791601 - "Assertion failure: checkmTextrun ? !mTextRun : !Properties().Get(UninflatedTextRunProperty())" with font inflation, -moz-column

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: testcase (requires pref) (crashes Firefox)

With:
  user_pref("font.size.inflation.emPerLine", 15);
Loading the testcase triggers:

Assertion failure: checkmTextrun ? !mTextRun : !Properties().Get(UninflatedTextRunProperty()), at /Users/jruderman/trees/mozilla-central/layout/generic/nsTextFrameThebes.cpp:4373

Security-sensitive because the assertion was added in a security-sensitive bug.

Comment 1

[Jesse Ruderman]

Attachment: stack

Comment 2

[Mats Palmgren (inactive)]

Attachment: stack + frame tree

RemoveInFlows is trying to destroy the text frame marked red.
http://hg.mozilla.org/mozilla-central/annotate/3f0587ce1774/layout/generic/nsTextFrameThebes.cpp#l7273

This is pretty much the same situation as in bug 756241,
but with a block parent.  On line 7300, we explicitly excluded
this case...  I don't recall why we did that - maybe we thought
that FRAMES_ARE_EMPTY would prevent the same crash in blocks?

roc, do you recall?  do you see any reason why doing that block
unconditionally won't work?  (it fixes the crash locally)

Comment 3

[Mats Palmgren (inactive)]

Attachment: fix

Comment 4

[Mats Palmgren (inactive)]

Attachment: fix (wdiff)

Comment 5

[Mats Palmgren (inactive)]

Attachment: before/after frame trees with the fix

Comment 6

[Mats Palmgren (inactive)]

https://tbpl.mozilla.org/?tree=Try&rev=9f066ebad681

Comment 7

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

Comment on attachment 661861 [details] [diff] [review]
fix

Review of attachment 661861 [details] [diff] [review]:
-----------------------------------------------------------------

I don't know why we excluded the parent block here.

Clearing text runs should be safe in any case.

Comment 8

[Mats Palmgren (inactive)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1d8ff0faf0fa

Comment 9

[(no longer active)]

http://hg.mozilla.org/mozilla-central/rev/1d8ff0faf0fa

Comment 10

[Robert Kaiser]

This is marked sec-critical and tracking-firefox17+, are we planning to land this on beta?

Comment 11

[Mats Palmgren (inactive)]

Comment on attachment 661861 [details] [diff] [review]
fix

[Approval Request Comment]
Bug caused by (feature/regressing bug #): 
User impact if declined: sec-critical crash
Testing completed (on m-c, etc.): on trunk/aurora since 2012-09-18
Risk to taking this patch (and alternatives if risky): low risk
String or UUID changes made by this patch: none

Comment 12

[Lukas Blakk [:lsblakk] use ?needinfo]

Please also mark sec-approval? on this patch since it's sec-critical and looking for uplift we'll want to get that feedback as well to be able to time landing.

Comment 13

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 661861 [details] [diff] [review]
fix

Again, my bad, this doesn't need sec-approval please go ahead with uplift.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/7f450abfd318

Comment 15

[u279076]

Kamil, can you please make an attempt at verifying this is fixed? It should be fixed in Firefox 17, 17esr, and 18beta.

Comment 16

[Kamil Jozwiak [:kjozwiak]]

When going through the following issue using the test case HTML file attached above, I receive the following crash every single time:

###!!! ASSERTION: math on NS_UNCONSTRAINEDSIZE: 'NS_UNCONSTRAINEDSIZE != aState.mReflowState.mComputedBorderPadding.left && NS_UNCONSTRAINEDSIZE != aState.mReflowState.ComputedWidth()', file e:/builds/moz2_slave/m-rel-w32-dbg/build/layout/generic/nsBlockFrame.cpp, line 1629

The error doesn't match the one that has been posted above, but it's making it difficult to test as firefox crashes whenever the .HTML file is opened. Is this the same issue as above just a different error message? Or is there something I am doing incorrectly?

I also set the following user preference as mentioned above: ("font.size.inflation.emPerLine", 15)

Received the above assertion crash when going through the following builds:

Firefox 16: http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012/10/2012-10-11-mozilla-release-debug/
Firefox 17: http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012/11/2012-11-20-mozilla-release-debug/
Firefox 18: http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012-12-16-mozilla-beta-debug/

Comment 17

[Jesse Ruderman]

(In reply to Kamil Jozwiak [:kjozwiak] from comment #16)
> ###!!! ASSERTION: math on NS_UNCONSTRAINEDSIZE

That's a non-fatal assertion, so you shouldn't experience it as a crash.  If you're on Windows, you might need to run with XPCOM_DEBUG_BREAK=warn.

(I tend to ignore any assertion that mentions NS_UNCONSTRAINEDSIZE, due to bug 575011.)

Comment 18

[u279076]

Kamil, can you please retest with XPCOM_DEBUG_BREAK=warn as Jesse suggests?

Jesse, is this an environment variable that needs to be set or does this need to be set elsewhere?

Comment 19

[Jesse Ruderman]

Yes, it's an environment variable.

https://developer.mozilla.org/en-US/docs/XPCOM_DEBUG_BREAK

https://developer.mozilla.org/en-US/docs/Debugging_Mozilla_on_Windows_FAQ#Disabling_ASSERTIONS

Comment 20

[Kamil Jozwiak [:kjozwiak]]

Following prerequisites used before testing builds:
- Environment Variable: XPCOM_DEBUG_BREAK=warn
- about:config: ("font.size.inflation.emPerLine", 15)

Firefox 16 (Issue Reproduced): http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012/10/2012-10-11-mozilla-release-debug/
Received Error: Assertion failure: checkmTextrun ? !mTextRun : !Properties().Get(UninflatedTextRunProperty()), at e:/builds/moz2_slave/m-rel-w32-dbg/build/layout/generic/nsTextFrameThebes.cpp:4248

Once the issue was reproduced several times with Firefox 16, used the following builds to ensure that the issue has been resolved:

Firefox 17 (Resolved): http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012/11/2012-11-20-mozilla-release-debug/
- Ran through Firefox 17 and didn't receive the assertion crash that was reproduced with Firefox 16. Ensured that the above assertion failure didn't appear in the log file

Firefox 17esrpre BETA (Resolved): http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012/11/2012-11-20-mozilla-esr17-debug/
- Ran through Firefox 17esrpre BETA and didn't receive the assertion crash that was reproduced with Firefox 16. Ensured that the above assertion failure didn't appear in the log file

Firefox 18 BETA (Resolved): http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2012-12-16-mozilla-beta-debug/
- Ran through Firefox 18 BETA and didn't receive the assertion crash that was reproduced with Firefox 16. Ensured that the above assertion failure didn't appear in the log file

Comment 21

[u279076]

Thank you, Kamil. I think this is sufficient to call this verified fixed. Jesse (or anyone else) feel free to revert my flag changes if there's something else we can do.

Comment 22

[Mats Palmgren (inactive)]

Crash test:
https://hg.mozilla.org/integration/mozilla-inbound/rev/cfaa2e423b1f

Comment 23

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/cfaa2e423b1f