'ight. | 11:48 |
fabbione | jbailey: i didn't manage to talk with elmo in a few days | 11:48 |
fabbione | jbailey: i will need to check with him | 11:48 |
jbailey | doko: I haven't. If we're adding it to build-essential, it just needs a give-back. | 11:48 |
jbailey | doko: (I've also been awake for about 10 minutes) | 11:49 |
doko | jbailey: as you like it, but for packages like zlib1g and others, we should explicitely add it. | 11:52 |
jbailey | Why? | 11:52 |
doko | push back a package to Debian and it will fail | 11:52 |
doko | it's not a big job to add that b-d for these few packages | 11:53 |
jbailey | True. I just hate b-d'ing on b-e packages. | 11:53 |
doko | so we should add amd64-libs-dev to b-e as well? | 11:55 |
jbailey | That's the question. =) | 11:55 |
jbailey | Or should sparc be made like the others and have sparc64 dropped from b-e? | 11:56 |
doko | no, then we have to change the gcc wrapper as well ... | 11:56 |
jbailey | We should be treating biarch configs consistently. | 11:57 |
doko | but what about an update for amd64-libs to match our current glibc before we go multiarch? | 11:57 |
doko | Mithrandir: ^^^ | 11:57 |
jbailey | But I tend to fall on the side of thinking that the sparc-gcc wrapper is teh suck. | 11:58 |
jbailey | right! That's what I was going to ask you yesterday! =) | 11:58 |
jbailey | To just clarify the include directories so I could do the glibc work for i386/amd64 and amd64/i386 biarch | 11:58 |
doko | jbailey: I don't like it either, but BenC defends it | 12:01 |
jbailey | Is benc the only one who defends it? At this point I'd take joshk's word over ben's for the sparc port. | 12:01 |
fabbione | jbailey: did you complete the kernel build with a ppc32 kernel? | 12:02 |
doko | I would have to look | 12:02 |
jbailey | fabbione: BAh, that's what I was supposed to do last night, sorry. No, lemme reboot and start it now. | 12:02 |
fabbione | eheh ok :) thanks | 12:02 |
fabbione | there is no rush really | 12:02 |
fabbione | i am not going to upload anytime soon | 12:02 |
fabbione | i need to go and prepare food | 12:02 |
fabbione | bbl | 12:02 |
jbailey | doko: I'm just updating the status of my stuff for the wiki. Want me to update ToolChainRoadmap? I was thinking "C++ transition still in progress, ppc64 uploaded" | 12:06 |
doko | sure | 12:06 |
jbailey | Mm, and java bits in main? | 12:07 |
jbailey | and C++ done for main | 12:08 |
=== jbailey adds b-d for now to ppc | ||
doko | still waiting for the OOo2 build ... | 12:21 |
jbailey | glibc with added build-dep uploaded | 12:23 |
=== doko_ [~doko___@dsl-084-059-033-037.arcor-ip.net] has joined #ubuntu-toolchain | ||
infinity | doko : Which OOo2 build are you waiting on? | 12:52 |
infinity | doko : It was successful on i386/ppc, failed on amd64/ia64. | 12:52 |
doko_ | infinity: oh, nice, I didn't see them succeed, you know, the upload was done some months ago ;) | 01:21 |
lamont | ln -sf /usr/include debian/libc6-dev/usr/hppa64-linux-gnu/include | 02:39 |
lamont | ln: creating symbolic link `debian/libc6-dev/usr/hppa64-linux-gnu/include' to `/usr/include': No such file or directory | 02:39 |
lamont | jbailey: that's the error I get from -5... | 02:40 |
lamont | seems kinda familiar.. :-) | 02:40 |
lamont | -6 is downloading now, build will happen sometime today, I expect. | 02:41 |
Mithrandir | doko: amd64-libs is utterly uninteresting to me and jbailey have some things up his sleeve to get rid of it, afaik. | 03:20 |
