mozilla :: #content

12 Sep 2017
09:29emiliosmaug: ping (hopefully the last time I ping you about that bag, promise)
09:30emilio(there's no rush either)
09:40smaug emilio: pong
09:40emiliosmaug: just submitted a review request instead
09:40emiliosmaug: we were leaking chrome-only nodes to the add-on test
09:41emiliosmaug: I don't know if that's a security concern or not
09:41emiliosmaug: but should be fixed with my patch
12:37nelsonhi
14:10bkellyhmm, more ghost windows
14:43bkellyI have steps to reproduce this leak if anyone wants to look at it: https://bugzilla.mozilla.org/show_bug.cgi?id=1397062#c14
14:43firebotBug 1397062 NEW, nobody@mozilla.org many ghost windows in nightly 57
17:11smaugemilio: so you use IsChrome when other code uses ChromeOnlyContent
17:11smaugtotally different things
17:12emiliosmaug: hmmm... You may be right
17:13emiliosmaug: I'm out of my computer now, will address your comments as soon as I'm back, thanks a lot for taking a look
17:13smaugok. If my queue is closed, just ping me and I'll take another look.
18:02mccr8bkelly: when I try your steps to reproduce in a debug build, I just get a permahang. :-/
18:02bkellymccr8: that seems even worse
18:04mccr8bkelly: by "back to the main page", do you mean clicking the back button in Firefox, or clicking on the link back to the main page on the article?
18:04bkellymccr8: I was clicking the back button
18:25bkellymccr8: oh, I missed the hang was in a debug build
18:46mccr8fortunately these ghost window STR seem to be pretty easy, so bisecting isn't too much of a nightmare
18:48bkellymccr8: yea... its probably the best strategy for a "unknown extra ref" thing
18:57smaugmystor: in hangs.html, the timeline, what does the lighter blue area mean?
18:57smaugwell, I don't understand the darker iether
18:57smaugeither
18:59smaugmystor: ha
19:00mystordthayer: <smaug> mystor: in hangs.html, the timeline, what does the lighter blue area mean?
19:00mystor<smaug> well, I don&#39;t understand the darker iether
19:00mystor<smaug> either
19:00smaugdthayer: want to explain hangs.html UI a bit :)
19:00mystordthayer: I figured you might be able to explain it better than me ^.^
19:01dthayerheh, yeah - and I&#39;ll happily accept any suggestions on how to make it more self-explanatory
19:02dthayerthe light area is the number of hangs prefixed by the focused node in the tree on that date, normalized by the usage hours we collected for that date
19:03dthayerso if you have (root) > RtlUserThreadStart > BaseThreadInitThunk selected, the light area will tell you how often hangs with that stack (or descendant stacks) occurred on that day
19:03smaughmm, now I don&#39;t see any timeline for content process
19:04dthayerdo you have a node selected?
19:04smaugah, I didn&#39;t
19:05smaugdthayer: so darker blue is then what?
19:06smaugdthayer: another question, is there some way to track whether the overall hangs have been reduced over time, say couple of months
19:06dthayerthe dark area, though, is the number of milliseconds actually hanging with that stack on that day. since we lump all hangs > 2s together, some of those hangs can be longer or shorter than others - the dark area is supposed to be an indicator of this
19:06smaugor at least couple of weeks
19:08* smaug thinks certain hangs are way rarer these days than just two weeks ago, but hard to verify
19:09dthayercurrently no, but so long as we&#39;re just interested in overall hangs i can write a separate spark job for that
19:09smaugand don&#39;t know whether overall number of hangs has gone down
19:10smaugdthayer: another option perhaps is to be able to see old data in hangs.html
19:11smaugjust to see how often certain things happened then
19:13dthayersmaug: yeah - i&#39;m working on performance issues right now to try to get a larger period. a month is probably as far as we&#39;re going to get though without sacrificing any detail
19:14dthayersmaug: we could potentially keep a list of stacks that we want to track over multiple months, and just run a separate spark job for that list
19:15smaugdthayer: I&#39;d be happy with any solution which let&#39;s one to somehow track the changes
19:15smaugright now I have no idea whether we&#39;re behaving better or worse than a month ago
19:16smaugI may check some telemetry probes to get some hint
19:16smaugbut I don&#39;t always know which ones
19:17dthayersmaug: another problem that we face right now is that we switched the BHR ping format on August 16th, and added interleaved pseudo stacks right around there, so tracking before that date is hard, and the quality of data is worse
19:17smaugah
19:20smaugdthayer: in general I&#39;m quite happy with the stack view
19:20dthayersmaug: thanks! :)
20:08mccr8bkelly: bisection complete! woooo
20:08bkellymccr8: oh?
