mozilla :: #build

7 Aug 2017
09:49tedAlexAltea: FYI if you're passing that like --with-libclang-path=... then msys won't translate it to a windows path
09:50AlexAlteated: I see, thanks :)
09:50tedpath translation is a PITA :-/
09:50AlexAlteain the end I just went for mach bootstrap
09:51AlexAlteawould be awesome to mention it somewhere in the windows building page :)
09:51tedyeah, we should definitely do that
11:20Piketed, any idea on how to do make -C ...branding.. export "right" in a toplevel build? bug 1388026
11:20firebothttps://bugzil.la/1388026 NEW, nobody@mozilla.org installers-% top-level entry point is TOPLEVEL, and breaks on branding for windows installer
17:21rforbesI am having strange issues trying to build firefox today. It started for me yesterday. Whenever I build I get 0:06.67 ValueError: No JSON object could be decoded
17:21rforbeshas anybody seen this before?
17:21rforbesthis happens in the configure stage.
17:23dmajorrforbes: can you pastebin some more context before the error, and your mozconfig?
17:25rforbeshttps://www.irccloud.com/pastebin/4CO9j0wY/
17:25rforbesdmajor: ^^
17:28dmajorgps: ^ it's in watchman_version, any idea?
17:28araican you check the value of `out` there?
17:29rforbeshow do i do that?
17:29gpsor run `watchman version` from the command line
17:29rforbesbtw, this was just me pulling down source and building. nothing fancy
17:29araiput `print(out)` before `res = json.loads(out)` line
17:29gpsi don't think that will work. configure is special.
17:29araioh
17:29rforbeswatchman doesn't seem to have a way to check the version.
17:29gpsthis will get fixed by bug 1385380
17:30firebothttps://bugzil.la/1385380 ASSIGNED, gps@mozilla.com Build failure when checking "watchman version"
17:30rforbesis there a way around it?
17:30gpsrename watchman.exe
17:30rforbesok, thanks.
17:31rforbesalthough, i am on linux
17:31rforbesbut i get the idea.
17:31gpsoh?
17:31gpswhat is the output of `watchman version`?
17:32rforbesit doesn't return anything.
17:32rforbesrforbes@ubuntu:~/src/firefox-plain$ watchman version
17:32rforbesrforbes@ubuntu:~/src/firefox-plain$
17:33gpswhat about `watchman --version`?
17:33gpsand i assume `watchman version` exits with non-0 exit code?
17:34rforbesso, first this is what flags watchman appears to have
17:34rforbeshttps://www.irccloud.com/pastebin/Nz0WwMUI/
17:34rforbesand when i do --version it doesn't like it.
17:34rforbestoo few arguments.
17:35gpsso your `watchman` is not https://facebook.github.io/watchman/
17:35rforbesit's either what comes with ubuntu or what ./mach bootstrap installed. I didn't do anything special.
17:35gpsO_o
17:35rforbesand nothing i have changed recently.
17:36gpsyou are the only person complaining about this failure. the watchman code has been in the repo for over a week.
17:36gpsi'm finding it difficult to believe that `mach bootstrap` did this
17:37rforbesi find it weird too, clearly something is wrong on end but all i did was pull down m-c, run mach bootstrap and mach build
17:37gpsi believe you. it's... just wonky
17:38rforbesheh, my feelings exactly. :)
18:20gpsugh. emma is starting to shame components with untriaged bugs
18:33froydnjand so it begins
18:51rforbesgps: definitely something weird about this system. i spun up an ubuntu instance on ec2 to test it out and it seems to be working fine.
19:01rforbesI tried deleting my .mozbuild directory but that didn't help.
19:51dmajorgps: did you ever figure out that selftest.py failure thing from some days ago?
20:01gpsdmajor: i can't recall which selftest.py thing you are referring to :/
20:14dmajorgps: some win64 failure on a push you were looking at with glandium. I'm forgetting the details too, and the logs page seems to be down.
21:39mshalgps: is bug 969744 still relevant? I was poking around for open install manifest bugs
21:39firebothttps://bugzil.la/969744 NEW, nobody@mozilla.org Better support for install manifests
21:44gpsmshal: possibly. but i'm sure the patch has been bit rotted. so i'm going to wontfix it
21:45gpsi do have an in-progress series to overhaul how links are handled
21:45RyanVMglandium: is the linux64 base-toolchains job yours?
21:45gpsmore generally, overhaul how file "installs" are treated
21:45gpsallows you to do things like switch the file install "policy" from symlink to copy
21:46mshaloic
21:47gpsi'll polish that off once version control fires die down
21:47mshalany idea if there's a reason not to use --track everywhere?
21:48gpswhat is --track again?
21:48mshalit was added in bug 1230060 for the faster make backend
21:48firebothttps://bugzil.la/1230060 FIXED, mh+mozilla@glandium.org mach build faster doesn't remove files when they are removed from moz.build or jar.mn
21:48mshalI guess it's a glandium question :)
21:49mshalIIUC it's so we can correctly remove files without blasting away the whole target directory
22:25glandiumrforbes: fwiw, there is no ubuntu package providing a "watchman" binary
22:25glandiumRyanVM: yes, why?
22:26RyanVMjust trying to figure out why it's running on Beta
22:26rforbesglandium: yeah, that is what i figured out. Not sure why that VM image is having so much issues. It started happening this weekend.
22:27RyanVMglandium: also, a bit odd that it's on the Linux x64 lines on TH instead of with the other toolchain jobs
22:27glandiumRyanVM: it's not a toolchain job
22:27RyanVMglandium: I guess they need the bug 1360609 treatment?
22:27firebothttps://bugzil.la/1360609 FIXED, mh+mozilla@glandium.org Only run toolchain jobs on trunk
22:27RyanVMoh, right
22:27RyanVMthis is the "build Firefox with minimum supported toolchains" right?
22:28glandiumyup
22:28glandiumand I think they should run on branches
22:28RyanVMk, guess that makes more sense then, yeah
22:28RyanVMlinux64 debug job on Beta has been going for 145min, though :\
22:29glandiumoh, bug 1360609 reminds me that run-on-projects should be removed from the toolchain jobs now
22:32glandiumah no, because of optimize_target_tasks=false
22:38PikeI'm looking for a reviewer for bug 1385227 for installers-% foo. I think the hardest part is to not look at the code that I remove. The most important part is that the code doesn't move in the wrong direction. gps? glandium?
22:38firebothttps://bugzil.la/1385227 NEW, nobody@mozilla.org Make l10n repack builds do what they're supposed to do
22:49emilioglandium: wow, so ironic that the code that crashes clang-cl is the code for which we actually would want clang-cl (the Servo integration stuff) :)
8 Aug 2017
No messages
   
Last message: 14 days and 22 hours ago