mozilla :: #jsapi

18 May 2017
04:37shujandem-away: ping when you're up
05:02shuunending torments
08:46Yoricshu: I intend to rewrite my code to take as argument an ESTree-like AST specification.
08:47YoricLess hardcoding should be good.
08:59pboneHi team! Can I get a high-5. I made changes, they compile, and the test suite is now running. Things are looking good.
08:59pboneOr maybe high-5 for first commit to SM is when things hit nightly or "try" likes them>
08:59pbone>
08:59pbone?
09:00pboneBTW. what do the 4 columns mean for the JS test suite? green, red, dark green? and grey.
09:47jandem-awayshu: pto until Friday so won't be around much
09:48jandem-awayphone: don't forget the fuzzers :)
09:49jandem-awayoffhand i think these columns are probably pass, fail, timeout, skipped
10:31pbonejandem-away: thanka.
11:04pbonejonco: ping.
11:55joncopbone: oh hey sorry I missed your ping I was at lunch
11:55joncopbone: are you still around?
11:56pboneYeah,
11:56pboneI'm arround. I put stuff in e-mail
11:57pboneI can join you on vidyo.
11:58joncopbone: I'm online now
11:59joncopbone: https://wiki.mozilla.org/JavaScript:SpiderMonkey:C%2B%2B_Coding_Style
12:45evilpiebaku: are you actually going to work on modules?
12:46evilpiepbone: congrats and belated welcome!
12:46bakuevilpie:eh... that was my initial plan. It seems I'm not, right now. At least not in the next month.
12:47evilpie:(
12:55yurytill: hi, just to let you know, pdf.js got gsoc student to convert stuff to streams api. is there a plan to land streams in couple of months?
12:55tillyury: that's great to hear!
12:56yurytill: yeah, it will be nice to see that in action
12:56tillyury: yes, that'll happen. I talked to baku today and we'll have streams support in fetch in that time frame
12:57yuryawesome, it promises less memory and more responsive in pdf.js
13:29nbpyury: is pdf.js still alive?
13:30RyanVMyep
13:30yuryvery much alive
13:30nbpwasn't there a project to replace it with the pdf viewer from chrome?
13:30tcampbellI thought there was a new plan for 57?
13:30yurymortar is acitve, yes
13:31nbpI am not sure to understand the motivation for a gsoc then. >/
13:31yurynbp: streams api rocks!
13:32nbpyury: is that a way to compete between improving pdf.js vs. mortar project?
13:32nbpyury: or testing streams on an existing project to learn more about streams?
13:33yurynbp: for me it's about testing ff's streams implementation, pdf.js is complex enough to surface its imperfections
13:34nbpok.
13:36yuryalso, pdf.js is moving to pure es2015 syntax, so I hope it might help to test https://bugzilla.mozilla.org/show_bug.cgi?id=1167472
13:36firebotBug 1167472 NEW, tcampbell@mozilla.com Make classes work in the JITs.
13:36tcampbellah
13:38yurytcampbell: btw, pdf.js now is used as is in es2015 syntax, but contributors afraid to convert to es6 class due to the bug above
13:38yurytcampbell: https://github.com/mozilla/pdf.js/pull/8119#issuecomment-294128860
13:40tcampbellgood to know. webextensions and devtools have been asking about it too. It really probably is time to have at least baseline support.
13:49joncosfink: ping
13:49pboneevilpie: Thanks :-)
13:54tcampbellevilpie: good point on MConcat, so I made the test case worse
16:43shuhm who knows TI anymore
16:53bzshu: bhackett?
16:53shubz: i should hope so :)
16:53shubz: was hoping for more someone i could talk to synchronously
18:00bzshu: maybe jandem?
18:11nbpshu: Could we have an emptyGlobalScope with a scope kind which is ScopeKind::NonSyntactic? Or I can safely use the fact that the envChain is empty to deduce the scopeKind ahead?
18:13nbpshu: I am trying to split the JS::Evaluate call from dom/base/nsJSUtils.cpp into a JS::Compile or JS::CompileForNonSyntactic & JS_Execute
18:16nbpshu: http://searchfox.org/mozilla-central/source/js/src/vm/GlobalObject.cpp#326-329 I guess this implies that I can use the envChain length to deduce the scopeKind ahead.
18:29shunbp: you can, yes
18:30shunbp: http://searchfox.org/mozilla-central/source/js/src/jsapi.cpp#3527
18:30jandem-awayshu: yeah happy to talk about that when I'm around
18:32shujandem-away: ok i'm gonna try to fix it in the meantime
18:42leobaltershu: I'm looking forward for your feedback on that email. I'm want to make sure I'm prioritizing the right thing.
18:43shuon my list today!
18:43shushould do that before i go back down that jit bug rabbithole
18:59djvjshu: what about TI?
