/irc-logs / mozilla / #accessibility / 2015-01-28 / end
Options:
Previous day, Next day
- # Session Start: Wed Jan 28 00:00:00 2015
- # Session Ident: #accessibility
- # [00:03] * Quits: maxli (maxli@moz-fi7.g4b.97.129.IP) (Quit: Leaving.)
- # [00:11] * Quits: newtron (newtron@moz-6lh.c31.71.199.IP) (Ping timeout: 121 seconds)
- # [00:15] * Quits: @surkov (surkov@moz-v8g78c.cpe.teksavvy.com) (Client exited)
- # [00:15] * Joins: maxli (maxli@moz-h8kluf.cs.uwaterloo.ca)
- # [00:22] * Quits: maxli (maxli@moz-h8kluf.cs.uwaterloo.ca) (Quit: Leaving.)
- # [00:22] * Joins: maxli (maxli@moz-h8kluf.cs.uwaterloo.ca)
- # [01:18] * Quits: maxli (maxli@moz-h8kluf.cs.uwaterloo.ca) (Quit: Leaving.)
- # [02:52] * Quits: satdav (uid15780@moz-0t6277.uxbridge.irccloud.com) (Quit: Connection closed for inactivity)
- # [02:56] * Quits: Justin_o (uid14648@moz-c6vipl.ealing.irccloud.com) (Quit: Connection closed for inactivity)
- # [03:03] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [03:04] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [03:05] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [03:05] * Quits: Gijs (chatzilla@moz-8e8.cgs.80.5.IP) (Quit: sleep)
- # [03:08] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Connection closed)
- # [03:08] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [03:09] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [03:09] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [03:19] * Joins: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca)
- # [03:26] <@firebot> New Core - Disability Access APIs bug 1126649 filed by dholbert@mozilla.com.
- # [03:26] <@firebot> https://bugzil.la/1126649 — NEW, nobody@mozilla.org — style-system mochitest test_pseudoelement_state.html triggers piles of "ASSERTION: We must reach doc
- # [04:05] * Quits: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca) (A TLS packet with unexpected length was received.)
- # [04:05] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Connection closed)
- # [04:06] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [04:06] * Joins: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca)
- # [04:07] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [04:07] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [05:42] * Joins: newtron (newtron@moz-epuckk.cpe.teksavvy.com)
- # [06:37] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Connection closed)
- # [06:37] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [06:39] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [06:39] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [06:41] * icaaq|afk is now known as icaaq_
- # [07:05] * Joins: ioanachiorean (ioanachiore@moz-ohb.65e.79.82.IP)
- # [07:24] * Quits: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca) (Quit: Leaving.)
- # [07:52] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Connection closed)
- # [07:53] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [07:57] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [08:01] * Tomcat|afk is now known as Tomcat|sheriffduty
- # [08:02] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [08:47] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [08:48] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [08:51] * Joins: yliao_ (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [08:51] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [09:34] * Joins: MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de)
- # [09:34] * ChanServ sets mode: +ao MarcoZ MarcoZ
- # [09:55] <@firebot> mzehe@mozilla.com requested needinfo from gijskruitbosch+bugs@gmail.com on bug 1126516.
- # [09:55] <@firebot> https://bugzil.la/1126516 — UNCONFIRMED, nobody@mozilla.org — Setting accessibility.blockautorefresh to true blocks login to impots.gouv.fr
- # [10:09] * Quits: &MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de) (Quit: l8er)
- # [10:24] * Joins: Gijs (chatzilla@moz-8e8.cgs.80.5.IP)
- # [10:24] * Joins: slee (chatzilla@moz-gatgc9.range109-157.btcentralplus.com)
- # [11:01] <@firebot> gijskruitbosch+bugs@gmail.com requested needinfo from vincent-moz@vinc17.net on bug 1126516.
