505059 - [FIX]Crash [@ CSSLoaderImpl::LoadSheet(SheetLoadData*, StyleSheetState) ]

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect)

Description

[Michael Kraft [:morac]]

Attachment: crash when trying to install this while offline. May not be trigger though.

My network connection was down (I disabled the wireless network card on my laptop) and any attempt to install the attached addon resulted in a crash.  The add-on did not install.  Another add-on installed without issue.

Once I enabled the network, everything was fine.

http://crash-stats.mozilla.com/report/index/7b9dfade-be82-4efa-a264-bdcfc2090718

Comment 1

[Jesse Ruderman on Windows]

dbaron, can you figure this out based on the stack trace?

Comment 2

[Boris Zbarsky [:bzbarsky]]

The above could perhaps happen if someone calls into the stylesheet service from far enough into shutdown that the layout module has already shut down...  or something.

Comment 3

[Jesse Ruderman on Windows]

This is actually a fairly frequent crash (300 per week), so I'm not going to mark it as INCO just yet.

Comment 4

[Jesse Ruderman on Windows]

Filed bug 524256 in the hope of getting better stack traces.

Comment 5

[Jesse Ruderman]

3.7a1pre stacks are good now.  It's definitely happening during shutdown:

http://crash-stats.mozilla.com/report/index/180ab197-098b-48a2-b378-5ab9e2091113

bz, does this help more?

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Sounds like what Boris said in comment 2.  A JS stack might be interesting.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Yeah, just having some idea of what js is running would be good...

I wonder what some of these extensions are.

That said, maybe we should just bulletproof things against this eventuality (e.g. null-check in the stylesheet service) and be done with it?

Comment 8

[Boris Zbarsky [:bzbarsky]]

Or keep the layout module alive if someone's holding a ref to the stylesheet service... we do that for DOM nodes already, iirc.

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Keeping the layout module alive sounds fine.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Hmm.  Then again, after xpcom-shutdown we'll have a null security manager here.  I think we should probably just bulletproof the cssloader or stylesheet service and be done with it.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Nevermind comment 10; I was misreading code.

Comment 12

[Boris Zbarsky [:bzbarsky]]

Attachment: Fix

Comment 13

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 413026 [details] [diff] [review]
Fix

r=dbaron

Comment 14

[Boris Zbarsky [:bzbarsky]]

Pushed http://hg.mozilla.org/mozilla-central/rev/9e225083676f

I'd want to bake this for a bit before landing on 1.9.2, assuming we do that at all.

Comment 15

[Boris Zbarsky [:bzbarsky]]

And http://hg.mozilla.org/mozilla-central/rev/b55fcd30070a to fix build bustage.

Comment 16

[Marcia Knous [:marcia]]

Looks as if a few users on Windows and Mac hit this crash in 3.6 B5 -> http://tinyurl.com/yd6e3ra.

Comment 17

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 413026 [details] [diff] [review]
Fix

I guess we've baked this enough to take on branch at this point, yeah.

Comment 18

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

Comment on attachment 413026 [details] [diff] [review]
Fix

We should take this for 1.9.2.3

Comment 19

[Daniel Veditz [:dveditz]]

Comment on attachment 413026 [details] [diff] [review]
Fix

Approved for 1.9.2.4, a=dveditz for release-drivers

Comment 20

[Boris Zbarsky [:bzbarsky]]

Pushed http://hg.mozilla.org/releases/mozilla-1.9.2/rev/5d64ab943a25

Comment 21

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

This reads a lot like a top crasher with no way to verify it. We never had a reproduction scenario for this, did we, we just fixed the underlying issue?