mozilla :: #activity-stream

12 Jul 2017
00:54dmoseMardak: ROFL
00:54dmosegood times!
17:22Mardakdmose/mconley: interestingly talos seems to have been relatively happy/non-crashy in all of yesterday's pine pushes...?
17:23dmoseyeah, huh
17:23dmosethat was monday
17:23dmoseno pushes yesterday
17:23dmosei'm gonna push right now to pine just to see the current state of the world
17:54Mardakdmose: your current-nightly pine push didn't enabling activity stream ?
17:55dmoseMardak: so I did "npm run buildmc"; that should apply all the patches to m-c, correct?
17:55dmoseoh, right, it can't
17:55Mardaki can push to pine
17:55dmoseok, go for it
18:03Mardakdo we still need/want to disable prelaunch if it doesn't actually help with our test failures?
18:03dmoseMardak: good question. i think we have to ask gabor
18:04dmosebut yeah, maybe it doesn't really make sense
18:14Mardakdmose,ursula : can i get a rs for removing the patch that landed in m-c already?
18:15as-github-botMardak: Hey! Someone just assigned you a PR for review:
18:16Mardakursula: thanks. dmose: looks like taskcluster is broken?
18:16Mardakdecision task failed
18:16MardakHTTPError: 403 Client Error: Forbidden for url: http://taskcluster/queue/v1/task/ZNIZrl6iQqW7RAnrSP5_bQ
18:16Mardakalthough builds are still running anyway..
18:16Mardakso maybe we're okay for now
18:16dmoseMardak: yeah, i noticed that. i've seen it before.
18:17dmosewhere it runs anyway with that error message
18:17dmosenot sure what it means
18:17Mardakok well with the just merged PR, it should do the usual pine push to see if that's any better
19:08dmosemconley: yo
19:16tspurwaydmose, mardak: activitystream vidyo
19:18mconleydmose / Mardak: fun times - I can't reproduce the crash on the loaner hardware. :/
19:19dmose mconley: :-/ we're chatting a bit with mossop right now about handle to some of the stickier bugs
19:20dmoses/handle/how to timer/
19:25dmoses/handle/how to handle/
20:08as-github-botk88hudson: Hey! Someone just assigned you a PR for review:
20:08mconleydmose / Mardak: I might have something. Incoming comment...
20:09* dmose waits with bated breath
20:12mconleydmose / Mardak: here ya go:
20:14dmosemconley: wow, nice find
20:14* dmose considers the storage questio
20:14mconleywhat does Activity Stream use for storage?
20:15dmosei didn't think the stuff in m-c was actually using any yet
20:16dmosei mean, there's some places stuff, but i'm not sure if we'd be tickling that
20:16dmosek88hudson or ursula might know more...
20:17ursulamconley: what kind of storage?
20:18mconleyhow about local storage
20:18k88hudsonmconley: no local storage
20:18ursulathe system addon doesn't use local storage anymore
20:18k88hudsonsorry just in a meeting
20:18ursulathe only it really uses is places
20:18k88hudsonwe will need indexedDB for snppets
20:19dmoseand we write a few prefs too
20:20mconleydmose: putting a printf or something right above this IPC failure:
20:20mconleyand pushing last Friday's pine to try might help bolster this theory
20:20k88hudsonfile system storage for thumbnails
20:20mconleyWith what I'm hearing here, I have no idea why Activity Stream would tickle this
20:20mconleyperhaps timing
20:21mconleyperhaps it's initting itself at a time that moves something _else_ that starts accessing local storage further down the line or something
20:21mconleysome kinda clusterf(*%
20:21dmoseyeah, maybe one of the other talos tests
20:21dmoseor the browser
20:21dmoseor something
20:22Mardakdmose: #e10s
20:22dmosemconley: maybe instead of a printf, we force a stacktrace to be written somewhere?
20:25mconleyeither would be fine
20:25mconleyanything to confirm or refute the idea that PStorageParent::RecvAsyncPreload returning the IPC Error is why we're crashing
20:25mconley(IPC errors usually lead to intentional crashes)
20:25mconleylike, we killhard the content process anytime an IPC message screws up
20:32dmosewell, just for that , presumably following billm's suggestion of a reason seems better
20:32dmosethe idea of a stacktrace is so that we can maybe get some handle on who is tickling storage
20:37mconleynote that a reason just gives you a string
20:37mconleyat least, I think that's what happens
20:37mconleygetting some logging or stacktraces on accessing storage would be good too
20:37mconleyalso, knowing where getting at storage is failing sounds like the big win to me.
20:39dmosemconley: i imagine there's some way to turn on NS_ENSURE_* spew for opt builds.
20:39dmosemconley: though maybe it's easier to just manually put printf's in that Init function you pointed to and see if that does the trick
20:42as-github-botncloudioj: Hey! Someone just assigned you a PR for review:
20:43mconleydmose: either or
20:56Mardakdmose: do you know what needs to be printfed where?
21:11dmoseMardak: yeah
21:11dmoseMardak: vidyo?
21:16Mardakdmose: ok
21:21Mardakmconley: are any talos written as webextension?
21:21mconleyMardak: I don't believe so.
21:21mconleyI think they're making baby steps to migrate
21:21dmosemconley: or does it write out any of the perf data to storage?
21:22mconleybut so far, I think they're all mostly bootstrap add-ons
21:22mconleydmose: I don't believe so. I think it writes all of it to the console.
21:23mconleyyeah, I see no usage of any kind of DOM storage stuff in talos
21:32dmoseMardak: so on going back to the old commit; how would you approach that? reverting all subsequent commits? cutting a branch?
21:35Mardakdmose: yeah just branch from friday's pine push instead of monday's
21:36Mardakwell i'm using git
21:36Mardaki think you can just `hg up` to a commit and commit from there
21:37Mardakhere's the pine push from friday
21:37Mardakand this is the parent
21:39Mardakwe don't care about the extra heads on pine anyway
21:42Mardakk88hudson: lucky you! devtool perma orange is gone ;)
21:44dmoseMardak: thanks!
21:44dmoseMardak: is it possible to "try" from git?
21:45Mardakdmose: at least with git cinnabar, ./mach try
21:45dmoseMardak: oh, nice!
21:45Mardaki've been using this for talos ./mach try -b o -p win64 -u none -t all
13 Jul 2017
No messages
Last message: 11 days and 19 hours ago