- # [11:01] <@firebot> https://bugzil.la/1126516 — UNCONFIRMED, nobody@mozilla.org — Setting accessibility.blockautorefresh to true blocks login to impots.gouv.fr
- # [11:07] * Joins: agibson (agibson@moz-j04gi9.cable.virginm.net)
- # [11:30] * Joins: marcoz (Adium@moz-74kv4p.dip0.t-ipconnect.de)
- # [11:30] * ChanServ sets mode: +ao marcoz marcoz
- # [11:55] <@firebot> vincent-moz@vinc17.net cancelled needinfo?(vincent-moz@vinc17.net) on bug 1126516.
- # [11:55] <@firebot> https://bugzil.la/1126516 — UNCONFIRMED, nobody@mozilla.org — Setting accessibility.blockautorefresh to true blocks login to impots.gouv.fr
- # [11:55] * Quits: yliao_ (yliao@moz-2hf62a.HINET-IP.hinet.net) (Connection closed)
- # [11:56] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [12:00] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [12:25] * Quits: &marcoz (Adium@moz-74kv4p.dip0.t-ipconnect.de) (Quit: Leaving.)
- # [12:26] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [12:31] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [12:36] * Joins: MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de)
- # [12:36] * ChanServ sets mode: +ao MarcoZ MarcoZ
- # [12:53] * Joins: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net)
- # [12:55] * icaaq_ is now known as icaaq|afk
- # [12:57] * icaaq|afk is now known as icaaq_
- # [12:59] * Quits: yliao (yliao@moz-aj9ejj.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [13:01] * agibson is now known as agibson|afk
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com changed the Resolution on bug 1126516 from --- to DUPLICATE.
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com changed the Status on bug 1126516 from UNCONFIRMED to RESOLVED.
- # [13:02] <@firebot> https://bugzil.la/1126516 — DUPLICATE, nobody@mozilla.org — Setting accessibility.blockautorefresh to true blocks login to impots.gouv.fr
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com requested in-testsuite from the wind on bug 1070713.
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com denied qe-verify on bug 1070713.
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com granted firefox-backlog on bug 1070713.
- # [13:02] <@firebot> gijskruitbosch+bugs@gmail.com changed the Status on bug 1070713 from UNCONFIRMED to NEW.
- # [13:02] <Gijs> MarcoZ: that pref is annoying. :(
- # [13:11] <&MarcoZ> Gijs: I know. But when these pages were still modern, this was the only way for screen reader users to have a chance to read the contents at all.
- # [13:11] <&MarcoZ> This is for those pages or framesets were a page was really reloading every few seconds, causing the screen reader to completely lose and re-render the contents.
- # [13:13] <Gijs> MarcoZ: right, but paranoid people (not saying that particular bugreporter, but I've seen a few) are now using it to get manual control over redirects and get a little uppity about it.
- # [13:15] <&MarcoZ> Not sure what we can do.
- # [13:15] <&MarcoZ> In the case of the duplicate bug, the user was really unable to use the site because of this.
- # [13:16] * &MarcoZ is going for lunch.
- # [13:22] <Gijs> MarcoZ: enjoy lunch; when you get back: is there still a good reason to keep this preference from an accessibility PoV?
- # [13:58] * Joins: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca)
- # [13:59] * Quits: newtron (newtron@moz-epuckk.cpe.teksavvy.com) (Connection closed)
- # [14:23] * agibson|afk is now known as agibson
- # [14:26] * Joins: Justin_o (uid14648@moz-c6vipl.ealing.irccloud.com)
- # [14:36] * Joins: yzen (yzen@moz-ql1r6k.cpe.pppoe.ca)
- # [14:36] * ChanServ sets mode: +o yzen
- # [14:49] * Joins: newtron (newtron@moz-6lh.c31.71.199.IP)
- # [14:52] * Joins: anvk (anovak@moz-871.mvk.11.204.IP)
- # [15:04] * yzen is now known as yzen_
- # [15:05] <&MarcoZ> Gijs: Hhmmm I am unsure. I rarely see these refresh warnings myself any more. What I do see are the automatic redirect warnings, e. g. when opening a link from a mail within Gmail, for example. But refresh warnings are very rare nowadays.
