| pnorman | I'm a maintainer on a PPA which keeps getting virtual memory exhausted errors when it tries to build a package. This isn't surprising, given the huge codebase of heavily templated C++ code that takes half an hour to build. We're looking at switching from g++ to clang++ which might fix this, but in the mean time, is it possible to supply our own VM for building the package so we can get a more recent version up? | 02:22 |
|---|---|---|
| wgrant | pnorman: It's not possible to do that. But if you're getting VM exhaustion errors it sounds like you're running out of 32-bit address space, not running out of RAM. | 02:26 |
| wgrant | A bigger virtual machine won't help won't help on 32-bit architectures if you can't compile within 4GiB of address space. | 02:27 |
| wgrant | We normally see such problems during linking rather than compilation, though. Are you really managing to exhaust the address space in a single compilation unit? | 02:30 |
| pnorman | You're right, that was from the 32 bit build. Let me find an error for the 64 bit | 02:33 |
| pnorman | https://launchpad.net/~mapnik/+archive/ubuntu/nightly-trunk/+build/6304351 is an amd64 and I see virtual memory exhausted: Cannot allocate memory towards the end | 02:36 |
| wgrant | pnorman: Are you sure something's not very wrong? The builders all now have twice the RAM they did a month ago; this could never have built before then. | 02:39 |
| pnorman | I'm building a more recent commit at work. Let me check when last build is (the one in the PPA) | 02:41 |
| pnorman | Let me try that failing build on my server | 02:53 |
| === s8321414_ is now known as s8321414 | ||
| pnorman | building, but with -j6 so it shouldn't take the hour that the builders do | 03:37 |
| pnorman | Well 10 minutes in and it's using ~6GB of RAM. How much RAM do the builders have? | 03:48 |
| wgrant | pnorman: 4GB plus some swap. | 04:33 |
| wgrant | I don't recall how much swap. | 04:33 |
| wgrant | But that seems like an implausibly large amount of RAM for a single compilation unit to require, even if it's horrid C++. | 04:34 |
| pnorman | clang++ does it in about 2.5GB | 05:07 |
| pnorman | My build log inclues the message clang: error: unable to execute command: Killed and tells me to file a debian bug (https://launchpadlibrarian.net/183499083/buildlog_ubuntu-trusty-i386.mapnik_3.0.0%2Bdev20140829.git.8faa358-2~trusty1_FAILEDTOBUILD.txt.gz) | 07:20 |
| wgrant | pnorman: It probably got OOM-killed. | 07:23 |
| pnorman | I also have an amd64 build that's apparently failed, without a build log: https://launchpad.net/~pnorman/+archive/ubuntu/mapnik-testing/+build/6308209 | 07:24 |
| pnorman | Hmm. I'd say it's time to abandon any thought of i386 on the PPAs. Which is probably fine, most users using the nightly trunk builds will have 8GB+ of RAM, and many will have 64GB+, so they're pretty obviously not i386 | 07:25 |
| wgrant | I think your amd64 build might be driving them into swapdeath. | 07:27 |
| pnorman | Any way to diagnose? It'd surprise me, since clang++ built in <4GB https://github.com/mapnik/debian/issues/2#issuecomment-53838008 | 07:28 |
| wgrant | Are you sure you're using the same version of clang? | 07:28 |
| pnorman | Ubuntu clang version 3.4-1ubuntu3 on my server, and I'm depending on clang-3.4. Without a build log, I can't be *sure*, but I think I am | 07:30 |
| pnorman | ah, there we go. someone rebuilt it and I see a buildlog now, which includes the same error as i386. I wonder if I should bisect this, see if something pushed up compile memory usage | 07:45 |
| wgrant | pnorman: Have you tried building it in a trusty chroot? | 07:47 |
| wgrant | If this ever built before, there is a serious regression somewhere. | 07:47 |
| pnorman | I've built it not in a chroot just now | 07:48 |
| wgrant | On trusty? | 07:48 |
| pnorman | ya. | 07:49 |
| geser | btw: your utopic build used g++ while only your trusty test build used clang | 07:54 |
| pnorman | hmm. while I'm not immediately concerned with utopic (precise and trusty are what's used most), I can't think of anything that would cause that, as there's no release-specific stuff going on | 07:56 |
| geser | "Build-Depends: g++-4.8" (from https://launchpadlibrarian.net/183497914/buildlog_ubuntu-utopic-amd64.mapnik_3.0.0%2Bdev20140829.git.8faa358-1~utopic1_FAILEDTOBUILD.txt.gz) | 07:58 |
| pnorman | rules passes CXX=/usr/bin/clang++ to ./configure | 08:03 |
| pnorman | Oh, that's -1. That was still on g++. -2 is clang++ | 08:03 |
| mpt | Launchpad Janitor claims to have confirmed bug 1308321, but didn’t | 09:21 |
| ubot5 | bug 1308321 in unity (Ubuntu) "Can't open app from unity panel twice, only first time works" [Undecided,New] https://launchpad.net/bugs/1308321 | 09:21 |
| mpt | I am confused | 09:22 |
| === Trevinho_ is now known as Trevinho | ||
| === vidplace7_ is now known as vidplace7 | ||
| === Quintasan_ is now known as Quintasan | ||
| === daker_ is now known as daker | ||
| geser | change conflict perhaps? according to the activity log you both changed something in the same second | 10:05 |
| === semiosis_ is now known as semiosis | ||
| === elijah__ is now known as elijah | ||
| === zoktar_ is now known as zoktar | ||
| === popey_ is now known as popey | ||
| === FourDollars_ is now known as FourDollars | ||
| === cprov__ is now known as cprov | ||
| === dch_ is now known as dch | ||
| === ctracey_ is now known as ctracey | ||
| === jjox` is now known as jjox | ||
| === FloSoft_ is now known as FloSoft | ||
| === heroux_ is now known as heroux | ||
| === ral_ is now known as ral | ||
| === BradCrittenden is now known as bac | ||
| === beuno_ is now known as beuno | ||
| === Pici` is now known as Pici | ||
| === mthaddon` is now known as mthaddon | ||
| === dobey_ is now known as dobey | ||
| === TRB143__ is now known as TRB143 | ||
| === cjwatson_ is now known as cjwatson | ||
| sergio-br2 | i put one folder to bazaar ignore | 19:51 |
| sergio-br2 | * i put the folder in the .bzrignore | 19:51 |
| sergio-br2 | but if i delete some file from that folder, it does not ignores... | 19:51 |
| sergio-br2 | what's the magic i need to do? | 19:52 |
| sergio-br2 | i tried | 19:53 |
| sergio-br2 | folder | 19:53 |
| sergio-br2 | folder* | 19:53 |
| sergio-br2 | folder/* | 19:53 |
| === RagingComput is now known as RagingComputer | ||
| === jjox` is now known as jjox | ||
| === NikitaKonovalov2 is now known as NikitaKonovalov | ||
| === Quintasan_ is now known as Quintasan | ||
| meth | does launchpad have pull requests or is it just bug reports that act like both issues and pull requests? | 21:59 |
| === lifeless1 is now known as lifeless | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!