Closed Bug 887975 Opened 7 years ago Closed 7 years ago
Flickering when panning content and the richgrid on the start UI
When you pan the start UI, the richgrid will periodically disappear. See attached video.
This is all changing with the async pan zoom controller, we should re-evaluate if this is a bug when that lands and is switched on.
I'm pretty sure this is in layers code. If we can assure ourselves it's not in the front end we should update the product/component so gfx folks see it. jwilde, can you reproduce this in mc tip with "gfx.direct2d.disabled" set to true?
No longer blocks: metrov1triage
Updated to mc tip and clobbered/rebuilt: Without gfx.direct2d.disabled: flicker. With gfx.direct2d.disabled: no flicker.
Component: General → Graphics: Layers
Product: Firefox for Metro → Core
Version: unspecified → Trunk
reproducible on this page: https://developer.mozilla.org/en-US/docs/HTML/HTML_Elements/textarea#HTML_Content down near the bottom where the page has an embedded iframe with a text area in it.
Summary: Flickering when panning richgrid on the start UI → Flickering when panning content and the richgrid on the start UI
(In reply to Milan Sreckovic [:milan] from comment #7) > Bas, thoughts? Hrm, I'm not sure what the behavior of Metro is at the moment with Direct2D disabled, definitely looks like a graphics problem but I have no initial idea on what the cause is. Also no Windows 8 machine with me in SF to try and debug this.
Hi Bas, we have Metro team members in SF who could show you on a Win 8 device.
(In reply to Marco Mucci [:MarcoM] from comment #10) > Hi Bas, we have Metro team members in SF who could show you on a Win 8 > device. I'd be more than happy to look at it, I'm at the 7th floor on Elika's desk. I can probably give some pointers as well but I doubt I can do significant debugging right now.
Hi Ally, can you see Comments 9,10,11.
Bas, headed your way
Had a look at Ally's machine. ContentHostDoubleBuffered::UpdateThebes and ThebesLayerBuffer::BeginPaint is kind of where all the mess of moving the buffer around in memory and making sure front and back buffer are in sync happens. This bug looks to me like sometimes the syncing is not correct (i.e. the content in the middle, which is not being redrawn, is in one of the buffers but not in the other). It's particularly interesting that as you scroll it can re-appear, meaning it either gets redrawn or we're just ending up with 'the other' buffer of the swap chain where the content is still present.
Thanks for looking into it Bas. Anything you can do to help you would great.
on latest tip (Wed Jul 24 05:36:15 2013 PDT), I can't reproduce.
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WORKSFORME
Oddly enough, this is now back with apz enabled.
Status: RESOLVED → REOPENED
Resolution: WORKSFORME → ---
I'm also seeing this again with APZC disabled, though for a while it went away.
On machines with Atom processors the tiles disappear when you pan left or right.
I get that sometime with my sony vaio which is intel
Does it work with the patch from bug 900742?
(In reply to Milan Sreckovic [:milan] from comment #23) > Does it work with the patch from bug 900742? Doesn't look like an event related patch like that would help but I can take it for a spin to see. One interesting thing I noticed about this today, when we scroll via the mouse wheel we don't see a flicker. Mouse wheel translates into scrollBy calls on the start ui panel. Not sure why it would not reproduce in that case, but reproduce when dragging the scrollbar with the mouse to scroll.
I discovered today that removing the background image ("Firefox" watermark) from the start page fixes the flickering on my system. Is this an acceptable band-aid until we can find a fix or workaround that lets us scroll correctly with the image in place?
That's acceptable with me. The improved performance is definitely a bigger win here.
We should spin this patch off into a metro bug so the gfx bug stays open (or change up this bug and file a new gfx bug.) We still want a bug open on the compositor / layers painting problem. (I've seen this occasionally in content as well.)
Let's spin a new gfx bug for that and CC kats, that issue will probably be one he, or BenWa has to fix. It would be ideal if we could use jing to capture a short video of the problem happening on content. It's just a software that you install, press a button to record your screen, and it gives you a link you can use online.
Jim if you can reproduce that easily you can also try tweaking some of these prefs: http://mxr.mozilla.org/mozilla-central/source/gfx/layers/ipc/AsyncPanZoomController.cpp#146
Comment on attachment 786642 [details] [diff] [review] work-around: remove "watermark" background image Review of attachment 786642 [details] [diff] [review]: ----------------------------------------------------------------- Does what it says on the box. Though I hope we get a real gfx fix soon. :)
Attachment #786642 - Flags: review?(ally) → review+
https://hg.mozilla.org/integration/fx-team/rev/4e2e139c6b83 As discussed above, we'll close this bug and open a new bug to track the underlying graphics or layout issue.
Status: ASSIGNED → RESOLVED
Closed: 7 years ago → 7 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla26
(In reply to Matt Brubeck (:mbrubeck) from comment #32) > https://hg.mozilla.org/integration/fx-team/rev/4e2e139c6b83 > > As discussed above, we'll close this bug and open a new bug to track the > underlying graphics or layout issue. Did we file a new bug yet?
(In reply to Jim Mathies [:jimm] from comment #34) > (In reply to Matt Brubeck (:mbrubeck) from comment #32) > > https://hg.mozilla.org/integration/fx-team/rev/4e2e139c6b83 > > > > As discussed above, we'll close this bug and open a new bug to track the > > underlying graphics or layout issue. > > Did we file a new bug yet? filed bug 903421.
You need to log in before you can comment on or make changes to this bug.