20:08bkellyis it my fault?
20:09mccr8hah. no.
20:09mccr8https://bugzilla.mozilla.org/show_bug.cgi?id=1358060
20:09firebotBug 1358060 FIXED, honzab.moz@firemni.cz Postpone tracker requests processing after non-trackers
20:10bkellymccr8: I wonder if bug 1398671 fixes it
20:11firebothttps://bugzil.la/1398671 NEW, kechang@mozilla.com Call AsyncAbort when a tail request is canceled
20:11bkellymccr8: does it reproduce with one of the builds here? https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=a8d001a6ee41840e38414a313579a2b0c2799300
20:11mccr8great, both people involved are on PTO. :P
20:12bkellymccr8: someone is working on it... they just pushed something today
20:12mccr8bkelly: cool
20:12mccr8&quot;When a tail request is canceled at [1], we have to call AsyncAbort to avoid memory leaks.&quot; yeah that&#39;s believable.
20:13bkellymccr8: how will you feel if that is the fix... you could have just not done the bisect since it landed in inbound 2 hours ago :-\
20:13bkellyI guess its good to know
20:14mccr8bkelly: hah. that&#39;s fine! I&#39;d much rather have wasted an hour or two for something that was fixed than to have something not fixed.
20:18jdmmayo is smiling beatifically right now and doesn&#39;t even know why
20:19froydnjthe bug bringing down 57 has already been filed
20:24bzHas anyone seen peterv?
20:27mccr8he posted on an XSLT bug earlier today. :)
20:27bzmmmhm
20:27bzWhy is he not reviewing my patches, then? ;)
20:27mccr8I often ask myself that.
20:29bkellymaybe the XSLT blinded him
20:30bzstun damage, you think?
20:30bkellydoes mercurial have a way to select specific hunks within a patch when doing a qref?
20:30bkellyI want to split a patch with multiple changes to the same file
20:30bzbkelly: qref -i
20:30bzbkelly: interactive mode
20:31* bkelly does not see interactive mode in hg help qref.
20:31bkellymaybe its a plugin or something
20:31bzhmm
20:31* bz has it...
20:31bzone sec
20:32bzstill have it if I turn off mqext
20:32mccr8you should use git /unhelpful
20:33bzWhat version of hg?
20:33bkelly3.7.3
20:33* bz has 4.2.3; maybe that&#39;s why...
20:34bzAnyway, if you have that option then for uncommitted changes you can use it
20:36bkellybz: what hg plugins do you have?
20:36emiliosmaug: thanks a lot for the reviews :)
20:36bzbkelly: um... a bunch
20:36bkellybz: do you have Record?
20:37bzyes!
20:37bzhttps://pastebin.mozilla.org/9032259 fwiw
20:37bzyes
20:37bzrecord is the one that adds that option
20:38bkellynow to figure out how to install a mercurial extension
20:38bzrecord is a built-in extension
20:39bkellybz: magic
20:39bzJust add &quot;record = &quot; to the [extensions] bit in your hgrc
20:39bkelly -i --interactive interactively select changes to refresh
20:39bkellybz: thanks!
20:39bzNo problem
21:11oddcoderHi I was curios is there good start guide for learning About DOM
21:11oddcoderI want to get their implementation, but all I know for know is that they exist
21:11oddcoderso if there is a guide for end user, then a guide for developer it will be cool!
21:12oddcoderfor now I started reading this
21:12oddcoderhttps://developer.mozilla.org/en-US/docs/Web/API/Document_Object_Model/Introduction
21:27bzoddcoder: Do you mean &quot;DOM&quot; in the broad sense of &quot;APIs browsers expose&quot;, or in the narrow sense of &quot;the forest of Node objects&quot;?
21:28oddcoderAPIs browser exposes
21:28oddcoderI am not sure what is the other thing
21:31bzthe other thing is the things representing elements and textnodes
21:31bzFor the general &quot;APIs browser exposes&quot;....
21:32bzThe implementation really depends on the API
21:32bzFor the Gecko implementation, https://developer.mozilla.org/en-US/docs/Mozilla/WebIDL_bindings might be relevant in terms of describing the mechanism that goes from JS to the underlying C++ implementation
21:33bzThen you can use http://searchfox.org/ to look at code
21:48oddcoderthanks
13 Sep 2017
No messages
   
Last message: 7 days and 4 hours ago