Someone who understands what regions cairo is actually repainting needs to adjust the patch from bug 220953 to work in cairo builds.
Not gonna block 1.9 on this. roc, could we possibly do this in the viewmanager instead of widgets? That would also flash the area we really are repainting...
...and would be cross-platform, which would mean we'd stop having to rewrite it.
With our newfangled alpha support, we could even draw the actual update region in a translucent color, overlaying the page content. It would be a permanent thing on the rendered display, as opposed to just a flash, but I think that would actually be preferable. How's that sound?
(In reply to comment #3) > With our newfangled alpha support, we could even draw the actual update region > in a translucent color, overlaying the page content. It would be a permanent > thing on the rendered display, as opposed to just a flash, but I think that > would actually be preferable. How's that sound? Wouldn't that mean that if I paint area A and then area B, both A and B have the "update" overlay, so I can't tell which one was the last one to be repainted? It's hard to do a temporary flash from non-widget code because that stuff's all double buffered in the normal paint path. We could create a rendering context and do an immediate draw but I'd actually like to eliminate drawing outside of the paint event. I'm not sure what the right way to go is.
> It would be a permanent thing on the rendered display That's not really desirable, because then we'd just end up with the whole window showing it... In my experience paint flashin is most useful for things like scrolling, DHTML, UI interaction, where every so often the whole thing gets invalidated.
For Xgl there's a little tool that does this for any and all applications (probably works in aiglx too). The only problem is that it shows damage due to scrolling as well as invalidation, which I don't want, but that may not worry other people. I think OSX has something similar. A related question is whether we want the app to pause while flashing. Currently it does, which can be good and bad...
Do we really still not have paint flashing on X?
Last I checked, we do not. Nor do we really have it on Mac, though QuartzDebug can help with that a bit.
Paint flashing belongs more in layers now than in Cairo drawing code; IIRC we have something approximating it too.
I don't see anything approximating this in the code. I'm going to try hacking this up as a learning exercise. I don't see a logical bottleneck in the Layer classes and my plan is to tack it on to gfxContext::Paint(). Will that work?
This probably needs to be done as part of bug 690064.
Created attachment 574190 [details] [diff] [review] Introduce paint flashing This patch introduces 'paint flashing' what is does is it draws a transparent 'shade' in a random color (varying per paint) over the area that was invalidated.
Comment on attachment 574190 [details] [diff] [review] Introduce paint flashing Review of attachment 574190 [details] [diff] [review]: ----------------------------------------------------------------- ::: layout/base/FrameLayerBuilder.cpp @@ +2130,5 @@ > aContext->Restore(); > } > + > +#ifdef DEBUG > + if (Preferences::GetBool("nglayout.debug.paint_flashing", false)) { Use AddBoolVarCache and make this non-#ifdef DEBUG
Created attachment 574521 [details] [diff] [review] Introduce paint flashing v2 Updated as per review comments.
https://hg.mozilla.org/mozilla-central/rev/6235e0c97c3b To save time/mistakes when merging, please could you set assignee + target milestone when landing on inbound (https://wiki.mozilla.org/Tree_Rules/Inbound#Please_do_the_following_after_pushing_to_inbound). Thanks :-)
Created attachment 594628 [details] [diff] [review] Add nglayout.debug.paint_flashing to all.js We should have done this originally
https://hg.mozilla.org/integration/mozilla-inbound/rev/677bf71d3030
https://hg.mozilla.org/mozilla-central/rev/677bf71d3030
*** Bug 723716 has been marked as a duplicate of this bug. ***
or
