+++ This bug was initially created as a clone of Bug #1105137 +++ The text appears behind the image in page https://dl.dropboxusercontent.com/u/95157096/85f61cf7/ItJnkL8E9x.html. Screenshot https://dl.dropboxusercontent.com/u/95157096/85f61cf7/Qxj3jZf7uG.png. Bug 1105137 does not fix the problem...

Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=506226a2e6c0&tochange=f1d7da5ff1ed Last Good: 1de2c87f94e5 First Bad: 2a2316981708 Triggered by: 2a2316981708 Simon Montagu — Bug 1062963 patch 3: make nsFloatManager's origin a LogicalPoint, adapt GetFlowAreas, AddFloats, ClearFloats, etc. to use it and make nsFloatManager region functions work with logical region. r=jfkthame

This also happens in https://dl.dropboxusercontent.com/u/95157096/85f61cf7/uidssg5u5s.html. Screenshot https://dl.dropboxusercontent.com/u/95157096/85f61cf7/2sjkgf65kg.png I think this happens if the picture within the frame.

STR 1. Load reduced testcase and reload ---- observe layout 2. Zoom In and then Zoom Out ---- observe layout Actual Results: Layout is broken at step 1. layout is corrected at step 2, after reflow(ZoomIn/Out)

[Tracking Requested - why for this release]: layout is broken due to regression of Bug 1062963

I don't really understand this testcase, and it doesn't show the bug when loaded locally, so I'm not sure if I want to check it in, but here it is as a reftest so we can decide what to do with it. The test from bug 1114329 may be sufficient anyway.

Comment on attachment 8540208 [details] [diff] [review] Reftest based on the testcase Review of attachment 8540208 [details] [diff] [review]: ----------------------------------------------------------------- It looks like the wrong files ended up in this patch... As for the testcase in this bug, note that it doesn't fail on initial load; you have to hit Reload to trigger the bad rendering. I'm not sure whether we'd be able to reproduce that behavior in the reftest framework, or if it'd need to be converted to a mochitest form.

Important layout regression, tracking

Comment on attachment 8540205 [details] [diff] [review] Patch: make GetLogicalNormalPosition return the correct logical origin of the frame, not just convert the physical origin to logical coordinates Review of attachment 8540205 [details] [diff] [review]: ----------------------------------------------------------------- Makes sense - and fixes the bug, according to local testing. :)

https://hg.mozilla.org/integration/mozilla-inbound/rev/bf55c6f6424d https://hg.mozilla.org/integration/mozilla-inbound/rev/860181fd59cf With the reftest from bug 1114329, since this patch fixes that test, but not the original problem reported in the bug.

Simon, can we have an uplift request for aurora? Thanks

Comment on attachment 8540205 [details] [diff] [review] Patch: make GetLogicalNormalPosition return the correct logical origin of the frame, not just convert the physical origin to logical coordinates Approval Request Comment [Feature/regressing bug #]: 1062963 [User impact if declined]: Elements in right-to-left pages can be invisible or mispositioned. [Describe test coverage new/current, TBPL]: Baked on trunk since 2014-12-24, checkin includes regression test [Risks and why]: Low risk one-line fix. Left-to-right pages will not be affected by the patch at all. [String/UUID change made/needed]: None

Comment on attachment 8540205 [details] [diff] [review] Patch: make GetLogicalNormalPosition return the correct logical origin of the frame, not just convert the physical origin to logical coordinates thanks!

https://hg.mozilla.org/releases/mozilla-aurora/rev/85a90208e544 https://hg.mozilla.org/releases/mozilla-aurora/rev/998fb156b184