• Waldo
    my ability to push things has not been disabled \o/
  • pbone
    Hi Waldo
  • » Waldo waits for philor to snark
  • Waldo
    hi pbone
  • sfink
    can CustomElementConstructor::CustomElementConstructor gc?
  • mrgiggles
    sfink: No, |void mozilla::dom::CustomElementConstructor::CustomElementConstructor(mozilla::dom::CallbackFunction*)| cannot GC
  • pbone
    Waldo: to answer your question. Yeah, I started in May.
  • pbone
    So far so good ;-)
  • Waldo
    nice
  • pbone
    I work with sfink, jonco and chia-hung on the GC.
  • pbone
    You? What did you do and what do you do now? for work and fun?
  • Waldo
    pbone: I generally do VM and runtime-y sorts of things, and jack-of-all-trades as many other things as I can, more or less
  • Waldo
    pbone: I also offer puns as a service
  • pbone
    Oh nice. Puns as a service reminds me of my freind who runs a rick astly hotline (as a service I suppose).
  • Waldo
    mfbt/ in broader Gecko-land (or maybe more primordial, seeing as it's used by SpiderMonkey and non-SpiderMonkey Gecko both
  • Waldo
    quality
  • Waldo
    and I've dabbled in other areas of Gecko -- the initial window.postMessage and background-size implementations were my fault
  • pbone
    Cool. Yeah, I'm still getting familiar with mfbt. I havn't done C++ for about 15 years before I started here, so that was a little eye-opening.
  • sfink
    Waldo also fixes my security bugs
  • Waldo
    I also might have written the most-used JS HTTP server in the world, which was pretty fun
  • pbone
    I knew that C++ could do some of these things, and that these kinds of things were normal-ish. I'd heard of Boost for example. But hadn't actually tried to understand any of it.
  • Waldo
    sfink: don't look at me, automation⊙bmo.tld revoked my javascript-security-team permission and the infra bug hasn't been fixed yet to restore it :-)
  • sfink
    so you're saying that if I CC you, I'll be first in line?
  • pbone
    So if you're not working for Mozilla now, and do Firefox/SM for shits and giggles, then what do you do for money?
  • Waldo
    sfink: that's not what I--SQUIRREL!
  • Waldo
    pbone: I'm back at Mozilla now -- took off May 8-September 29 to thru-hike the PCT Mexico to Canada
  • Waldo
    240h PTO, 3.5 months unpaid leave
  • pbone
    Ah cool. that's even cooler. Welcome back.
  • pbone
    I didn't do it this year. I felt that I was too busy, and havn't the last few years because my son was too long and we didn't want to leave him with a babysitter that long.
  • pbone
    but my anual pilgramage/spiritual rejuvenation is downhill skiing.
  • sfink
    yeah, I've left my kids on the rack for too long too
  • Waldo
    sfink: you know, they'll stretch out on their own if you'd just be patient
  • pbone
    So hopefully next year will leave Alfie with his grandparents and go skiing for a weekend or so.
  • Waldo
    good fun
  • pbone
    oh. s/too long/too young/
  • Waldo
    :-D
  • Waldo
    I grew up in Michigan, family did cross-country skiing but not really anything downhill
  • Waldo
    outskirts of the Detroit area, so we had seasons but not super-wintry winter
  • sfink
    no typo shall go unpunished
  • pbone
    Ah, My dad, brother and I are adreniline junkies when it comes to this stuff.
  • Waldo
    no, they will all go very punnished
  • pbone
    Although my brother went to martial arts and roller derby, and dad seems to have left skiing (possibly due to age, I didn't ask).
  • pbone
    but I've kept it up and introduced my wife to it.
  • pbone
    She's been a few times now and the last time we went she just got to the stage where she got to the bottom of the run and said "LETS DO THAT AGAIN!"
  • Waldo
    my uncle did ski patrol at a nearby hill (not all that much of one), I went and tried it once, but I think that's the sum extent of my downhill skiing
  • pbone
    then we had Alfie and didn't go skiing...
