833689 - Thunderbird Crash [@ nsRuleNode::WalkRuleTree(nsStyleStructID, nsStyleContext*) ] (EXCEPTION_STACK_OVERFLOW in frame construction recursion)

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Wayne Mery (:wsmwk)]

+++ This bug was initially created as a clone of Bug #724234 +++

splitting this from bug 724234  because the Thunderbird issue is a regression, potentially fixed by bug 824926.

bp-90af1fba-b8d9-42e1-b001-9b6202130117 TB17 
"I keep getting my Emails crashin when I want to send them, and I really am frustrated with Thunderbird at the moment, and am thinking of changing."
0	xul.dll	nsRuleNode::WalkRuleTree	layout/style/nsRuleNode.cpp:1817
1	xul.dll	nsRuleNode::GetStyleVisibility	layout/style/nsStyleStructList.h:61
2	xul.dll	nsIFrame::GetStyleVisibility	layout/style/nsStyleStructList.h:61
3	xul.dll	nsFrame::DidSetStyleContext	layout/generic/nsFrame.cpp:813
4	xul.dll	nsFrame::Init	layout/generic/nsFrame.cpp:535
5	xul.dll	nsSplittableFrame::Init	layout/generic/nsSplittableFrame.cpp:23
6	xul.dll	nsContainerFrame::Init	layout/generic/nsContainerFrame.cpp:65
7	xul.dll	nsInlineFrame::Init	layout/generic/nsInlineFrame.cpp:48
8	xul.dll	nsCSSFrameConstructor::InitAndRestoreFrame	layout/base/nsCSSFrameConstructor.cpp:4543
9	xul.dll	nsCSSFrameConstructor::ConstructInline	layout/base/nsCSSFrameConstructor.cpp:11102 

bp-2fd140ab-23d4-463a-b517-1e4392130121 reporter also has multiple crashes

Comment 1

[Wayne Mery (:wsmwk)]

most users' comments cite changing fonts.
so yet another sig for font crashes - making this collection an Uber Topcrash

Comment 2

[Joe Sabash [:JoeS1]]

bug 824926 "handles" the nested fonts.
but the core backout in bug 812638 prevents those redundant tags.
That backout never made it into gecko 20 afaict.

Comment 3

[Tracy Walker [:tracy]]

This bug went explosive yesterday. from 8 -> 34 -> 750 crashed per 1m ADU.

Many reports of Thunderbird crashes when changing font size in mail composition.

Comment 4

[Wayne Mery (:wsmwk)]

(In reply to Tracy Walker [:tracy] from comment #3)
> This bug went explosive yesterday. from 8 -> 34 -> 750 crashed per 1m ADU.
> 
> Many reports of Thunderbird crashes when changing font size in mail
> composition.

Thanks for the info.  I'm not sure we can trust stats on a holiday because of the ADU swings. And just eyeballing raw stats I don't see anything unusual.

That said, where can one access explosive reports?

Note, if we had a released fix for this topcrash we wouldn't be having this discussion :)

Comment 5

[Tracy Walker [:tracy]]

Wayne, I actually ran across this is Kairo's Fx explosiveness reports, which he lists here: https://crash-analysis.mozilla.com/rkaiser/0000.overview.html.

Comment 6

[Wayne Mery (:wsmwk)]

(In reply to Tracy Walker [:tracy] from comment #5)
> Wayne, I actually ran across this is Kairo's Fx explosiveness reports, which
> he lists here: https://crash-analysis.mozilla.com/rkaiser/0000.overview.html.

ah thanks. so it is - https://crash-analysis.mozilla.com/rkaiser/2013-03-31/2013-03-31.firefox.aurora.explosiveness.html ... I didn't think such reports existed for thunderbird.

Comment 7

[Wayne Mery (:wsmwk)]

(In reply to Joe Sabash from comment #2)
> bug 824926 "handles" the nested fonts.
> but the core backout in bug 812638 prevents those redundant tags.
> That backout never made it into gecko 20 afaict.

what is needed to get this fixed this in *THUNDERBIRD 17* ?
does bug 855898 fix only the explosive bit?  or the whole shootin match?

(not "picking on" standard8, but I don't know who else to ask)

Comment 8

[Mark Banner (:standard8)]

My first question would be have we been able to verify that this has been fixed and if so, in which version?

Comment 9

[Wayne Mery (:wsmwk)]

(In reply to Mark Banner (:standard8) from comment #8)
> My first question would be have we been able to verify that this has been
> fixed and if so, in which version?

A fair question.  It was unclear to me if Neil's patch was going to address the crash issue because no developers commented directly on the crash aspect, and for some odd reason I don't think we ever had users of non-release builds crashing.

But bug 824926 *finally* landed on TB17 and crash is now #41 in TB17.0.5 whereas in prior releases it was solid at #6.  But 
- we've only shipped it a little over a week ago, so we may need to look again in a few more days
- I'd like a second opinion from someone (ludo, mark) of whether  the remaining crashes https://crash-stats.mozilla.com/query/query?product=Thunderbird&version=Thunderbird%3A17.0.5esr&version=Thunderbird%3A17.0.5&version=Thunderbird%3A23.0a1&version=Thunderbird%3A22.0a2&version=Thunderbird%3A22.0a1&range_value=6&range_unit=weeks&date=04%2F10%2F2013+12%3A07%3A27&query_search=signature&query_type=exact&query=nsRuleNode%3A%3AWalkRuleTree%28nsStyleStructID%2C+nsStyleContext*%29&reason=&build_id=&process_type=any&hang_type=any&do_query=1 are due to McAfee - 4 of 5 [1] that I randomly chose have the extension.


As a post mortem (assuming this crash is winding down) this is another example of a thunderbird problem that was difficult to assess before bug 824926 landed in an actual release because

1. developer comments to date in related bug reports pretty much focused on firefox  - so there was no insight from developers on the effect on Thunderbird 

2. for various reasons it was difficult to judge the effect of patches on thunderbird in non-release branches 

I assume bug 802997, reported almost 6 months ago, is related.  Someone should check that it and other bugs' crash rate to see that they are also reduced


[1] 
https://crash-stats.mozilla.com/report/index/5bb1e075-3f63-4e08-8085-499eb2130410
https://crash-stats.mozilla.com/report/index/f854e119-0adb-4947-8b39-b78a72130410
https://crash-stats.mozilla.com/report/index/08287895-6702-4528-b151-22edc2130410
https://crash-stats.mozilla.com/report/index/7c12037d-4e71-48b0-94fc-205242130409
https://crash-stats.mozilla.com/report/index/6774515f-fd6e-490e-b72d-4095b2130409

Comment 10

[Wayne Mery (:wsmwk)]

Followup to comment 9...

A week later situation still as described.  TB17.0.5 Crash rank #38.  About 70% of crashes have McAfee addon (bug 803460).  I've pinged some users who do and don't have McAfee to get some feedback. 

So fixed by bug 824926. And if needed will file follow up bug for any remaining crashes.