doko | Mithrandir: ok, would it be possible to keep building gcc-4.0 biarch in unstable ? would be a good thing to point people to it as a test for 64bit compatibility | 03:22 |
Mithrandir | doko: yes, I think that would be a good thing. | 03:23 |
doko | Mithrandir: asking, because it's NOT possible to build gcc-4.0 biarch in breezy at the moment | 03:25 |
doko | I think we face the same problem in unstable once glibc get's updated? | 03:25 |
jbailey | lamont: Don't bother with -6 then. | 03:51 |
Mithrandir | doko: why isn't that possible? | 03:56 |
doko | Mithrandir: build failure, just try to build the package with an installed amd64-libs-dev. | 03:58 |
doko | IIRC, even if biarch is disabled | 03:58 |
=== lamont__ [~lamont@15.238.5.160] has joined #ubuntu-toolchain | ||
jbailey | doko: ln: creating symbolic link `debian/libc6-dev/usr/hppa64-linux-gnu/include' to `/usr/include': No such file or directory | 04:12 |
jbailey | doko: Who provides /usr/hppa64-linux-gnu ? | 04:12 |
jbailey | Do we? | 04:12 |
doko | hmm, glibc should provide that | 04:12 |
jbailey | 'kay, I'll add the mkdir. | 04:13 |
doko | it doesn't hurt. ok, it' currently not yet available, because I didn't build a hppa64 compiler yet. | 04:13 |
=== bhome is now known as BenM | ||
doko | infinity, lamont, lamont__: please requeue binutils for powerpc, if this is built and in the archives, then gcc-3.3 and gcc-3.4 for powerpc. | 05:49 |
lamont__ | doko: binutils kicked | 05:51 |
doko | lamont__: thanks | 05:52 |
=== lamont__ sighs, updates the chroots on ppc, and prepares to give binutils back again | ||
jbailey | doko: Thanks | 06:18 |
jbailey | lamont__: If you can do the same for ppc that would be lovely. | 06:18 |
jbailey | err | 06:18 |
jbailey | gdb on ppc | 06:18 |
lamont__ | jbailey: ok. but grumbling about gdb just doesn't seem kind until after an incomplete statement of work. :-) | 06:25 |
=== lamont__ grumbles about gdb. | ||
lamont__ | there -you happy> | 06:25 |
jbailey | Wha...? =) | 06:25 |
lamont__ | you just want it given back? or do I need to actually _do_ something to the chroot first? | 06:26 |
lamont__ | ah, same 'needs new procfs.h' issue | 06:26 |
jbailey | Sorry, I had assumed this was in the queue after the "updates chroots on ppc" =) | 06:26 |
lamont__ | lol | 06:26 |
jbailey | Bad optimiser! Stop applying yourself to human interactions! | 06:27 |
=== lamont__ is fighting with ipsec. makes for a cranky mood | ||
=== Seveas [~seveas@seveas.demon.nl] has joined #ubuntu-toolchain | ||
elmo | doko: have you seen this fun "gcc-4 miscompiles glibc" bug? | 07:46 |
jbailey | elmo: Doesn't affect us, we use gcc-3.4 to compile glibc. | 08:36 |
elmo | score | 08:40 |
elmo | can we make it a goal to not be using gcc-3.4 for anything by breezy? | 08:41 |
elmo | or is there anything that's unlikely to be compilable by 4.0 by release? | 08:41 |
jbailey | I think we'd planned on glibc and the kernel staying with 3.4 | 08:41 |
jbailey | We still need 3.4 in the archive for g77 anyway, so we're not losing much. | 08:41 |
elmo | whine | 08:41 |
elmo | ok | 08:41 |
jbailey | Neither upstream is advocating gcc 4 anyway. glibc 2.3.5 doesn't compile with gcc-3.4 without patches. The bug you're thinking of is for CVS HEAD of glibc. | 08:42 |
elmo | yeah, it just makes me cry that we're carrying 4 compilers around | 08:44 |