  • Waldo
    :-D
  • Waldo
    give it a few years, start him early :-)
  • pbone
    yeah.
  • pbone
    He's old enough that he'd LOVE snow. but would still struggle to ski.
  • pbone
    but a couple of years and he can try.
  • pbone
    I started young.
  • Waldo
    my family's outdoor pursuits were mostly cross-country skiing MLK weekend, a lot of biking, and (other than my mom) backpacking
  • pbone
    I honnestly say that I can ski better than I can walk. But that's mostly because just after Liz and I got together I went skiing for a week without her, she was worried I'd hurt myself, then I came back and before arriving at her parents' place to see her that night I triped walking down the street and cut open my knee.
  • pbone
    so I walked into her home with blood comming out, after saying I wouldn't hurt myself.
  • Waldo
    heh
  • pbone
    I had to explain it didn't happen skiing, I just tripped on the curb a block from her family's house.
  • » Waldo vaguely remembers an MLB player (?) who fractured an arm tripping on a curb, or maybe doing it repeatedly just after it'd healed the first time
  • pbone
    Murphy's law.
  • Waldo
    one of those stupidest-IR-reasons-ever sorts of articles :-)
  • » philor looks at Waldo's bustage
  • philor
    rats, just our crappy network
  • Waldo
    oh, did I tempt fate?
  • Waldo
    ah, good, good
  • philor
    oddly, that sort of fate-tempting has very very good odds of getting away with it
  • Waldo
    those meddling kids
  • philor
    only things I can think of that don't are first push after a closure you've complained about, and a quick repush after a backout
  • Waldo
    /builds/worker/workspace/build/src/gfx/angle/src/common/platform.h:64:21: fatal error: d3d11_3.h: No such file or directory
  • Waldo
    do we really just have t3 broken builds on treeherder displaying by default these days?
  • philor
    broken this morning
  • Waldo
    I did some trying of a number of builds of those patches, testrun on linux64 even tho I wasn't changing any testy sort of code really, so I am cautiously optimistic
  • Waldo
    (which if memory serves, Al Franken once memorably explained should be interpreted as "we're screwed")
  • » philor looks at the unstarred count, settles in to do two full-time jobs after finishing his own full-time job
  • philor
    not that I'm bitter
  • » bz_dinner sees philor is sour, not bitter
  • Waldo
    well, he's certainly not hoppy
  • bz
    sfink: ping
  • arai
    I'm wondering the lifetime of type information
  • arai
  • arai
  • arai
    and it seems that the generation corresponds to sweep
  • arai
    and this says different generation value is stale dxr.mozilla.org/mozilla-central/rev…/src/vm/TypeInference.cpp#4566-4570
  • arai
    so, what's actually expected to be retained across GCs ?
  • jandem
    arai: type information can be retained across GCs
  • jandem
    when we keep type information, the generation field is used for incremental sweeping
  • jandem
    if it doesn't match the TypeZone generation, we still have to sweep it
  • jandem
    type sweeping is just moving all LifoAlloc data into a fresh LifoAlloc
  • jandem
    "compacting" it
  • arai
    when the incremental sweeping starts, the generation field is flipped, right?
  • jandem
    yeah, and when it ends everything should have a matching generation field
  • arai
    "type sweeping" is different than the sweeping for other values/objects ?
  • arai
    ah...
  • jandem
    type sweeping also discards stuff that points to dead GC things I think, so that part is similar to GC sweeping
  • jandem
    but it also compacts the type LifoAlloc
  • arai
    so, while type sweeping, not all type information is discarded?
  • arai
    and some of them are kept with new generation value?
  • jandem
    yes exactly
  • arai
    makes sense
  • arai
    thanks :)
  • jandem
    type sweeping and the generation applies to both ObjectGroup and TypeScript
  • djvj
    nbp: jit team meeting happening today?
  • nbp
    djvj: no
  • nbp
    djvj: the next one is in 4 weeks
  • nbp
    djvj: I annouced that a minute ago on #ionmonkey
  • nbp
    djvj: see you next week, right?
