All users were logged out of Bugzilla on October 13th, 2018
Status
()
People
(Reporter: jruderman, Assigned: bzbarsky)
Tracking
(Blocks: 1 bug, {testcase})
Firefox Tracking Flags
(Not tracked)
Details
Attachments
(3 attachments, 1 obsolete attachment)
Created attachment 410317 [details]
testcase (dynamic)
Based on layout/reftests/bugs/424236-5.html. Probably a regression.
|
(Reporter) | |
Comment 1•9 years ago
|
||
Created attachment 410318 [details]
reference
|
||
Comment 2•9 years ago
|
||
Regressed between Linux x86_64 nightlies: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=736619af9701&tochange=00924c08d33d Presumably bug 501847.
Blocks: 501847
|
(Assignee) | |
Comment 3•9 years ago
|
||
Yes. The key part is that the two testcases have different frame trees. I guess I can actually make this case (the trailing inline is empty and a block is the first thing appended) generate the same frame tree as the static case. Let me give that a shot.
|
|
(Assignee) | |
Comment 4•9 years ago
|
||
Hmm. So here's a question about the continuation model. Is it possible to have an inline whose first continuation has no kids but which has some other continuation that does have kids?
|
|
(Assignee) | |
Comment 5•9 years ago
|
||
Created attachment 413782 [details] [diff] [review] Fix, assuming the answer to comment 4 is "yes" Though I suspect that if that answer is "yes" some of our other append code might not be quite correct... so I hope it's actually "no".
I think the answer is "no".
|
|
(Assignee) | |
Comment 7•9 years ago
|
||
Created attachment 413791 [details] [diff] [review] "no" == simpler fix
Attachment #413782 -
Attachment is obsolete: true
Attachment #413791 -
Flags: review?(roc)
Attachment #413782 -
Flags: review?(roc)
|
|
(Assignee) | |
Comment 8•9 years ago
|
||
Though I guess the other approach might be faster if we have lots of continuations for that trailing inline..... whichever one you like more?
Attachment #413791 -
Flags: review?(roc) → review+
|
|
||
Comment 9•9 years ago
|
||
Why didn't the outline merging code detect this case, anyway? Should it have? (Should we be fixing this bug in two different ways?)
|
|
(Assignee) | |
Comment 10•9 years ago
|
||
Possibly. How should outline merging generally work for ib splits? Where do we merge outlines, in general?
|
|
||
Comment 11•9 years ago
|
||
I suppose the little unioning we do (which isn't nearly what we should) is in nsCSSRendering::PaintOutline, about 20 lines in.
|
|
(Assignee) | |
Comment 12•9 years ago
|
||
Ah. That's not actual unioning; that's just drawing the outline around a different rect. It seems to me that we don't actually coalesce outlines in a sane way at all. Consider: data:text/html,<!DOCTYPE html><span style="outline: 5px solid black">aaa<br>bbb</span>
|
|
(Assignee) | |
Comment 13•9 years ago
|
||
Pushed http://hg.mozilla.org/mozilla-central/rev/1126b08afa75
Status: ASSIGNED → RESOLVED
Last Resolved: 9 years ago
Flags: in-testsuite+
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•