mozilla :: #jsapi

17 Jul 2017
00:50docbrownHi folks! Where does the build system define JS_NUNBOX32 or JS_PUNBOX64?
01:27tcampbelldocbrown: I believe this is it: https://searchfox.org/mozilla-central/source/js/src/old-configure.in#883
01:27pbonemorning tcampbell
01:28tcampbellhey pbone!
01:28docbrowntcampbell: I thought that script wasn't used anymore. (Sorry, new to SpiderMonkey)
01:29tcampbelldocbrown: I think one day old-configure will be removed, but it is has been in transition for a while. I think the top-level gecko configure has mostly migrated to python. SpiderMonkey will be a holdout for a while
01:30docbrowntcampbell: Alright, thanks!
01:33tcampbell(also, it's an appropriate question for this channel. depending on timezones and workload questions may go unanswered though)
15:49sfinkyeah, the original configure.in got renamed to old-configure.in but is still used. Everything that could be easily migrated to the new system has been, but old-configure still does everything that is too messy or too much work to do in the new system. (I believe the intention is for it to eventually shrink down to nothingness.)
16:01joncosfink: ping
16:02sfinkjonco: pong
16:02joncosfink: hey, how far did you get with the incremental weak marking patches?
16:03joncolooking at making the gray marking incremental
16:04sfinkit's implemented, most tests pass but many do not, and the implementation feels more complicated than it ought to be
16:04joncosfink: ok
16:04sfinkI'll post my latest patch so you can take a look
16:05joncothanks
16:05sfinktogether with my description of the algorithm and invariants, which ought to be turned into a comment somewhere but I never did that
16:05joncook great, I'll take a look
16:10sfinkok, posted
16:10sfinkbug 1167452
16:10firebothttps://bugzil.la/1167452 ASSIGNED, sphink@gmail.com Incrementalize weak marking phase
16:11joncosfink: cheers!
16:30sfinkjonco: just posted a rebased version (that compiles)
16:35joncook cool
16:35docbrownI'm trying to grok the SpiderMonkey build process... It builds with C++ exceptions enabled, but doesn't use them and then on Windows, it defines _HAS_EXCEPTIONS=0 so that the STL won't throw. Is that correct?
16:35sfinkI thought we compiled with -fno-exceptions on non-Windows
16:36docbrownsfink: That could be. I'm mostly looking at the MSVC bits.
16:36sfinkwhat you describe certainly sounds plausible, but I couldn't tell you for sure
16:37docbrownSo when Visual Studio 2017 warns about "no matching operator delete found; memory will not be freed if initialization throws an exception", I can just ignore that because exceptions won't be thrown anyway?
16:40sfinkbleh, that's funky. https://social.msdn.microsoft.com/Forums/vstudio/en-US/a5b5305d-551c-4b79-badc-2989ac11b8a5/the-usage-of-placement-new-produces-a-c4291-warning?forum=vclanguage seems to suggest that we could make the warning go away by using ::new instead of unqualified new.
16:41sfinkbut it sounds like you're right, it's ignorable in this case. Spidermonkey won't work if something underneath it throws an exception that unwinds through SM code anyway.
16:43sfinkI had never heard of 'placement delete' before
16:47docbrownsfink: thanks for the info
16:48docbrownI'm trying to graft SpiderMonkey into an existing CMake-based build. Fun times. :P
16:49sfinkmay the computing gods have mercy on your poor soul
16:49docbrownlol
16:52anbaDoes anyone know why it costs so much to null the ArrayIteratorSlotIteratedObject slot of ArrayIterators? At least in micro-benchmarks, nulling the slot is about 25% of the time. For example this -benchmark https://pastebin.mozilla.org/9027361 improves from 475ms to 350ms with this patch applied https://pastebin.mozilla.org/9027360.
17:15mrgigglesthe mozilla-inbound tree is now closed (bustage from bug 1364908 )
17:15firebothttps://bugzil.la/1364908 NEW, nobody@mozilla.org Support inlining functions that use arguments[x]
17:47mrgigglesthe mozilla-inbound tree is now open
19:25tcampbellnaveed: have any time today to sync up?
20:18docbrownIs the top-level old-configure.in used at all during a SM build or just the one in js/?
20:27sfinkit appears to not be used, at least as determined by looking at our standalone SM package, which doesn't include it
20:27sfinkbut the toplevel configure.py and moz.build files are used
20:39docbrownsfink: great, thanks
23:22Caspy7curious, how's the outlook for Quantum DOM?
