716443 - Large Data Submission causes crash: [@ mozalloc_abort(char const* const) | mozalloc_handle_oom() | xul.dll@0x126615 ] via js::Shape::set

Status: RESOLVED WORKSFORME

(Core :: General, defect)

Description

[rtrappman]

User Agent: Mozilla/5.0 (Windows NT 6.1; rv:9.0.1) Gecko/20100101 Firefox/9.0.1
Build ID: 20111220165912

Steps to reproduce:

Submit 500+ Domains to community.opendns.com/domaintagging/submit


Actual results:

The website will process the domains for a while then randomly crashes.


Expected results:

All submissions submitted without crash.

Comment 1

[rtrappman]

Firefox Crash Reports [@ EMPTY: no crashing thread identified; corrupt dump ] is a "topcrash"

Comment 2

[rtrappman]

Finally got Firefox to provide a non null crash signature for this error by putting it in a tab group. This error creates data loss, as no information is recoverable upon crash restart.

Actual Crash Signature: 
[@ mozalloc_abort(char const* const) | mozalloc_handle_oom() | xul.dll@0x126615 ]

Crash ID:
bp-28189b85-6886-445c-a3ce-5036e2120108

Comment 3

[rtrappman]

Due to the nature of this bug adding priority "critical" like Crash Signature Bug.
Also, I have a feeling that users could get these two crash signatures on any large data submission.

Comment 4

[Scoobidiver (away)]

The corrupt dump crash signature is for anything, so it must not be taken into account.
With only 78 crashes over the last week, the xul crash signature is not a top crasher.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Hmm.  There's not much to go on in that stack... :(

Are we running out of memory while rendering the website, or while submitting the form?  That is, does the server take whatever action is involved before the crash actually happens?

Comment 6

[rtrappman]

(In reply to Boris Zbarsky (:bz) from comment #5)
> Hmm.  There's not much to go on in that stack... :(
> 
> Are we running out of memory while rendering the website, or while
> submitting the form?  That is, does the server take whatever action is
> involved before the crash actually happens?

The server processes all submission up until the point in which FireFox would crash.

Comment 7

[Boris Zbarsky [:bzbarsky]]

That doesn't quite answer my question, though....

If you look at the actual HTTP traffic, do we crash before we finish sending (or at least before we get a response from the server)?

Comment 8

[rtrappman]

(In reply to Boris Zbarsky (:bz) from comment #7)
> That doesn't quite answer my question, though....
> 
> If you look at the actual HTTP traffic, do we crash before we finish sending
> (or at least before we get a response from the server)?

The connection is still open and trying to actively send data when the crash occurs. Yes, we crash before we finish sending; The server does not respond with the category status (already in category, or new in category) for the domains submitted just before the crash.

Comment 9

[Boris Zbarsky [:bzbarsky]]

OK, thanks.  And this is with no extensions installed, right?  At least the crash reports don't list any...

Comment 10

[Wayne Mery (:wsmwk)]

 rtrappman, does this reproduce with recent nightly build?

Comment 11

[rtrappman]

(In reply to Wayne Mery (:wsmwk) from comment #10)
>  rtrappman, does this reproduce with recent nightly build?

I have not noticed this issue in the current stable release, I have not tested the nightly though.

Comment 12

[Wayne Mery (:wsmwk)]

WFM per comment 11. Please reopen if you see agina in future.