- # [15:06] <Gijs> MarcoZ: are you sure those warnings aren't done by google?
- # [15:06] <Gijs> MarcoZ: Firefox's own refresh/redirect warnings are controlled by the same pref, though
- # [15:07] <Gijs> MarcoZ: do you even have that pref turned on? :)
- # [15:07] <Gijs> MarcoZ: (I guess my real question is, do you have it turned on and if so, why - do a11y tools still cope poorly with it?)
- # [15:08] <&MarcoZ> Gijs: Well, I do get a doorhanger warning about the automatic redirecting. So it's something done by Firefox, not Google.
- # [15:08] <&MarcoZ> Doorhanger or Info bar or whatever.
- # [15:08] * Quits: @yzen_ (yzen@moz-ql1r6k.cpe.pppoe.ca) (Ping timeout: 121 seconds)
- # [15:08] <&MarcoZ> It says "Firefox prevented this site from automatically redirecting..."
- # [15:09] <&MarcoZ> Gijs: And the *refresh* thingie is different, that is about loading the same page over and over again. That is something that, to a screen reader, always looks like a whole new page that loaded. And there's nothing that can be done about that particular scenario really.
- # [15:10] <Gijs> MarcoZ: different it may be, the pref that controls it is the same, I'm pretty sure
- # [15:10] * Quits: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca) (Quit: Leaving.)
- # [15:10] * Joins: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca)
- # [15:10] <Gijs> MarcoZ: so you get the automatic redirecting doorhanger because of that pref.
- # [15:11] <Gijs> MarcoZ: is there an accessibility reason to have the redirect warnings?
- # [15:12] <&MarcoZ> Gijs: To be honest, I do not know. This was there long before I joined Mozilla. And I don't even remember ever having touched that pref. Yet when I look in about:config, it says the pref is user-set to true, which leads me to believe the default may be false.
- # [15:12] <Gijs> MarcoZ: indeed, the default is false :)
- # [15:12] <&MarcoZ> So if I never touched it, why is it set to true for me?
- # [15:12] <Gijs> MarcoZ: I don't know :)
- # [15:13] <Gijs> MarcoZ: there is UI for the pref, in the options pane...
- # [15:13] <Gijs> MarcoZ: "Warn me when websites try to redirect or reload the page"
- # [15:13] <Gijs> (Advanced > General)
- # [15:13] <&MarcoZ> Gijs: Yes I know, but whenever I looked at it, it was set to block the redirects/refreshes.
- # [15:13] <Gijs> hmm
- # [15:13] <Gijs> unsure!
- # [15:14] <&MarcoZ> Gijs: This was implemented long before I joined, so I don't know the history of it. Aaron Leventhal was in charge of this back then.
- # [15:14] <Gijs> MarcoZ: fair. I'll check with davidb when he's here, I guess?
- # [15:14] <&MarcoZ> Gijs: There may be some auto-detection going on, e. g. turn this pref to True once we detect that something invokes Accessibility.
- # [15:14] <Gijs> MarcoZ: right.
- # [15:16] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [15:20] <@firebot> New Core - Disability Access APIs bug 1126786 filed by surkov.alexander@gmail.com.
- # [15:20] <@firebot> https://bugzil.la/1126786 — NEW, surkov.alexander@gmail.com — remove NS_DEBUG_X rudiment
- # [15:21] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [15:31] * Joins: yzen (yzen@moz-i5m.05u.207.66.IP)
- # [15:31] * ChanServ sets mode: +o yzen
- # [15:32] * yzen is now known as yzen_
- # [15:32] * yzen_ is now known as yzen
- # [15:32] * Quits: anvk (anovak@moz-871.mvk.11.204.IP) (Connection closed)
- # [15:51] * Joins: davidb (davidb@moz-i5m.05u.207.66.IP)
- # [15:51] * ChanServ sets mode: +qo davidb davidb
- # [15:51] <~davidb> heyo!