  • djvj
    nbp: yeah. not on #ionmonkey so didn't see it.
  • djvj
    nbp: see you next week :)
  • tcampbell
    till: ping. I have some small questions about promises
  • till
    tcampbell: pong, I'm all ears
  • tcampbell
  • firebot
    Bug 1402596 — NEW, nobody⊙mozilla.org — Crash [@ JS::Rooted<JS::Value>::registerWithRootLists] with evalInCooperativeThread and drainJobQueu
  • tcampbell
    the issue is that cooperative threads don't init promise stuff
  • tcampbell
    no callback and no internal jobqueue
  • till
    ah
  • tcampbell
    does it make sense to just forword the callback from the siblingContext?
  • tcampbell
    I tried just UseInternalJobQueues, but since runtime is already initialized, that doesn't make sense
  • till
    I don't know enough about how the cooperative threads stuff works
  • till
    yeah, that's definitely not right
  • till
    you'll still have to let the browser schedule the promise jobs, but the question is how
  • tcampbell
    oh, what/who is the coorperative threads stuff?
  • till
    ah, this is in the shell
  • till
    sorry, I didn't actually look at the test case :(
  • tcampbell
    yeah, shell only. it *seems* like browser always registers promise infrastructure for any jscontext
  • till
    ok, that's good
  • till
    hmm, so I guess we'll want to have cooperative threads use their own queues in case the promise queue is handled internally
  • till
    it'd be somewhat weird not to have that, given that the entire reason for doing the cooperative threads thing is to improve scheduling :)
  • tcampbell
    heh
  • » tcampbell has zero idea what the cooperative threads stuff is about
  • till
    tcampbell: bug 1323066
  • firebot
    bugzil.la/1323066 — NEW, bhackett1024⊙gmail.com — [meta] Multithreaded JS runtime
  • tcampbell
    I can try added a cooperative flag to UseInternalJobQueues that creates a new context queue, but leaves the runtime structures alone
  • till
    tcampbell: it looks like the relevant fields are already thread-local searchfox.org/mozilla-central/source/js/src/jscntxt.h#912-925
  • till
    tcampbell: so I guess initializing them during creation of a coop thread might work
  • tcampbell
    till: alright, I'll poke around with that and see if we missed anything obvious
  • till
    tcampbell: great. I unfortunately really don't have time to work on this right now, otherwise I'd do it
  • till
    tcampbell: in case you're interested, here's more information on the cooperative scheduling: wiki.mozilla.org/Quantum/DOM and hacks.mozilla.org/2017/06/an-inside-look-at-quantum-dom-scheduling
  • tcampbell
    ooh, will read those
  • tcampbell
    thanks
  • jonco
    smaug: hey, did you make any GC changes recently? I'm trying to track down an intermittent test failure that started on 4th october…
  • tcampbell
    jandem: so with the this/singleton stuff, you observed that if we called constructor from jit -> interp, we never set the singleton flag previously?
  • jandem
    tcampbell: yeah
  • tcampbell
    interesting
  • jandem
    tcampbell: actually the jits create |this| before making the call, so it's possible it happened to work
  • jandem
    tcampbell: then the main issue would be VM -> interp constructor calls
  • tcampbell
    when do we have that? and how is it handled now. It looked like all your caller-side patches were within interp/jit
  • tcampbell
    jandem: ^
  • tcampbell
    oh, InternalCallOrConstruct. I see
  • jandem
    yeah was just going to link to searchfox
  • tcampbell
    :)
  • smaug
    jonco: I don't think I've done anything in FF58
  • smaug
    and I don't recall mccr8 making changes either
  • smaug
    thinking
  • smaug
    hmm, or was there one...
  • smaug
    there was bugzilla.mozilla.org/1377914, but that got backed out
  • firebot
    Bug 1377914 — NEW, bugs⊙pettay.fi — Don't trigger a full GC in nsGlobalWindow::DetachFromDocShell()
  • jonco
    smaug: ok, thanks for looking
  • nbp
    jorendorff: Are you interested in "Integrating Cretonne in SM" on the Sheet, and chose the wrong column?