23:27selenamarieis there a tracking bug for "import" anywhere?
23:28selenamarieis it https://bugzilla.mozilla.org/show_bug.cgi?id=1044467 ?
23:28firebotBug 1044467 NEW, nobody@mozilla.org Reflect.parse: Support new `import * as X from "Y"` alternative of `module X from Y`
23:29sfinkI don't think so. I think it would be under https://bugzilla.mozilla.org/showdependencytree.cgi?id=568953&hide_resolved=1
23:29sfinkis that 1342012? that looks like a new dynamic thing
23:30sfinkbug 1342012
23:30firebothttps://bugzil.la/1342012 NEW, nobody@mozilla.org Implement the dynamic import() proposal
23:31sfinkI don't know if the state has progressed past bug 1240072 comment 34
23:31firebothttps://bugzil.la/1240072 NEW, nobody@mozilla.org Implement milestone 0 <script type=&quot;module&quot;>
23:31shuselenamarie: yeah, depends on which import
23:42selenamarieshu: sfink: thank you! dynamic import() was what i was looking for :)
23:43shuselenamarie: it&#39;s stage 3, so we should get a move on, yes
23:43shuselenamarie: is somebody asking for it?
23:44selenamarieshu: was just chatting with joe h. about long term plans. i don&#39;t have an immediate request, and would like to know what i&#39;d be displacing.
23:45selenamarieshu: but i might come back in the next few weeks with a specific ask :)
23:45shuselenamarie: jonco&#39;s the modules guy
23:45selenamariesorry to be vague.. i&#39;m mostly just feeling things out right now
23:45selenamariei saw! :)
23:45shuselenamarie: so in the immediate, it&#39;d most likely displace the ongoing GC work
23:46sfinkshu: do you know if the spec issues with static modules would also get in the way of the dynamic import?
23:46selenamarieshu: that doesn&#39;t sound good to me.
23:46pboneMorning.
23:46pboneshu: glad to see that you&#39;re feeling better.
23:46sfinkgood timeofday
23:46shupbone: thank you
23:47shusfink: what issues?
23:47pbonesfink: I was about to head into the co-working space in the city. Should I hang around here for a little while.
23:47sfinkoh, just last I talked to jonco, he seemed to be waiting on resolution of some weird problems with cyclic modules and errors. Or something. And bug 1240072 comment 34 sounded similar.
23:47firebothttps://bugzil.la/1240072 NEW, nobody@mozilla.org Implement milestone 0 <script type=&quot;module&quot;>
23:48sfinkpbone: I&#39;ll be around for a while
23:48shusfink: i guess they&#39;re orthogonal
23:49sfinkpbone: do you think you could add that pref for max nursery size?
23:49pboneProbably.
23:49shusfink: import() afaik hooks into the same host-specific loading resolution stuff
23:49sfinkit would be nice to be able to experiment with that for speedometer
23:49shusfink: so once jonco&#39;s bugs are fixed, it should be fixed for both kinds of imports?
23:50shu3
23:50shuoops
23:50sfinkshu: I don&#39;t know. I thought the issue was that the spec hadn&#39;t been decided, not that there were bugs.
23:50pbonesfink: You did the profiler marker things for some GCMinor markers? is that right?
23:50sfinkbut I don&#39;t really know what I&#39;m talking about
23:50shusfink: oh, the circularity thing was a bug
23:51sfinkpbone: yes, adding the reason and 3 or 4 other bits of information
23:51pbonesfink: I&#39;ll hang around at home for a while then, and go to the city a little later.
23:51sfinksadly, I also broke startup profilng as an unhappy side effect
23:54sfinkpbone: so I think I&#39;ll request review from you on the fix, so you can learn about the joys of our review system. ;-)
23:55pbonehaha... ...okay.
23:56shusfink: i haven&#39;t been keeping up with the phabricator business
23:56shusfink: are we switching everything to phabricator or something?
23:56sfinkthat looks to still be the current plan
23:56shuheh
23:57sfinkamid much wailing and gnashing of teeth, but we&#39;ll see how it goes
23:58shuare we losing much productivity over splinter
23:58shuthe only problem with it, and it&#39;s an egregious one, is that the review is saved locally
23:58shu(imho)
23:59sfinkthe motivations aren&#39;t entirely clear, but it sounds like some of the impetus is from needing better control over the patches that actually land, and how they&#39;re allowed in, not so much productivity loss due to splinter/mozreview
18 Jul 2017
No messages
   
Last message: 66 days and 8 hours ago