- # [15:52] <&MarcoZ> Heyo davidb!
- # [15:52] <~davidb> hi hi
- # [16:05] <~davidb> back in about 30
- # [16:05] * davidb is now known as davidb|afk
- # [16:21] * Joins: clown (clown@moz-v6m.c3o.211.205.IP)
- # [16:23] <@firebot> florin.mezei@softvisioninc.eu denied qe-verify on bug 1123163.
- # [16:23] <@firebot> https://bugzil.la/1123163 — FIXED, surkov.alexander@gmail.com — Hit testing broken on Google Search results page
- # [16:24] * davidb|afk is now known as davidb
- # [16:31] * Joins: anvk (anovak@moz-871.mvk.11.204.IP)
- # [16:51] <&MarcoZ> davidb: Gijs and I had a question re the pref accessibility.blockautorefresh. The default is false, but it is set to true for some users automatically. At least it is for me, and I was wondering if you know/remember the history about that.
- # [16:51] <~davidb> i don't recall
- # [16:51] <&MarcoZ> davidb: I at least never consciously set that pref to true.
- # [16:51] <Gijs> MarcoZ: I do not. I looked in MXR, and I don't see anything setting it automatically. :s
- # [16:51] <&MarcoZ> davidb: Ah OK...
- # [16:52] * ~davidb summons surkov
- # [16:52] * Joins: surkov (surkov@moz-i5m.05u.207.66.IP)
- # [16:52] * ChanServ sets mode: +o surkov
- # [16:52] <&MarcoZ> Gijs: Interesting. I don't recall having ever set that pref to true in any profile.
- # [16:52] <~davidb> surkov, do you recall "accessibility.blockautorefresh" and why it defaults to false?
- # [16:52] <Gijs> MarcoZ: it's possible that earlier versions did or that external software did, or (respectfully) that you did it a very very long time ago and forgot :)
- # [16:52] * Gijs found, for instance, that somehow his windows profile didn't have DNT set to on, which was a surprise
- # [16:53] <@surkov> davidb: I don’t remeber that pref at all :) what’s it about?
- # [16:53] <~davidb> ditto
- # [16:53] <&MarcoZ> davidb: surkov: That is a pref that controls whether Firefox allows automatically refreshing or reloading pages, or redirecting to different pages.
- # [16:53] <Gijs> it blocks some types of automatic redirecting/refreshing
- # [16:53] <&MarcoZ> davidb: surkov: Pref has UI in Tools/Options/Advanced/General tab, under Accessibility.
- # [16:54] <@surkov> bug 520431
- # [16:54] <@firebot> https://bugzil.la/520431 — INCOMPLETE, nobody@mozilla.org — Can't stop javascript:location.href redirects - accessibility.blockautorefresh ineffective
- # [16:54] <Gijs> from bugzilla, at least, it seems more people are trying to use it to control websites that do things that "just" annoy them, than people who have a genuine a11y need
- # [16:55] <@surkov> actually bug 83265
- # [16:55] <@firebot> https://bugzil.la/83265 — FIXED, pilgrim@gmail.com — UAAG: Add a way to disable HTTP-EQUIV=refresh (block automatic meta redirection, lock on current pag
- # [16:57] <@surkov> Gijs: I’ve got same impression
- # [16:58] <@surkov> but in general it feels like this pref is a big simplification of the things
- # [16:58] <Gijs> yeah, it's near impossible to make work right anyway
- # [16:59] <@surkov> is there data how many people use it?
- # [16:59] <&MarcoZ> Gijs: I've now disabled that checkbox, and about:config now shows it as "default false", too.
- # [16:59] <&MarcoZ> Gijs: I will see if it somehow gets reenabled upon my next Firefox launch or so.
- # [17:00] <Gijs> if I do setTimeout(() => location.href = location.href + "&plonk", 5 * 60000), then that's "evil", but if I set location.href to something else when I click a link, it's not...
