542136 - "ASSERTION: Incorrect aState.mPrevChild before inserting line at end"

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: testcase

###!!! ASSERTION: Incorrect aState.mPrevChild before inserting line at end: 'aState.mPrevChild == mFrames.LastChild()', file /Users/jruderman/mozilla-central/layout/generic/nsBlockFrame.cpp, line 2162

bz marked a previous bug with this assertion as security-sensitive (bug 522170), so I'm filing this one as security-sensitive.

Comment 1

[Timothy Nikkel (:tnikkel)]

On trunk this assertion shouldn't itself lead to a security problem. On 3.6 and earlier it is likely to cause a messed up frame tree, so a security problem is likely just around the corner.

Comment 2

[Timothy Nikkel (:tnikkel)]

So I think the problem lies with this bit of code in nsBlockFrame::ReflowInlineFrame http://mxr.mozilla.org/mozilla-central/source/layout/generic/nsBlockFrame.cpp#3838

    // If we just ended a first-letter frame or reflowed a placeholder then 
    // don't split the line and don't stop the line reflow...
    if (!(frameReflowStatus & NS_INLINE_BREAK_FIRST_LETTER_COMPLETE) && 
        nsGkAtoms::placeholderFrame != frameType) {
      // Split line after the current frame
      *aLineReflowStatus = LINE_REFLOW_STOP;
      rv = SplitLine(aState, aLineLayout, aLine, aFrame->GetNextSibling(), aLineReflowStatus);
      NS_ENSURE_SUCCESS(rv, rv);
    }

At this point *aLineReflowStatus is already LINE_REFLOW_STOP because the reflow status was break after for the frame we just reflowed, but we don't split the line because of the first letter complete check. And so we stop reflow after 1 of 2 frames on the line and when we get back to nsBlockFrame::ReflowDirtyLines mPrevChild is not the last frame on the line. Adding "|| *aLineReflowStatus == LINE_REFLOW_STOP" to the if condition fixes the asserts for me.

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Sounds right.

Comment 4

[Timothy Nikkel (:tnikkel)]

Attachment: patch

Comment 5

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/mozilla-central/rev/537f60b8e8df

Comment 6

[Timothy Nikkel (:tnikkel)]

After some bake time on trunk this is something that would probably be good to have on the branches.

Comment 7

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 423615 [details] [diff] [review]
patch

a=beltzner for 1.9.2.2 and 1.9.1.9

Comment 8

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/dda32ceb3b5b

Comment 9

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/9185a0d68573

Comment 10

[Al Billings [:abillings - ex-MoCo]]

I notice with the testcase post-fix that I'm not seeing the assert above anymore on OS X but I do get:

###!!! ASSERTION: overflow list w/o frames: 'mFrames.NotEmpty()', file /Users/albill/mozilla-192/layout/generic/nsInlineFrame.cpp, line 374


Is this a non-issue?

Comment 11

[Timothy Nikkel (:tnikkel)]

I don't get that assertion on trunk, I would guess because of bug 533379 which is fixed on trunk but not 1.9.2.

Comment 12

[Al Billings [:abillings - ex-MoCo]]

Well, that isn't *this* assertion so I will call this verified1.9.2.

Comment 13

[Carsten Book [:Tomcat]]

verified 1.9.2-2 with Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US; rv:1.9.2.2) Gecko/20100321 Firefox/3.6.2 ID:20100321170417 (debug build) - the testcase didn't caused the assertion from comment #0 - for comment 10 i have made a comment in bug 533379 - maybe that bug need to land on 1.9.2 too

Comment 14

[Timothy Nikkel (:tnikkel)]

Added crashtest
http://hg.mozilla.org/mozilla-central/rev/f8c56496ae7d