  • jorendorff
    nbp: maybe? can you give me a link to the sheet
  • nbp
    done
  • jorendorff
    as a convenience, we should cast votes for the people who haven't voted yet
  • jorendorff
    based on what we think they would like to talk about
  • nbp
    jorendorff: or spam them ;)
  • florian
    Hi. Is there a way for JS code running in a worker to give a hint to the JS engine that GC'ing now would be a good idea? I have a page where I typically have workers taking more than 1GB of memory each. Pressing "Minimize memory usage" in about:memory releases that memory.
  • mrgiggles
    the mozilla-inbound tree is now closed (test failure for bug 1406041 and bug 1406897)
  • firebot
    bugzil.la/1406041 — ASSIGNED, bbouvier — Crash in js::wasm::Code::tiers when profiling Tank's demo
  • firebot
    bugzil.la/1406897 — ASSIGNED, bbouvier — Crash [@ js::wasm::Code::tiers] due to over-recursion with evalInCooperativeThread
  • nbp
    florian: You should ask either jonco / sfink / pbone.
  • jonco
    florian: there's no way to do this at the moment
  • jonco
    florian: in general it's better to leave it up to the engine to decide, as it has to do this anyway
  • florian
    All of that memory is in array buffers. Is it possible that the engine is not gc'ing hard enough when there's plenty of data around stored in array buffers?
  • nbp
    jonco: Do we have GC when the browser is going idle?
  • jonco
    nbp: we gc workers quite agressively IIRC
  • jonco
    nbp: eager gc when entering idle state is something we're looking at at the moment
  • nbp
    ok :)
  • jonco
    florian: it's possible that there is a bug in how we handle this
  • florian
    what the script does is extract the pixel data of all frames of a video, and it sends batches of 10 frames to workers. Workers compare the frames, and send back to the main thread only the frames that are significantly different from the previous one.
  • jonco
    florian: ok
  • mrgiggles
    the mozilla-inbound tree is now open
  • jonco
    florian: do you have a test case
  • florian
    so the large amount of garbage I have in workers is from array buffers that were received, but not sent back (because they are useless frames)
  • jonco
    ?
  • jonco
    florian: or a way that I can try this?
  • jonco
    florian: oh … interesting, we might not account for received buffers correctly
  • florian
  • florian
    download queze.net/goinfre/tmp/blob-testcase/Segment_0001-big.mp4 (67MB) and then drop it on the test page (after setting the needed pref in about:config)
  • jonco
    florian: cheers!
  • florian
    and look at about:memory before the processing ends
  • florian
    (when it's done, the script terminates the workers)
  • bz
    florian: are these buffers structured cloned, or transferred?
  • jonco
    florian: oh yeah, something seems wrong there
  • jonco
    florian: I'll file a bug
  • bz
    Looks like transferred
  • florian
    bz: For each batch, 10 are transferred and one is cloned
  • » bz bets it's the transferred ones that are busted
  • bz
    lemme check something
  • florian
    (because there's one frame that needs to be accessible to 2 different workers)
  • bz
    So JS_StealArrayBufferContents does not decrement any memory counters, which is maybe ok
  • bz
    But JS_StealArrayBufferContents does not bump any counters either
  • bz
    er, JS_NewArrayBufferWithContents
  • florian
    jonco: when looking at about:memory on that page, don't worry about filing a bug for the large heap-unclassified in the main process at the end, I already filed bug 1407636
  • firebot
    bugzil.la/1407636 — NEW, nobody⊙mozilla.org — Using blobs in a page loaded in a content process causes a large amount of heap-unclassified memory
  • bz
    florian++
  • jonco
    florian: great
  • » bz was going to do some DMD bits, then saw florian already did
  • jonco
    bz: yeah, we probably forgot to update the counter
  • florian
    bz: I only filed a bug for the unclassified memory in the parent process. There's also some in the content process, but it disappears as soon as the script is done processing so it's harder to catch
  • » bz not sure whether JS_NewMappedArrayBufferWithContents should
  • jonco
    bz: I'll look into it
  • florian
    bz: DMD gives me pastebin.mozilla.org/9069838 for the unclassified memory in the content process.