elmo | and using one of them to only compile basically two packages | 08:44 |
jbailey | 4? | 08:45 |
elmo | 2.95/3.3/3.4/4.0 | 08:45 |
jbailey | I thought we had dropped 2.95 already | 08:45 |
elmo | gcc-2.95 | 2.95.4.ds15-22 | breezy/universe | source | 08:45 |
jbailey | Oh, hmm. | 08:45 |
jbailey | NEeded to compile silo for sparc still. | 08:45 |
elmo | it's kind of nice for some commerical crap too | 08:45 |
jbailey | Feh | 08:46 |
elmo | well, if you have a better way to monitor hardware raid array on big 3 servers, lemme know :p | 08:46 |
jbailey | But 3.4 will probably need to stick around until the 4.2 timeframe when fortran catches up. | 08:46 |
elmo | that long? ouch | 08:46 |
jbailey | Follow 4.1, it doesn't look like fortran is moving fast to get the old g77 specific bits in. | 08:47 |
jbailey | s/Follow/I've been following | 08:47 |
jbailey | Is it build errors in software that you can change, or linking problems from precompiled modules? | 08:47 |
elmo | latter | 08:48 |
elmo | not modules tho, just binary progs | 08:48 |
elmo | usually linked to old libstdc++ | 08:48 |
elmo | it's not a huge problem, there are newer versions for FC 4 and so on, but I've had other problems with them | 08:49 |
elmo | (glibc symbol errors RUN AWAY) | 08:49 |
jbailey | afk a sec | 08:49 |
jbailey | Right. Those should probably all work fine against the breezy glibc. | 09:03 |
jbailey | elmo: (And everyone knows that backports of glibc are a good idea.. Right kids?) | 09:03 |
lamont__ | jbailey: B4CKP0R75 RUL3Z! | 09:07 |
lamont__ | hey, I know. lets backport breezy to Bo. | 09:08 |
lamont__ | All of it. | 09:08 |
jbailey | lamont__: package by package or feature by feature? | 09:08 |
lamont__ | package-by-package in one fell swoop | 09:08 |
lamont__ | and build it twice, of course. :-) | 09:09 |
jbailey | (/me fears how hard allowing bo to seamlessly dist-upgrade to breezy would be) | 09:09 |
elmo | jbailey: chicken | 09:09 |
lamont__ | jbailey: best accomplished via rapid injection of lead into the crainal cavity | 09:10 |
lamont__ | cranial? | 09:10 |
jbailey | I'm a tofudebeast, a far fiercer animal. | 09:10 |
lamont__ | jbailey: call the result frankenstein - people will expect seams that way. | 09:10 |
fabbione | jbailey: right now silo is building with the default gcc... i removed the strict b-d to allow silo in main | 09:26 |
jbailey | fabbione: Ah? I thought benc said it didn't work? | 09:28 |
fabbione | it doesn't ALWAYS work | 09:28 |
fabbione | it does most of the time | 09:28 |
jbailey | But.. it's *silo* | 09:30 |
jbailey | That's the way it is even with gcc-2.95 compiling it. | 09:30 |
fabbione | jbailey: tough.. i am not going to commit to support gcc-2.95 in main because of silo | 09:37 |
fabbione | if it boots.. good.. otherwise bitch upstream to fix silo :) | 09:37 |
jbailey | benc doesn't answer *questions* I have, I can't imagine him answeing me asking him to do actual work. =) | 09:37 |
fabbione | jbailey: he is busy working to make his own business | 09:38 |
fabbione | but i can still ping him on irc when he is around | 09:38 |
fabbione | : idle : 61 hours 24 mins 0 secs (signon: Sun Jun 12 08:14:43 2005) | 09:38 |
fabbione | anyway | 09:39 |
fabbione | time to be with my wife :) | 09:39 |
fabbione | cya tomorrow guys | 09:39 |
=== Seveas [~seveas@seveas.demon.nl] has joined #ubuntu-toolchain |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!