119038 - Assertion: document demuxed from FastLoad file more than once?: 'entry->mNextSegmentOffset != 0' nsFastLoadFile.cpp#527

Status: RESOLVED FIXED

(Core :: XUL, defect)

Description

[timeless]

I'll attach the stack trace and nspr log.

Comment 1

[timeless]

Attachment: assertion stack

Comment 2

[timeless]

i accidentally let mozilla clobber my log leaving me with a shorter and less 
interesting log, i'll try again later.

Comment 3

[John Morrison]

Oh, wait. Is this only FreeBSD? Is this the trunk? Does this happen on each 
startup (or every second?).

Comment 4

[John Morrison]

OK, I'll answer my own question. In my opt. build [with a printf for the 
assertion], I get this condition every second time I start the browser on 
Linux.

I suspect that this may be [part of] the problem with window and pageload 
regression on tinderboxen (see bug 119066).

Comment 5

[John Morrison]

This applies to today's comm. builds on Mac and win32 as well.

Comment 6

[timeless]

ok, then i'm not going to keep ownership of this bug :) yes cvs tip.

Comment 7

[Brendan Eich [:brendan]]

Zoiks.  Does all become well if harishd's changes are backed out?

All this stuff is fragile, as the best hacks tend to be.

/be

Comment 8

[John Morrison]

> Does all become well if harishd's changes are backed out?

I'll answer in the opposite direction. A build pulled by date Jan 8 17:36 PST,
just before harishd's first batch of checkins, did not blow away XUL.mfasl on 
every second startup (with that assertion showing). When I applied harishd's 
first flight of checkins, at 17:37 PST, and rebuilt from the top, this was no
longer the case; XUL.mfasl would be blown away on every second startup with 
that assertion.

(Sidenote: but for 1MB of bloat reduction and faster startup, these are the 
kind of regressions I almost kinda like tracking down :-]).

Comment 9

[Brendan Eich [:brendan]]

harishd is loggin WillResume calls against Start/EndMuxedDocument (WillResume
calls SelectMuxedDocument), and comparing with and without his patch.  I suspect
something in how WillResume is called changed.

/be

Comment 10

[Brendan Eich [:brendan]]

Harish is all over this, thanks to my WillResume suspicion.

/be

Comment 11

[harishd]

Attachment: patch v1.1 [ fix assertion ]

Comment 12

[harishd]

Attachment: patch v1.2 [ Removed default values from WillResumeParse() & WillInterruptParse() ]

Comment 13

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Comment on attachment 64416 [details] [diff] [review]
patch v1.2 [ Removed default values from WillResumeParse() & WillInterruptParse() ]

r=heikki

Comment 14

[Brendan Eich [:brendan]]

Comment on attachment 64416 [details] [diff] [review]
patch v1.2 [ Removed default values from WillResumeParse() & WillInterruptParse() ]

sr=[e-mail]

Comment 15

[harishd]

fixed.