mozilla :: #build

17 Jul 2017
12:30igoldanhi
12:31igoldanhave there been any changes to the Windows 7 PGO build configuration?
12:31igoldanbetween July 10 & 11?
12:31igoldanthere are performance regressions on several of our Talos tests
12:32igoldanand they appear only on the Win7 platform
12:32igoldan+ they are totally unrelated to the other platforms
15:12mjfAlex_Gaynor:ping
15:14Alex_Gaynormjf: pong
15:15mjfFriday, we talked about Bug 1380132 causing peerconnection mochitest failures when obj-dir is not in source-dir. I noticed that the patches landed and tried a new build. Im still seeing failures.
15:15firebothttps://bugzil.la/1380132 FIXED, haftandilian@mozilla.com SSL info in url bar broken when launching from symlinked path or when objdir outside of repo
15:16mjfAlex_Gaynor: Should I open a new bug for this?
15:16Alex_Gaynormjf: Yes please, in the sandboxing component, and please CC me and haik
15:16Alex_Gaynormjf: and mark it as a blocker for let-level-3-ride-the-trains bug
15:17mjfAlex_Gaynor: Will do. And just fyi, it shows the NSS_NoDB_Init failed message.
15:18Alex_Gaynormjf: :-( ok, good to know. (FWIW, in addition to the objdir fixes haik landed, I'm testing an alternate approach involving hardlinks that might dramaticly reduce the number of edge cases like $objdir not in $srcdir)
15:19mjfAlex_Gaynor: thank you for all the help! I really do appreciate it!
15:19Alex_Gaynormjf: we broke it, we really ought to help fix it :-)
15:20mjf:-)
15:21mjfAlex_Gaynor: Bug1332190 is the level 3 bug you mentioned?
15:21firebothttps://bugzil.la/1332190 FIXED, haftandilian@mozilla.com [Mac] Enable level 3 Mac content sandbox, removing filesystem read access
15:21Alex_Gaynormjf: 1377522
15:21mjfAh! Literally Let the Level 3 content sandbox ride the trains! bug. LOL
15:28mjfAlex_Gaynor: Bug 1381517
15:28firebothttps://bugzil.la/1381517 NEW, nobody@mozilla.org Level 3 sandboxing for OSX builds with obj-dir outside of source-dir breaks PeerConnection mochitest
15:30mjfAlex_Gaynor: if there is any additional info you need from me to help, or to help test. Just let me know. Also, :ng is interested and tracking this along with me. Hes west coast, so slightly later availability that me at times if anyone needs confirmation.
15:30Alex_Gaynormjf: sounds good, I expect either haik or I will look into it today
15:55zeeHey Mozilla devs..im getting a cross over custom installer image which shouldn't be thr...i changed the firefox.icns and got this... any ideas how to fix it? http://imgur.com/a/5bxC1
16:15* ted shakes off the vacation dust
16:24dmajorted: you missed an interesting week :p
16:24tedyeah?
16:25dmajorwith the mac symbols problems and whatnot
16:25tedi have not yet seen that
16:28tedah
16:28tedmshal, gps: thanks for tackling that
16:41tedwish this was easier to reproduce locally
17:33mshalnp! It doesn't crash when you run it locally? It crashes pretty much every time if you run it in a one-click loaner, if that helps
17:33mshalI don't think I got around to testing it on my machine
17:35gpsted: don't worry - we found all the symbols/crashreports/sccache bugs while you were away. law of vacations is that things you are responsible for never break when you are around: just when you are on vacation :)
17:41tedhah, of course
17:41tedthe mac dsymutil thing had shown up before
17:41tedbut i was never able to track it down
18:26sfosterDoing my first build on a new windows, machine. It fails with msvcrt.lib(chkstk.obj) : fatal error LNK1112: module machine type 'X86' conflicts with target machine type 'x64'
18:26RyanVMsfoster: that's because it's trying to use the x64 rust toolchain while building 32-bit Firefox
18:27RyanVMeasiest fix is to build Firefox x64
18:27RyanVMhttps://developer.mozilla.org/en-US/docs/Mozilla/Developer_guide/Build_Instructions/Windows_Prerequisites#Common_problems_hints_and_restrictions
18:27sfosterRyanVM: ah ha. Why is it trying to build 32-bit firefox. Did I select that at some point?