19:00shudjvj: not at xo
19:00shuerr
19:00shunot at computer right now
19:01shudjvj: but i found that group->maybeInterpretedFunction()->group() doesn't round trip
19:01shudjvj: not sure if bug or by design
19:03djvjshu: so the group is a singleton group for a function?
19:03djvjshu: or the singleton group for the instances of a function?
19:03djvjer.. second question shouldn't include "singleton"
19:07djvjshu: so it's the former.. group is the singleton group for a function.
19:07djvjshu: seems like a bug, actually
19:07djvjshu: see: http://searchfox.org/mozilla-central/source/js/src/vm/ObjectGroup.cpp#339
19:08djvjthe code for the group creation sets obj->group_ = group; after creating group->setInterpretedFunction(obj)
19:09shudjvj: what is a singleton group? like a lazy thing? not that
19:10djvjshu: yeah, lazy groups are singleton groups
19:10djvjshu: basically, idea being that if you have a singleton group, you don't have to actually materialize it until you need to "do something" with it.
19:10djvjyou can track the group in typesets by just using the object pointer.
19:10shuyes i'm familiar
19:11shunot that case
19:11djvjoh, what do you mean then?
19:26shudjvj: the nonsingleton case
19:26shudjvj: like the regular case, a group of objects that share a canonical interpreted fun
19:28djvjshu: what about that case?
19:29djvjshu: that case won't use maybeInterpretedFunction()
19:29shudjvj: it does!
19:30shudjvj: for poly inline dispatch
19:30djvjshu: when you say a group of objects that share a canonical interpreted fun
19:30djvjshu: I assumed you meant a group of instance objects that share a constructor function
19:30shudjvj: no, i mean an ObjectGroup
19:31shuin the TI sense of group
19:31djvjshu: right, an ObjectGroup describing instances created for a function F using "new F()"
19:31djvjshu: ?
19:32djvjthere's two things here. There's the ObjectGroup for the function object itself, and the ObjectGroup for the instances of the function created using 'new'. I'm not sure which you're referring to.
19:34shudjvj: the ObjectGroup of the functions themselves
19:34shudjvj: not for instances if used as constructor
19:34djvjah, right.
19:35shudjvj: i'm looking at a bug where the poly inline dispatch is dispatching the wrong functions
19:35djvjyeah, in that case, group == group->maybeInterpretedFunction()->group() should hold.
19:35djvjprobably a bug if it doesn't..
19:35djvjah..
19:35evilpiewe just started using this in more cases with bug 1354582
19:35firebothttps://bugzil.la/1354582 FIXED, jdemooij@mozilla.com call/apply inlining should support non-singleton callees
19:36evilpieI just started wondering how this interacts with arguments.caller though
19:40djvjevilpie: we'd have to pass in the actual function instance in the args
19:41djvjevilpie: which is fine, because the callee is passed in the frame header
19:41evilpieah that makes sense
19:43evilpiebut we don't do this when inlining
19:44djvjah, yeah. We'd have to push that in as a MDefinition that's used by arguments.caller prop-get.
19:47djvjevilpie: in fact, I think that's what we do.
19:47djvjevilpie: I wrote the first polymorphic inlining impl in Ion
19:48evilpieI wish this stuff had actually comments that explained what it is ..
20:36kitcambridgehi all, i see nightly supports https://github.com/tc39/proposal-async-iteration. is it OK to use that in JSMs?
20:39araino
20:40araiit's nightly-only
20:40araiand also web-content only
20:40araiso it's not available in browser code
20:43araiit's just for testing the spec
20:43araiand the spec is about to change, iiuc
20:43arai(the semantics part
20:46leobalterkitcambridge: what arai just said. You might use it for experimentations, at least, knowing they might stop working soon for semantics fixes.
20:46kitcambridgeleobalter: arai: thanks!
21:12mrgigglesthe mozilla-inbound tree is now closed (windows bustage from bug 1180561)
21:12firebothttps://bugzil.la/1180561 REOPENED, dmajor@mozilla.com startup crash in [@ msmpeg2vdec.dll@0x25a3fb ]
21:18mrgigglesthe mozilla-inbound tree is now open
21:21mrgigglesthe mozilla-inbound tree is now closed (bustage from bug 1365940)
21:21firebothttps://bugzil.la/1365940 NEW, amarchesini@mozilla.com struct WindowAction in WorkerPrivate.cpp should be marked MOZ_RAII
21:31mrgigglesthe mozilla-inbound tree is now open
23:38gkwshu: ping re: landing fuzzblocker bug 1364573 ?
23:38firebothttps://bugzil.la/1364573 NEW, shu@mozilla.com Assertion failure: !fun->infallibleIsDefaultClassConstructor(cx) || fun->compartment()->behaviors().
23:39shugkw: all right let me land a test262 sync first, i'll do those after
23:42gkwshu: thanks!
19 May 2017
No messages
   
Last message: 6 days and 7 hours ago