mozilla :: #apz

13 Sep 2017
13:23petrutaHello! Does anyone know if there's a bug on file for this Facebook scrolling issue: ?
13:25petrutaBasically after login, on some pages that have several links on the right side, scrolling the timelime will move that div up/down while it should be fixed on page. This happens on older Fx versions too, but I couldn't find the regression so I'm assuming it might be a recent modification on their side.
14:11katspetruta: no bug on file that i'm aware of for that
14:13petrutaThanks kats, I'm filling one now.
14:25firebotBug 1399495 NEW, Jerkiness while scrolling on Facebook pages
16:05rhuntbotond: ping
16:06botond|homerhunt: pong
16:12rhuntbotond: so I think you've addressed all the things I saw in your scrollbar dragging patches, and I also did a little bit of testing on your laptop and couldn't find anything obvious so I think you should be good to land
16:12botond|homerhunt: great, thanks! i'm just doing some final testing locally, and will land soon
16:13botond|homerhunt: thanks for the reviews :)
16:14rhuntbotond: awesome, your patches were well written so once I knew the existing code it was easy :)
19:14rhuntkats: ping
19:14katsrhunt: pong
19:14rhuntkats: I was using botond's touch laptop and noticed a weird behavior with zooming and tracked it down to bug 789906
19:14firebot NEW, When using touch UI (for example on Win8), zooming shouldn't reflow, but just scale
19:15rhuntkats: I saw you mentioned it'd be a bit of work to fix, do you have a summary of what's involved?
19:16katsrhunt: it's been a while since i looked, but i think the main problem was that trying to do pinch-zooming with containerless scrollframes doesn't work very well
19:16katsrhunt: we have containerless scrollframes on desktop but not on android
19:16katsso on android it works
19:17rhuntkats: a containerless scrollframe?
19:17katsand there was some reason we couldn't use containerful scrollframes on desktop, but i can't remember what it was
19:18katswith containerfull scrolling we have a container layer wrapping all the things that scroll, and the scroll metadata goes on that layer
19:18katswith containerless there is no container layer, and the metadata is put on each child layer instead
19:18kats(that's a simplified version, i don't even remember all the details)
19:20rhuntkats: interesting
19:21katsrhunt: did you try running with apz.allow_zooming set to true?
19:23rhuntkats: no, I didn't. I'll give it a shot
19:27katsrhunt: what was the zooming problem you saw?
19:27rhuntkats: not a problem, but just that it does full zoom when I think most people expect touch to work like on their phone
19:28katsgetting pinch zoom working on desktop has been on our to-do list for a while but it keeps getting pushed back
19:30rhuntkats: so for some reason i'm not able to trigger pinch zooming with apz.allow_zoom enabled
19:30katsrhunt: hm, maybe it's broken more since the last time i tried
19:31rhuntkats: what was broken before, if you remember?
19:32katsrhunt: i don't recall offhand. it's probably been a year at least since i tried it
19:32rhuntkats: np, just curious
19:33botond|homerhunt: you're getting a pinch gesture causing full-zoom on the touchscreen laptop?
19:33botond|homerhunt: what page?
19:34katsthat's the expected behaviour currently
19:34rhuntbotond: yeah, but it has nothing to do with your patches I was just testing out touch input
19:34botond|homekats: do you know how it's implemented?
19:35katsbotond: through here
19:35botond|homekats: ah, right, NotifyPunchGesture. i should remember that since i reviewed it :p
19:35katsPunch gesture! we should totally add that
19:35katspunch the screen to file a bug
19:37botond|homekats: do you know if the handling of that on the content side is platform-specific?
19:41katsbotond: it looks like it should be shared
19:42katsalthough i'm not totally sure
19:43botond|homekats: i'm not getting full-zoom on my Linux touch laptop... was just wondering if that's expected. after all, getting full-zoom is better than nothing
19:44katsbotond: that might be worth looking into
19:44katsi thought it was supposed to work
20:30botond|homekats: found the culprit:
20:30botond|homekats: so looks like it's Windows-only for now
20:31katsbotond: ah. presumably you can modify the pref locally and see what happens
20:38botond|homekats: seems to work after setting the pref locally
20:39botond|homekats: i think we can probably just enable this on Linux. the comment for why it's disabled refers to "issues with track pad input
20:39botond|home", but i don't believe track pads generate pinch gesture on Linux
20:39botond|home(also it dates back to 2011, which is long before we had any touch support on Linux)
20:40katsbotond: sounds ok to me
20:41katswe can always re-disable it if there's issues
21:27botond|homemstange: ping?
21:45botond|homerhunt: i think i know why you're not seeing pinch-zoom behaviour with apz.allow_zooming set to true
21:46botond|homerhunt: you need to also set dom.meta-viewport.enabled to true, because of this:
21:47botond|homerhunt: with both of those prefs set, i can perform pinch-zooming
21:47botond|homerhunt: the most prominent issue that i notice is that scrollbars aren't drawn after you zoom in
22:21mstangebotond|home: pong
22:22botond|homemstange: hey! i was going to ask you the question i wrote up in if you happen to have any insight about it, feel free to share, otherwise i'll wait for CJ to weigh in
22:22firebotBug 1382534 NEW, Black gaps when scrolling BBC photo gallery
22:23mstangebotond|home: I see, I don't have any insight about it
22:23botond|homemstange: ok, no worries
22:23mstangebut the answer is probably "there's currently no faster version to get bounds for the actual path, but maybe you can add one"
22:24botond|homemstange: is speed the only reason i may not want to create a DrawTarget at display list building time?
22:25mstangebotond|home: I can't think of another reason
22:27botond|homemstange: ok. i thought perhaps we may not have the necessary information to choose the kind of DT to create, or other such prerequisites, at display list building time
22:28botond|homemstange: but if that's not the case, i can try creating a DT, at least for now as a temporary solution
22:28mstangewe have places where we use the ScreenReferenceDT to obtain path bounds, I think
22:29botond|homemstange: ooh, that sounds useful!
22:29mstangeSVGGeometryFrame::GetBBoxContribution does something like this
22:30botond|homemstange: indeed! and i call that function (indirectly) for the clip-path:url() case
22:30botond|homemstange: so, i suppose there's no reason i couldn't go with the same approach for basic-shape
22:30mstangeI agree
22:30botond|homemstange: thanks for pointing me to that!
22:30mstangeyou're welcome :)
22:33botond|homemstange: (you're more insightful that you give yourself creditor for :p)
22:33botond|homecredit* (can't spell today...)
14 Sep 2017
No messages
Last message: 7 days and 19 hours ago