mozilla :: #build

14 Mar 2017
00:07glandiumrillian: ask for a loaner? or ask releng to check it out directly?
00:15tedmy best guess would be "it's built targeting a newer OS X"
00:15dmajorglandium: I tried your suggestion to build stylo with asan in bug 1336013#c2 (on windows)
00:16glandiumted: that usually doesn't crash with a segfault
00:18glandiumdmajor: yeah, no, that patch doesn't work
00:19dmajorglandium: how come it worked for heycam?
00:19glandiumdmajor: it didn't
00:19dmajorglandium: what can I do?
00:20glandiumdmajor: for a local build, just remove the imply_option that prevents --disable-jemalloc along --enable-stylo
00:20* dmajor tries
00:25* dmajor got distracted; tries for real
00:29dmajorglandium: configure looks good. trying build. thanks!
01:54froydnjglandium: I guess? not sure
02:32RyanVMwoo, just cooked up a haaaaaaaaaaaaaaacked MozillaBuild w/ 64bit python
02:32RyanVMain't pretty, but it gets the job done
02:36RyanVMof course, glandium, the reason I'm even playing around with this is due to people having OOM issues with cinnabar
02:36RyanVMwhich my understanding from your blog post is that that's getting better these days anyway
02:36RyanVMoh well
08:19ewongdoes anyone know the 'right' way of getting a new variable in recognized in Makefiles?
08:23glandiumewong: the 'right' way is to not touch
08:24ewongglandium: then how do I specify browser-specific variables to be used in the makefiles?
08:24glandiumbe more specific as to what you want to do
08:29ewongglandium: I want to be able to specify both the MOZ_SFX_PACKAGE and MOZ_INSTALLER_PATH for instead of hardcoding paths in
08:31ewongglandium: then TB can specify that same thing in
08:31ewongas right now, it's busting because of the hardcoded path
08:32glandiumdon't use
08:32glandiumjust set your variables in mail/installer/
08:34ewongglandium: thanks!
14:12Callek_cloud9mshal: in the interests of not going down a wrong road, and not stepping on your toes, this is the current WIP I have for INNER_UNMAKE_PACKAGE for OSX...
14:13Callek_cloud9(if you like irccloud's pastebin better)
14:24Callek_cloud9mshal: any concerns that stick out to you in a skim (knowing this is WIP)
14:51mshalCallek_cloud9: as a WIP, I think it looks great! Why did you need to change rsync() in Was that for the /Applications symlink?
14:51Callek_cloud9mshal: because ".DS_Store" needed to be rsync'd, but its not a folder
14:52Callek_cloud9so by the append of "/" it failed
14:52Callek_cloud9(I could use basic file copy, but old code did rsync, so figured I'd copy)
14:52mshalI ran into problems with the /Applications symlink on the repackaging end, so I was curious if that was why (it was also dying in rsync, but I solved it a different way)
14:53Callek_cloud9mshal: Yea, the dmg repacking code doesn't support symlinks
14:53Callek_cloud9(that is the binaries)
14:53mshalbut creating the dmg adds that symlink, so if we go from dmg -> tar.gz, we'll have a symlink in the tar.gz
14:53mshalI just removed it before repacking into dmg :)
14:54mshalCallek_cloud9: do you think we can try to wire this up today and see if it all fits?
14:54Callek_cloud9mshal: not certain offhand... kim's working on the taskgraph stuff, and aki hasn't done the signing piece yet
14:54Callek_cloud9(the signing server piece is the easiest, just needed time carved)
14:55Callek_cloud9mshal: we can probably manually take a .tar.gz and run it, but I don't have the mozharness stuff written yet either (though i expect that to be pretty quick to do)
14:55mshalCallek_cloud9: and was there a bug covering the repackage work already? I just sorta dove right in :)
14:55mshalahh, ok
14:55Callek_cloud9I though there was, but couldn't find it offhand, so I do plan to look for it today, and get that bug on file for you
14:55mshalcool, thanks!
