[02:38] is there any secure way to verify the ssh host keys for git.launchpad.net ? === yp is now known as ypwong === daker_ is now known as daker [15:42] <_Groo_> hi/2 all [15:47] <_Groo_> when building an llvm git snapshot, the ppa is running out of memory and killing dpkg-buildpackage, is there any workaround for this? [15:51] <_Groo_> see https://launchpad.net/~paulo-miguel-dias/+archive/ubuntu/llvm-snapshot/+build/9669573 [15:51] <_Groo_> https://answers.launchpad.net/launchpad/+question/292934 [16:53] _Groo_: I've answered on your ticket [16:53] <_Groo_> cjwatson: nice, let me read it [16:54] * cjwatson corrects himself [16:54] <_Groo_> cjwatson: hum ok, ill check, but just a quick reply. am i right? is the builder killing the ld linkage? [16:54] Yep, you're just running out of memory because you're using a pretty enormous amount of it [16:55] Nothing magical [16:55] <_Groo_> cjwatson: yeah, llvm is a beast, it can easily consume 10GB or more during linkage [16:55] There must be ways to trim that down though [16:55] <_Groo_> cjwatson: ill have to check the docs to see HOW am i supposed to lower this amount [16:55] Giving builds that much would significantly reduce the number of builders we can operate [16:56] Which would make quite a dent in our overall throughput [16:56] <_Groo_> cjwatson: i know, it would be an exception, i only use this particular ppa for llvm purposed, precisely because its such a beast [16:56] Like I say, we may be able to bump to 8GiB at some point, but probably not until the builder clouds are next redeployed [16:56] We can't make exceptions [16:56] I mean literally can't, all builder instances are identical [16:56] <_Groo_> cjwatson: im not asking you to, i was just merely saying that this would be a case for an exception [16:57] And I'm saying we have no possibility of exceptions [16:57] <_Groo_> cjwatson: i understood that from the start, its a global configuration :D [16:57] <_Groo_> cjwatson: thanks for the quick reply, you are awesome :D [16:57] Anyway, llvm can be built in Ubuntu, so it must be possible [16:58] Since we use the same builders [16:58] <_Groo_> cjwatson: back to the documentation to see how am i going to fix this [16:58] <_Groo_> cjwatson: good luck in compiling 3.9 [16:58] I'm sure it will be managed [16:58] <_Groo_> cjwatson: it changed from make to cmake, and although it made stuff much simpler, it bumped the memory consumption quite a lot [16:59] <_Groo_> cjwatson: but yeah, we will tame the beast [16:59] How would the build system make much of a difference? Unless of course it dropped some memory-reduction flags or is linking more objects at once [16:59] Fortunately I don't myself maintain toolchain packages [16:59] <_Groo_> cjwatson: is there any way for me to see the logs of the actual builder? like memory consumption and the OS logs? [16:59] No, we don't retain those [17:00] <_Groo_> cjwatson: both i guess [17:00] <_Groo_> cjwatson: ugh it will be try and error then :P oh well... [17:00] You can dump out whatever information you like during the build as long as you can get hold of it as non-root [17:00] <_Groo_> cjwatson: how? [17:01] you already have a dh_override_auto_build target, right? [17:02] make it run actual-build-command || { dump-information; exit 1; } [17:02] pseudocode obviously [17:02] the process being killed is ld, not any of the packaging machinery, so you have an opportunity to handle it [17:04] <_Groo_> cjwatson: ok, tks for the tip [18:06] hi folks, anyone know how the LP git web hooks work? and if they can be used to trigger builds on Jenkins?