- # [17:00] <Gijs> it's like popup blocking, but its own massive can of worms
- # [17:00] <Gijs> probably bigger than popup blocking because this doesn't involve opening a new window at all
- # [17:00] <Gijs> just navigation
- # [17:01] <Gijs> MarcoZ: cool, thanks
- # [17:01] <Gijs> surkov: I doubt it... we could try to find out
- # [17:01] <&MarcoZ> Actually let me try that right now. brb!
- # [17:02] * Quits: &MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de) (Quit: brb)
- # [17:03] * Joins: MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de)
- # [17:03] * ChanServ sets mode: +ao MarcoZ MarcoZ
- # [17:03] <&MarcoZ> Gijs: Yes, this is definitely being set to true again for me upon the next Firefox restart.
- # [17:04] <&MarcoZ> I just closed Firefox and relaunched it, and the pref was set to true again without me doing anything.
- # [17:05] * Tomcat|sheriffduty is now known as Tomcat
- # [17:05] * &MarcoZ suspects something under accessible/
- # [17:09] <&MarcoZ> Nope.
- # [17:10] <Gijs> MarcoZ: if you create a new Firefox profile, is it set to true?
- # [17:10] <&MarcoZ> Gijs: Let me try. Need to log off because of not having set -no-remote
- # [17:10] * Quits: &MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de) (Quit: l8er)
- # [17:15] * Joins: MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de)
- # [17:15] * ChanServ sets mode: +ao MarcoZ MarcoZ
- # [17:15] <&MarcoZ> Gijs: No it is not. In a new profile, it is false and stays so. In my main profile, however, even when I set it to false, upon next relaunch, it is set to true.
- # [17:15] <&MarcoZ> And true is shown as user-set, not default.
- # [17:24] <Gijs> MarcoZ: hmm... any add-ons installed?
- # [17:24] * Quits: ~davidb (davidb@moz-i5m.05u.207.66.IP) (A TLS packet with unexpected length was received.)
- # [17:25] <&MarcoZ> Gijs: Yes, sure, but none that I know of that could enable this. Unless Adblock does it.
- # [17:28] <Gijs> MarcoZ: I don't know. Can you check that your prefs are being saved correctly otherwise?
- # [17:28] * Joins: davidb (davidb@moz-i5m.05u.207.66.IP)
- # [17:28] * ChanServ sets mode: +qo davidb davidb
- # [17:28] <&MarcoZ> Gijs: They definitely are. Let me try something else. brb.
- # [17:28] * Quits: &MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de) (Quit: brb)
- # [17:29] * icaaq_ is now known as icaaq|afk
- # [17:32] * Joins: MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de)
- # [17:32] * ChanServ sets mode: +ao MarcoZ MarcoZ
- # [17:32] * Quits: ~davidb (davidb@moz-i5m.05u.207.66.IP) (Quit: peace out)
- # [17:32] <&MarcoZ> Gijs: Weird weird weird... It seems kind of erratic when this pref gets reset to true and when not. Sometimes after a restart, it stays off, sometimes it comes back on, and the add-ons don't seem to affect that.
- # [17:34] <Gijs> MarcoZ: that's... very strange. :(
- # [17:37] <&MarcoZ> Gijs: I'm looking through the Mozilla sources to see if there are other places than the few instances under browser/ where this pref is touched.
- # [17:37] <&MarcoZ> (uses a local copy to do a grep)
- # [17:40] * Joins: davidb (davidb@moz-i5m.05u.207.66.IP)
- # [17:40] * ChanServ sets mode: +qo davidb davidb
- # [17:43] <Gijs> MarcoZ: I'm pretty sure there aren't, but I could be wrong...
- # [17:43] * icaaq|afk is now known as icaaq_
- # [17:46] <~davidb> surkov, yzen, tbsaunde, FWIW I might not be able to join for lunch
- # [17:46] <@yzen> ok
- # [17:50] <&MarcoZ> Gijs: There are references in docshell/base/nsDocShell.cpp and toolkit/mozapps/extensions/nsBlockListService.js.