  • bz
    florian: 1.8GB of blob data there...
  • bz
    And some IPC pickle stuff...
  • bz
    <sigh>
  • florian
    is it the actual blobs that the page can keep alive, or is it temporary objects that will be freed soon? I assumed it was the latter, but if it's the former it likely deserves a bug too
  • bz
    I don't know
  • florian
    that page keeps around ~2.3GB of data in blobs
  • bz
    we seem to have no memory reporting for blobs at all
  • florian
    I thought bug 1313859 added some
  • firebot
    bugzil.la/1313859 — FIXED, amarchesini⊙mozilla.com — large amount of heap-unclassified related to blobs
  • bz
    Worth a bug right there; cc :njn and :baku
  • » bz looks
  • bz
    oh, it's in file.cpp, not blob.cpp
  • bz
    er, in StreamBlobImpl
  • bz
    So dunno
  • Waldo
    jonco++
  • Waldo
    (header reductions)
  • bz
    we have so many blob impls...
  • jonco
    Waldo: cheers
  • jonco
    Waldo: also, we should do this for the rest of the engine… but one thing at a time
  • Waldo
    yeah
  • Waldo
    I try to do it incrementally when I touch stuff, but
  • bz
    Yeah
  • bz
    canvas toBlob does a MemoryBlobImpl
  • bz
    we report from BlobImplStream
  • » bz looks
  • bz
    er, now it's called StreamBlobImpl
  • bz
    Point is, we don't have a memory reporter for MemoryBlobImpl
  • bz
    erm...
  • bz
    maybe we do
  • bz
  • bz
    Then I dunno why it's not reporting your memory!
  • » bz writes testcase
  • » florian will be offline for about half an hour
  • jorendorff
    bz: we have email
  • bz
    florian: Ok, my testcase does show the blob memory....
  • bz
    florian: being reported
  • » bz pokes
  • bz
    Maybe you took your measurement while there were still encodes pending?
  • bz
    florian: ^
  • bz
    jorendorff: looking
  • bz
    wait
  • bz
    we already did a 56.0.1?
  • bz
    And didn't take this fix?
  • bz
    argh
  • bz
    jorendorff: And another. Do you want to reply to Liz, or should I try to, in terms of the security impact?
  • » bz doesn't have a concrete thing to point to other than "this is really scary"
  • jorendorff
    bz: I want to discuss with you :) /msg-ing
  • bz
    ok
  • florian
    bz: it's indeed possible there were still pending toBlob calls
  • Waldo
    gandalf: fwiw, as far as relative-time-formatting goes and using a delta -- that limits you to relative strings like "3 days ago" and prevents exposing something like "last Friday"
  • Waldo
    gandalf: I can certainly imagine some authors thinking the latter would be better
  • gandalf
    we're planning to add this in 2.0 of the RTF API
  • Waldo
    ah, good
  • gandalf
    I agree it would be good to have, and CLDR does have the data
  • gandalf
    :)
  • » Waldo saw tc39/proposal-intl-relative-time #15 and didn't see that deficiency especially mentioned anywhere, didn't feel like spamming a closed discussion
  • crowbot
    Issue #15: [discussion] epoch vs relative distance in ms as the value to format - tc39/proposal-intl-relative-time #15
  • Waldo
    yes, crowbot, that is the URL I mentioned, thanks for repeating it
  • Waldo
    mrgiggles: pun
  • mrgiggles
    Waldo: The new weed whacker is cutting-hedge technology.
  • gandalf
    :D
  • gandalf
  • crowbot
    Issue #20: [future] Add days of the week - tc39/proposal-intl-relative-time #20
  • gandalf
    yes crowbot , this one.
