Rendering and then scrolling the League Of Legends website to the very top will result in generating big black bars on either side of the page. 1. Turn on Azure-Thebes Wrapper 2. Load the url 3. Scroll Down 4. Scroll back to the top. Observe the black patches. Also, they go away if you shift the focus back and forth to Nightly.

Screenshot of the issue.

Here's another example: http://mozillamemes.tumblr.com/ Scroll the page and notice that the background turns black behind the columns.

I suspect this might be one of the component alpha bugs I introduced (and landed fixes for today on m-i). Try seeing if it's gone in tomorrow's nightly (if the patches make it to m-c in time).

Using an hourly cset: 576a14e57ea6, The issue remains.

Confirmed with gfx.content.azure.enabled=true in http://hg.mozilla.org/mozilla-central/rev/576a14e57ea6 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:14.0) Gecko/20120412 Firefox/14.0a1 ID:20120412160832 Regression pushlog: http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=c95b597b5f08&tochange=9fa58c6060c5

Last good : e36fabc31211 First bad : 9fa58c6060c5

I believe some of our component alpha fixes may have fixed this. I can no longer reproduce this on inbound, can anyone else?

(In reply to Leman Bennett (Omega X) from comment #2) > Here's another example: http://mozillamemes.tumblr.com/ > > Scroll the page and notice that the background turns black behind the > columns. Hrm, I do see an artifact here but only when I scroll down all the way on the last meme.. I really would like a minimized testcase, I have no idea what's going on :(

(In reply to Alice0775 White from comment #11) > Created attachment 616403 [details] > another sample html This is an excellent stand-alone testcase. Hopefully this will let me solve the issue.

When a ThebesLayer changes the content type of the primary surface, we need to mark the content of our surface invalid. While in this code I noticed right now we create a redundant DrawTarget here that will later be overwritten. This isn't a big issue but it is a waste, so I should fix it at some point. I'll file a bug for this.

Comment on attachment 616705 [details] [diff] [review] Properly mark ThebesLayer content invalid when changing content type. Are we likely to forget anymore of these?

(In reply to Jeff Muizelaar [:jrmuizel] from comment #14) > Comment on attachment 616705 [details] [diff] [review] > Properly mark ThebesLayer content invalid when changing content type. > > Are we likely to forget anymore of these? Not likely I don't think, I tried to re-check all of them. But you never know.

Bas It's possible this patch also fixed bug https://bugzilla.mozilla.org/show_bug.cgi?id=700088 which after updating to cset http://hg.mozilla.org/mozilla-central/rev/cc5254f9825f now works fine. Do you think so?