- # [17:51] <tbsaunde> davidb: ack
- # [17:51] <Gijs> MarcoZ: err, really? I don't see them. What lines?
- # [17:51] <Gijs> (link?)
- # [17:52] <Gijs> MarcoZ: find in page finds nothing for "blockautorefresh" in http://mxr.mozilla.org/mozilla-central/source/docshell/base/nsDocShell.cpp
- # [17:53] <Gijs> MarcoZ: nsBlockListService.js is a comment...
- # [17:53] <&MarcoZ> Gijs: Oh right, sorry, I was lazy and only looked for accessibility.block.
- # [17:54] <&MarcoZ> Gijs: But I just looked, and both instances are inconclusive, the one in nsDocShell.cpp checks for a pref I had never seen before, accessibility.blockjsredirection.
- # [17:57] <~davidb> ok back in about 45 mins
- # [17:57] * davidb is now known as davidb|afk
- # [17:57] * &MarcoZ has reached the end of his work day and will be going offline now, maybe watch the privacy talk later.
- # [17:58] * Quits: &MarcoZ (marco.zehe@moz-74kv4p.dip0.t-ipconnect.de) (Quit: Nightynight!)
- # [17:59] * clown is now known as clown_mtg
- # [18:12] * Joins: satdav (uid15780@moz-0t6277.uxbridge.irccloud.com)
- # [18:43] * davidb|afk is now known as davidb
- # [18:45] <@firebot> dholbert@mozilla.com denied in-testsuite on bug 1126541.
- # [18:45] <@firebot> https://bugzil.la/1126541 — ASSIGNED, dholbert@mozilla.com — fix implicit uint32_t to bool conversion in DocAccessibleParent::RecvShow return statement
- # [18:47] * icaaq_ is now known as icaaq|afk
- # [18:53] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [18:57] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # [18:57] * Quits: ioanachiorean (ioanachiore@moz-ohb.65e.79.82.IP) (Ping timeout: 121 seconds)
- # [19:00] * Gijs is now known as Gijs_away
- # [19:03] * clown_mtg is now known as clown
- # [19:33] * agibson is now known as agibson|afk
- # [19:34] * davidb is now known as davidb|afk
- # [19:37] * davidb|afk is now known as davidb
- # [19:39] <@eeejay> yzen: could you meet earlier today?
- # [19:39] <@yzen> any time
- # [19:39] <@eeejay> yzen: cool. i'll find a room
- # [19:41] * Quits: maxli (maxli@moz-fs8bmd.cs.uwaterloo.ca) (Quit: Leaving.)
- # [19:57] * Joins: maxli (maxli@moz-favcmk.cs.uwaterloo.ca)
- # [20:08] * Quits: agibson|afk (agibson@moz-j04gi9.cable.virginm.net) (Quit: )
- # [20:34] * davidb is now known as davidb|afk
- # [21:02] * Quits: slee (chatzilla@moz-gatgc9.range109-157.btcentralplus.com) (Quit: ChatZilla 0.9.91.1 [Firefox 35.0/20150108202552])
- # [21:08] * davidb|afk is now known as davidb
- # [21:20] * Quits: maxli (maxli@moz-favcmk.cs.uwaterloo.ca) (Quit: Leaving.)
- # [21:22] * Gijs_away is now known as Gijs
- # [21:32] * Joins: maxli (maxli@moz-favcmk.cs.uwaterloo.ca)
- # [21:51] * Joins: smaug (chatzilla@moz-7dv8tf.bb.dnainternet.fi)
- # [21:51] <smaug> hmm, accessibility.force_disabled doesn't work the way I thought it would
- # [21:52] <smaug> I thought setting it to -1 would enable a11y
- # [21:52] <~davidb> ah no
- # [21:52] <~davidb> would be handy
- # [21:52] <~davidb> smaug are you on linux?