  • Waldo
    gandalf: not just days of the week, tho, "last November" and so on as well
  • Waldo
    "last spring" for quarters too
  • Waldo
    maybe other units if I thought about it
  • gandalf
    we don't have that in CLDR yet :(
  • gandalf
  • gandalf
    but in the future I hope they'll add last November case
  • Waldo
    yeah, I'm thinking abstractly from a user perspective, not from what data we actually have available
  • Waldo
    it seems like data that *could* be tracked in CLDR, at least, so if we wanted to expose the capability, the function could be added to the libraries we use
  • bz
    Waldo: Gotta do at least the two different kinds of quarters
  • bz
    waldo: month mod 3, or season
  • bz
    Too bad they don't match up
  • bz
    (also, our 401 seems to have quarters ending at end of Jan, April, Jul, Oct for reasons that are not clear to me)
  • » Yoric is realizing how awful the syntax (and semantics) of `for..in` are.
  • Yoric
    Why, oh why is it defined as `for (expr in expr)`, with the left expr being repeatedly evaluated?
  • sfink
    oh, wow
  • sfink
    function f() { print("Called!"); return 'value' }; for (h[f()] in [10,20,30]) { print("yeehaw " + JSON.stringify(h)) }
  • sfink
    oh
  • sfink
    h = {} first, I guess
  • Yoric
    sfink: I have no clue why `f()` is called only once. That doesn't seem to match the semantics.
  • sfink
    er, what? In my test, it was called multiple tiems.
  • Yoric
    jorendorff: In the future, I have more chances of reading your replies if you do not leave MozReview mid-conversation :)
  • sfink
    hm, what shell was I using...
  • Yoric
    sfink: Weird. In mine, it was called only once.
  • Yoric
    And I copied/pasted your code.
  • jorendorff
    Yoric: oh no. I don't understand, I do that all the time.
  • jorendorff
    Does it trigger some lossiness in mozreview?
  • sfink
    which means I probably messed it up; I didn't run that actual code
  • Yoric
    sfink: Ah, it's somehow related to `print`. If calls several time if I replace it with `console.log`. I'm running this in MozReview.
  • sfink
    er, no, it does work, and it calls it 3 times
  • sfink
    ah!
  • Yoric
    er... in MozReview? I meant in DevTools.
  • sfink
    it might suppress identical outputs or something
  • Yoric
    Yeah.
  • Yoric
    jorendorff: Well, not technical lossiness, just difficulty of following the same conversation when it starts on one medium, then jumps to another one, while still being related to stuff that lives on the first one.
  • jorendorff
    Yoric: but my assumption was that when I hit publish, everything would be published to the bug
  • jorendorff
    and ... well, I certainly *intended* to publish all my comments
  • jorendorff
    it was all supposed to end up in bugzilla
  • jorendorff
    Yoric: utter delight @ "I'm running this in MozReview"
  • Yoric
    :)
  • Yoric
    jorendorff: Sure, but when you write directly in Bugzilla, nothing is reflected in MozReview, which is where the specific issues are opened.
  • Yoric
    jorendorff: I am ready to push a new version, because it has already changed quite a lot since your last review. I do not address all your comments yet, because that will require even more changes.
  • jorendorff
    ok
  • Yoric
    I am probably going to try and address a few more before I request a new review, although probably not all, because scope creep, etc.
  • Yoric
    Your call on whether you want to review it in the current state or wait for further updates.
  • jorendorff
    I'll at least look!
  • jorendorff
    Yoric: Will you be in Portugal?
  • Yoric
    jorendorff: Yep.
  • jorendorff
    sweet
  • jorendorff
    I'm taking the next two days off, but i'll see you there
  • Yoric
    jorendorff: Also, there's now a mailing-list for binast: lists.mozilla.org/listinfo/dev-binast
  • » jorendorff subscribes
  • Yoric
    jorendorff: Ok, pushed for review a new version. As you'll see... there are very few lines in common :)
  • jorendorff
    hahaha ok
  • Yoric
    The changes are basically your remarks, plus the side-effects, plus the side-effects of fuzzing after your remarks, plus...
