1. Launch the MiniBrowser and go to e.g. https://twitter.com/webkit
2. Click on a tweet to see its details.
Expected outcome: there's a pop-up window with the selected tweet.
Actual outcome: the main window is dimmed but no pop-up appears. The address bar shows the URL of the selected tweet (https://twitter.com/webkit/status/...)
This happens with 2.10.8 (tested in Debian and Arch Linux), but I haven't been able to reproduce it with 2.10.6 or 2.11.90.
This does not happen with all tweets. I can reproduced it with this one from March the 2nd: "After reaching consensus on Custom Elements API, we will begin implementing without a prefix and enabled by default".
We need to run a bisect in 2.10 branch then
Ok, this is not 2.10 specific, I can reproduce it with MB in trunk and 2.12 and with ephy depending in the window size. I can confirm this never happens reverting r196803. So, I suspect it has to do with the visible content size reporting, or the layer size/positioning.
Yes, it seems it has to do with the visible rectangle. If you open the MB and make the window larger, the popup works until you scroll more than ~ 500
This page is also broken with 2.10.8, but not with 2.10.7:
It's probably the same bug.
(In reply to comment #4)
> This page is also broken with 2.10.8, but not with 2.10.7:
> It's probably the same bug.
Yes, reverting r196803 also fixes that.
Maybe related to this, I also noticed that the top bar on Facebook disappears after scrolling up and down a few times. I think you need an account in order to reproduce this.
I've tried to figure out what the problem is, but I haven't been able to fix anything. It seems to be related with the visible contents rect transformation. In some cases, just returning the contents rect, like we do for the non composited contents layer, fixes the problem, but I don't really know why. Yoon has told me that he won't be able to look at this during this week, so I'm not sure what to do. The problem fixed by r196803 was very serious, but it has introduced more regressions than what we thought. Not only the ones mentioned here, we are also affected by bug #154912, for example, and I'm pretty sure more issues will be reported once we release 2.12. So, the question is whether we have broken more sites than fixed. In some cases the regressions are in popular web sites like twitter or facebook. What do you think? Is there anybody else apart from Yoon who could take a look at the issue?
I guess roll it out, release 2.10.9, and release again once we get it fixed again. It's a tough call because it fixes by far our most serious bug, but the sites that bug occurs on are much less popular than Twitter, I think we should prioritize Twitter.
Rolled out in r198266.
*** Bug 154879 has been marked as a duplicate of this bug. ***
*** Bug 155059 has been marked as a duplicate of this bug. ***