- # [21:53] <smaug> yes
- # [21:53] <~davidb> hmmm
- # [21:53] <smaug> we don't seem to actually use ePlatformIsForceEnabled for anything
- # [21:53] <~davidb> amazing
- # [21:53] <smaug> I guess I want GNOME_ACCESSIBILITY=1
- # [21:54] <smaug> env variable
- # [21:54] <~davidb> yeah
- # [21:54] <~davidb> sounds right
- # [21:54] <smaug> that works
- # [22:21] * Quits: @surkov (surkov@moz-i5m.05u.207.66.IP) (Client exited)
- # [22:36] * Quits: ~davidb (davidb@moz-i5m.05u.207.66.IP) (Quit: Blah blah blah)
- # [22:38] <tbsaunde> smaug: there's some gsettings / dconf thing too
- # [22:38] <tbsaunde> -1 is mac only I think
- # [22:38] <smaug> tbsaunde: had to enable a11y in gnome to see the crash
- # [22:38] <smaug> though, I'm not seeing the child process crash dholbert mentioned
- # [22:39] <tbsaunde> smaug: I can see it sometimes
- # [22:39] <tbsaunde> saddly not when under rr though
- # [22:45] * Parts: clown (clown@moz-v6m.c3o.211.205.IP)
- # [22:49] <smaug> hmm, ProxyAccessible::mChildren[0] is somehow special
- # [22:50] <tbsaunde> smaug: only in the case it holds a DocAccessibleParent* meaning its an accessible for a <iframe> or such
- # [22:52] <tbsaunde> smaug: so, the ownership model here is that the mAccessibles hashtable in the doc owns all the proxies in that document, and I think I'd prefer that random things can't keep other things alive past when they should die
- # [22:56] <smaug> I'm just super worried keeping raw pointers anywhere there
- # [22:56] <smaug> (but still trying to reproduce this child process crash)
- # [22:56] <tbsaunde> of course its unfortunate we can't make it more clear with types, it'd be nice if it could be hashtable<UniquePtr<>>, but I don't see how that's possible
- # [22:57] <smaug> oh, fun, I got a child process crash, because of necko
- # [22:57] <smaug> mozilla::net::PNeckoChild::SendPHttpChannelConstructor
- # [22:57] <tbsaunde> heh
- # [22:58] <tbsaunde> you may find rr doesn't work with a11y, and you don't want to debug the child process crash without it
- # [22:58] <tbsaunde> the root of the child process is we update the accessible tree in the child process, but we don't tell the parent there's no accessibles
- # [22:58] <tbsaunde> so DocAccessibleParent::RecvShow returns false
- # [23:01] <tbsaunde> smaug: ^
- # [23:01] * smaug has never used rr
- # [23:04] <tbsaunde> smaug: when it works its rather nice
- # [23:09] <smaug> tbsaunde: so why in particular do you not want refcounting?
- # [23:10] <smaug> also, do we have any case in a11y where something is happening in a different thread?
- # [23:10] <tbsaunde> smaug: no, its all single threaded
- # [23:11] <tbsaunde> smaug: for one thing the references that are most likely to point to dead things are the ones that can't be refptr's even if you want them to be because they are in unions
- # [23:11] <smaug> good, so if needed, more stuff can be made cycle collectable. (hopefully no need)
- # [23:11] <smaug> ProxyAccessible doesn't look like an union
- # [23:12] <tbsaunde> smaug: in generally I think its not great to make an ownership model less clear
- # [23:12] <smaug> and its lifetime management isn't trivial atm
- # [23:12] <tbsaunde> its pretty simple?
- # [23:12] <smaug> not to someone reading the code first time
- # [23:12] <smaug> you have these raw pointers hanging eveyrwhere
- # [23:12] <tbsaunde> look at what ProxyCreated / Destroyed in atk/AccessibleWrap.cpp does
- # [23:12] <smaug> everywhere
- # [23:13] * Quits: anvk (anovak@moz-871.mvk.11.204.IP) (Quit: Leaving.)