14:56Callek_cloud9the mozharness stuff I was sorta waiting until the ./mach repackage magic was written so I could use that as a guide for what I need to pass/do, etc
14:56mshalCallek_cloud9: here's my WIP:
14:56mshalit just does tar.gz -> dmg, so it would need to be fixed up to supports Windows too
14:57mshaland as the commit message implies, it needs a configure run at the moment
14:57mshalif we need to I think I can grab the MOZ_APP_DISPLAYNAME from application.ini instead of configure, and then we just have to pass in all the cross-tool info (I guess in the TC task definition?) for hfs/dmg/etc
14:59Callek_cloud9yea, lets do configure+tooltool for now, and we can invest in dropping configure+tooltool later, imo
14:59Callek_cloud9we certainly will need to support windows, but that can be done "next"
15:01mshalsounds good
15:02mshalCallek_cloud9: so kim is working on the taskgraph parts that will call out to 'mach repackage'? And is the mozharness stuff you're talking about the part that runs tooltool for it?
15:02mshalwondering if I should help out on those parts (or am already responsible for them :)
15:03kmoirI am working on adding the taskgraph to add mac nightlies which is done. Now I am just adding the bits for the repackage task
15:03kmoirfor mac only
15:03kmoirto add after the signing task
15:04firebotBug 1324052 NEW, add macosx64 nightly builds to the taskgraph
15:04kmoirI have newer patches locally
15:05mshaloh, nice!
15:08Callek_cloud9mshal: yea, the mozharness stuff will run tooltool and get those binaries, and will run configure before calling ./mach for this
15:09mshalCallek_cloud9: anything else I can help with at the moment? If not I can try to pretty up the existing patch
15:09Callek_cloud9mshal: I'd be interested if --disable-compile-environment in configure is good for running ./mach (as in whats the minimal set of configure args that are ok to do here)
15:09* mshal tries
15:19Callek_cloud9I can always reuse the build one though, just less ideal
15:52geekodour08I am having some build issues with I am using venv with python 2.7 on arch linux but It's failing because of this error. some issue with the uuid_mac_address. I can't even check the Makefile because it's happening directly from any help?
15:52geekodour08error pastbin link:
16:07mshalgeekodour08: can you add a print("Package name=%s" % name) after this line?
16:08mshalunfortunately the error message doesn't seem to say which package is actually failing
16:08mshalso that might help pinpoint the problem :)
16:08mshalI suspect we might just need to update the package list or something
16:19geekodour08mshal: Unable to run, mozboot.base not found. I am cloning the readonly gecko-dev repo.
16:20mshalgeekodour08: oh, don't run directly - just add the print line there and re-run bootstrap
16:33geekodour08mshal: here's the ouput after adding the print statement:
16:35mshalgeekodour08: sorry in a meeting at the moment. I'll take a look in a bit if you can hang around!
16:59mshalgeekodour08: so the last package you see is support/ConEmu.xml and then you get the error message?
17:01mshalhmm I'm confused why that would be a package
17:03geekodour08i am sending the whole log, just a minute. It does install packages and everything but fails towards the end.
17:04mshalfrom my reading of, it looks like it should install some system packages with pacman (autoconf2.13, base-devel, ccache, mercurial, etc)
17:04froydnjted: you look like you have steampunk goggles on
17:05mshaland then it dies trying to build uuid.tar.gz
17:05tedfroydnj: hah, the exposure is just very messed up at this camera angle
17:06mshalI guess I would've expected it to just say 'Package name=uuid' rather than all those other filenames it displays
17:17geekodour08mshal: here's the complete log: I'll try it using the uuid now
17:27mshalgeekodour08: I am able to run on my arch VM, though it's fairly old. One difference I see is that yours picks up 'c99' as the compiler, while mine is 'gcc'. It seems that enabling c99 disables the definition of 'struct ifreq' in net/if.h
17:27mshalgeekodour08: did you setup arch to use c99 by default somewhere or does that happen automatically now with arch? I forget when I updated mine last
17:27mshalif you can set the default to gcc temporarily, that might help you get further
17:28mshalI'm also not sure why we need to build uuid from source rather than getting it from libutil-linux. Maybe a specific version is required?
17:40geekodour08mshal: I did CC=gcc python It's not stuck on the pervious error now. Thanks a lot!
17:40geekodour08I'll check if I set it somewhere to be c99
18:10mshalgeekodour08: glad it helped! It might just be the default in arch nowadays
19:52froydnjwhat's the best place to file a bug on getting backtraces symbolicated during the xpcshell packaging step?
20:11tbsaundefroydnj: I seem to remember ted linking me to an existing one not that long ago
20:38jgilbertglandium, I updated msan and tsan manifests. Back over to you!
23:43rillianfroydnj: thanks for adding -C overflow-checks btw. Did you have a plan for being able to set this from Cargo?
15 Mar 2017
No messages
Last message: 160 days and 9 hours ago