1390389 - stylo: Assertion failure: parent->FirstContinuation() == aParent || (parent->StyleContext()->IsInheritingAnonBox() && parent->GetContent() == aParent->GetContent())

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Jesse Schwartzentruber (:truber)]

Attachment: testcase.html

The attached testcase causes an assertion in m-c rev df9beb78189 with stylo enabled by pref

Assertion failure: parent->FirstContinuation() == aParent || (parent->StyleContext()->IsInheritingAnonBox() && parent->GetContent() == aParent->GetContent()), at /home/worker/workspace/build/src/layout/base/ServoRestyleManager.cpp:208
# 0  libxul.so!mozilla::ServoRestyleState::ProcessMaybeNestedWrapperRestyle [ServoRestyleManager.cpp:183]
# 1  libxul.so!mozilla::ServoRestyleState::ProcessWrapperRestyles [ServoRestyleManager.cpp:168]
# 2  libxul.so!mozilla::ServoRestyleManager::ProcessPostTraversal [ServoRestyleManager.cpp:888]
# 3  libxul.so!mozilla::ServoRestyleManager::ProcessPostTraversal [ServoRestyleManager.cpp:873]
# 4  libxul.so!mozilla::ServoRestyleManager::DoProcessPendingRestyles [ServoRestyleManager.cpp:1094]
# 5  libxul.so!mozilla::PresShell::DoFlushPendingNotifications [PresShell.cpp:4200]

Comment 1

[Ting-Yu Lin [:TYLin] (PDT, UTC-7)]

Boris, you wrote this assertion. Any ideas?

Comment 2

[Bob Clary [:bc] (inactive)]

fyi, bughunter reproduces this on <https://id.yahoo.com/> and <https://maktoob.yahoo.com/?p=us>. STR: Load page, open Web Developer -> Web Console

Comment 3

[Xidorn Quan [:xidorn] UTC+11]

Attachment: Frame tree when the assertion happens

This is the partial frame tree when this assertion happens.

At the moment, in ServoRestyleState::ProcessMaybeNestedWrapperRestyle,
* "cur" points to the table frame (116aec9a1d8 in this dump),
* "parent", according to the context, should point to its first inline ancestor (116aec9a0b0 in this dump),
* "aParent" points to the first ib-split continuation of "parent" (116aec99de0 in this dump).

Comment 4

[Boris Zbarsky [:bzbarsky]]

Finally got to this.  Comment 3 is spot-on: we should be using FirstContinuationOrPartOfIBSplit in the assert, not just FirstContinuation.  Patch coming up.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: Fix anon box restyling assertions to correctly handle ib splits

MozReview-Commit-ID: AtBFlvLyg6b

Comment 6

[Boris Zbarsky [:bzbarsky]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=64441a69ff7f139b3ca627671f4ac17603c99f49

Comment 7

[Xidorn Quan [:xidorn] UTC+11]

Comment on attachment 8903011 [details] [diff] [review]
Fix anon box restyling assertions to correctly handle ib splits

Review of attachment 8903011 [details] [diff] [review]:
-----------------------------------------------------------------

Looks reasonable to me, but I don't have enough confident to review this code. Redirecting to emilio.

Comment 8

[Emilio Cobos Álvarez (:emilio)]

Comment on attachment 8903011 [details] [diff] [review]
Fix anon box restyling assertions to correctly handle ib splits

Review of attachment 8903011 [details] [diff] [review]:
-----------------------------------------------------------------

::: layout/base/ServoRestyleManager.cpp
@@ +214,3 @@
>    Maybe<ServoRestyleState> parentRestyleState;
> +  nsIFrame* parentForRestyle = aParent;
> +  if (FirstContinuationOrPartOfIBSplit(parent) != aParent) {

This means we need to move FirstContinuationOrPartOfIBSplit out of the `ifdef DEBUG` as well, right?

Comment 9

[Boris Zbarsky [:bzbarsky]]

> This means we need to move FirstContinuationOrPartOfIBSplit out of the `ifdef DEBUG` as well, right?

I actually just changed this to:

  if (nsLayoutUtils::FirstContinuationOrIBSplitSibling(parent) != aParent) {

locally last night when I found the opt build bustage, and thought I'd requested review with that change... clearly not.

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/ca0fcfb8cec2
Fix anon box restyling assertions to correctly handle ib splits. r=xidorn

Comment 11

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/ca0fcfb8cec2