- # [23:13] * yzen is now known as yzen_
- # [23:13] <tbsaunde> I guess it isn't really explained
- # [23:14] <smaug> tbsaunde: oh, I'm not talking about that part
- # [23:15] <smaug> tbsaunde: ProxyAccessible has mDoc and array mChildren
- # [23:15] <smaug> both member variables are raw pointers to objects
- # [23:15] <smaug> DocAccessibleParent extends ProxyAccessible, and it has mChildDocs and mParentDoc, again raw
- # [23:16] <smaug> and then there is mAccessibles hashtable, which contains raw pointers
- # [23:16] <smaug> all that setup isn't too clear
- # [23:16] <smaug> to someone not familiar with the code
- # [23:17] * Quits: @yzen_ (yzen@moz-i5m.05u.207.66.IP) (Ping timeout: 121 seconds)
- # [23:17] <tbsaunde> smaug: my point is the references I'd expect are most likely to point at dead objects are the ones in atk objects
- # [23:17] <tbsaunde> smaug: yeah, that's why I said I guess it isn't explained
- # [23:18] <smaug> why? atk objects are refcounted
- # [23:18] <tbsaunde> external code holds onto those not use, and there refs to proxies aren't strong
- # [23:21] <smaug> ah
- # [23:21] <smaug> I'm not there yet, so not worried about that yet :)
- # [23:21] * Quits: maxli (maxli@moz-favcmk.cs.uwaterloo.ca) (Quit: Leaving.)
- # [23:25] <smaug> tbsaunde: I argue that strong references self-document the ownership model. And raw refs can be used when someone doesn't own the other object, but there must be explicit Disconnect() for those cases
- # [23:31] <tbsaunde> smaug: they document a model, but the model of a whole bunch of things could theoretically own each other seems worse than a model of these things are owned by this own thing
- # [23:32] <tbsaunde> *one thing
- # [23:33] <smaug> tbsaunde: we're dealing with tree structures here
- # [23:34] <smaug> so natural ownership would be parent owns children
- # [23:34] <smaug> and everything else is raw + explicit Disconnect()
- # [23:34] * Quits: newtron (newtron@moz-6lh.c31.71.199.IP) (Ping timeout: 121 seconds)
- # [23:34] <smaug> well, IPC would also have strong ref in that case, for the stuff doing IPC
- # [23:36] <tbsaunde> smaug: but why should parent own children when that's actually controlled by doc?
- # [23:37] <smaug> well, if everything is controlled by doc (whatever that means here), then doc should have strong references to the objects it owns
- # [23:38] <tbsaunde> smaug: if I could make mAccessibles a hash table of UniquePTr I'd absolutely do that
- # [23:39] <tbsaunde> but I don't see why anything other than that hash table should be allowed to keep things alive
- # [23:40] <tbsaunde> we might actually be able to change ProxyEntry to use UniquePtr, that's definitely worth trying
- # [23:41] <smaug> tbsaunde: well, that keeps DocAccessibleParent objects alive? those are ProxyAccessible objects too. DocAccessibleParent is deleted by ContentParent::DeallocPDocAccessibleParent
- # [23:42] <smaug> it things were refcounted, ContentParent::DeallocPDocAccessibleParent would just Release
- # [23:42] <smaug> (IMO all ipdl things should be refcounted by default)
- # [23:43] <tbsaunde> smaug: no, those are owned by content parent
- # [23:44] <smaug> tbsaunde: yet you just said "but I don't see why anything other than that hash table should be allowed to keep things alive"
- # [23:44] <smaug> and we're talking about ProxyAccessible objects here
- # [23:45] <smaug> so DocAccessibleParent objects are special
- # [23:45] <tbsaunde> smaug: yeah, ok, so non docs are the only thing in the hash table, and the hash table owns them, docs are owned by ipdl
- # [23:53] * Joins: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net)
- # [23:57] * Quits: yliao (yliao@moz-2hf62a.HINET-IP.hinet.net) (Ping timeout: 121 seconds)
- # Session Close: Thu Jan 29 00:00:01 2015
Previous day, Next day
Think these logs are useful? Then please donate to show your gratitude (and keep them up, of course). Thanks! — Krijn