18:27RyanVMadd those two lines to your mozconfig
18:27RyanVMno, that's what it defaults to and we don't handle it well with the Rust toolchain
18:27RyanVMthere's a bug for making that better somewhere
18:28sfosterRyanVM: ok perfect, thanks
18:49dmajorRyanVM: I don't think I've built win32 (except on try) at all in the past year. I wonder if it's time to switch the default?
18:50RyanVMseems reasonable - we basically require a 64-bit host OS to build at this point (the next MozillaBuild will require it)
18:50RyanVMmaybe not until we're shipping it by default, though
18:50dmajorheh, I guess that's fair
19:14teddmajor: i've thought about changing that a few times and i think i came to the same conclusion as RyanVM: until we're shipping it by default it's probably not the right thing to do :-(
19:23sfosterfailed at the finish line. Same new windows machine: mt.exe : general error c101008d: Failed to write the updated manifest to the resource of file "xpcshell.exe". Access is denied.
19:24RyanVMsfoster: https://bugzilla.mozilla.org/show_bug.cgi?id=1356493
19:24firebotBug 1356493 ASSIGNED, bobowencode@gmail.com Cannot build Nightly with Windows 10 SDK (10.0.15063.0) due to "ERROR: Cannot find mt"
19:24RyanVMoh hrm, that's a different error
19:24RyanVMdunno about that one
19:42gandalfted: welcome back! Would you have the time to take a last look at bug 1362617? It got your r+ and gps's r+, but he suggested waiting for you so you get the last look :)
19:46tedgandalf: i will take a look first thing tomorrow, OK?
19:46gandalfgreat! I'll set an NI on you then :)
19:46gandalfthank you!
21:24Caspy7Is there a log of this channel somewhere?
21:25Caspy7I had a convo in here a week or two ago I'd like to reread if possible
21:25dmajor 0:06.49  Compiling winapi v0.2.8
21:25dmajorwell this looks fun
21:27dmajorCaspy7: http://logs.glob.uno/
21:27glandiumted: https://treeherder.mozilla.org/#/jobs?repo=try&revision=1761399f03d238950923d9a1e70a17b2c47cf49c
21:30Caspy7dmajor: ah, thanks. There was no note in the topic and I was thinking I'd looked there before.
22:03dmajorif I change rust compilers, how do I force all rust stuff to rebuild/
22:05glandiumdmajor: should rebuild on its own
22:07dmajorglandium: that seemed to work when I rustup'd from beta to nightly, but not when I went back to beta
22:12glandiumdmajor: ask #rust if cargo build is supposed to rebuild in that case
22:30glandiumgps: thoughts on https://bugzilla.mozilla.org/show_bug.cgi?id=1372697#c38 ? (and should I reset the r?)
22:30firebotBug 1372697 NEW, nobody@mozilla.org Build a release build configuration against trunk
22:41gpsglandium: you should definitely clear the r?. i'll reply on bug
23:04tedglandium: \o/ my hero
23:06gpsbug 1380690 is... ugh
23:06firebothttps://bugzil.la/1380690 NEW, nobody@mozilla.org [Mac] Automatically determine the repo dir so that MOZ_DEVELOPER_REPO_DIR isn't needed
23:07glandiumhuh
23:08gpstl;dr sandbox gets confused by symlinks. sandbox needs to know path to srcdir and objdir so it can whitelist
23:08gpsor something to that effect
23:08gpssee also recent comments in bug 1380416
23:08firebothttps://bugzil.la/1380416 NEW, nobody@mozilla.org Investigate using hardlinks instead of symlinks in the dev build on mac/linux
23:38haikWith the Mac builds now being cross-compiled, can I run a mac tool during the build from a Makefile? For example, plutil?
23:40nalexanderhaik: in general, no -- unless you can produce that tool on the host during hte build. That's what ted reverse engineered to produce .dmg files, for example.
23:41haiknalexander: thanks and noooo!
23:41haiknalexander: perhaps there's a gnu equivalent command, I shall look into it
23:46glandiumgps: funny how getting software version numbers can be hard sometimes ;)
18 Jul 2017
No messages
   
Last message: 6 days and 23 hours ago