  • Waldo
    out of semi-idle curiosity, do we still do inbound -> m-c merges around once or so a day these days?
  • RyanVM
    Waldo: give or take
  • Waldo
    k
  • RyanVM
    see coop's dev-platform post from a day or two ago
  • Waldo
    I assumed as much, just very mildly surprised patches I pushed yesterday hadn't merged yet, and I was assuming I was just being a little premature
  • RyanVM
    Waldo: looks like you just missed this morning's merges
  • RyanVM
    might be another one soon before the 3pm nightlies start (we're running those 2x daily now in case you missed that), but no guarantees
  • Waldo
    it's good
  • Waldo
    no rush really
  • gandalf
    till: ping
  • jorendorff
    sfink: ping
  • sfink
    pong
  • jorendorff
  • firebot
    Bug 1407236 — NEW, nobody⊙mozilla.org — Intermittent js/src/jit-test/tests/gc/bug-1109913.js | Unknown (code 0, args "--ion-eager --ion-offt
  • jorendorff
    sfink: what does "Unknown (code 0)" mean?
  • jorendorff
    seems weird
  • jorendorff
    "we have no reason to think this test failed, it's the craziest thing"
  • sfink
    wtf? That log file has a failure in check_spidermonkey_style.py
  • sfink
    but yes, I'm looking at the test "failure"
  • sfink
    wait, I killed off one test like this already. It's a race.
  • sfink
    (though I don't know why it presents itself so strangely)
  • sfink
    hm. I guess it's expected an oom error and not getting it.
  • sfink
    *expecting
  • sfink
    and that feels racy
  • sfink
    I guess Aryx intended to needinfo me on the bug
  • jorendorff
    oh
  • jorendorff
    yes, undoubtedly
  • jorendorff
    i think this is higher than P5
  • sfink
    heh
  • sfink
    yeah, it's more of a P1-why-hasn't-this-been-backed-out-or-disabled-yet?
  • sfink
    (if it's really >50%)
  • jorendorff
    naveed: triage opinion: "intermittent orange" bugs that happen often should be P1 or P2
  • naveed
    that is an excellent question
  • jorendorff
    because of the effect of spurious test failures on the whole universe
  • sfink
    you may need to define "often"
  • jorendorff
    like, if i pinged naveed every time it happened, would he be annoyed
  • jorendorff
    that's "often"
  • » naveed is already annoyed
  • naveed
    :)
  • naveed
    kidding
  • jorendorff
    lol
  • sfink
    I probably ought to stop uploading the artifact "CREATED-BY-AUTOSPIDER"
  • sfink
    or at least change it to a scary spider gif for Halloween or something
  • jorendorff
    more seriously: it is ok for these bugs to be high priority because they don't take long -- worst case you disable the test
  • sfink
    hm, this intermittent was filed 3 months ago, then closed a month ago because it was so rare. (bug 1381661)
  • firebot
    bugzil.la/1381661 — REOPENED, nobody⊙mozilla.org — Intermittent js/src/jit-test/tests/gc/bug-1109913.js | Unknown (code 0, args "--baseline-eager")
  • evilpie
    wait, the all hands location changed?
  • Waldo
    yes, because reasons I was briefly told in YVR and have since forgotten
  • Waldo
    travel warning, or something like that
  • jimb
    Waldo: Do you understand the role of the WrappedPtrOperations class?
  • jimb
  • jimb
    I'm wondering why `operator[]` is willing to give you a Handle, but the `back` method won't.
  • Waldo
    that sounds dubious
  • jimb
    "dubious" sounds like something Ella Fitzgerald would say when she forgets the lyrics
  • Waldo
    if I were hip I would make a Mariah Carey reference right now
  • Waldo
    seems like back() could return the same thing, really
  • jimb
    That's what I think, too.
  • pbone
    Morning.
  • RyanVM
    evening
  • pbone
    :-)