80025 - Composer crash when opening document with CSS

Status: VERIFIED INVALID

(Core :: DOM: Editor, defect, P1)

Description

[R.K.Aa.]

RH7.1, 7 hours old optimized CVS build:

crash when opening document in composer

To repro:

start mozilla
open composer
select File/Recent Pages and then select a file there
Crash. Tested this several times - same crash.

Comment 1

[R.K.Aa.]

Attachment: backtrace (non-debug)

Comment 2

[sujay]

Cc: shrir.

Comment 3

[R.K.Aa.]

tested some more: This only seems to happen when the document contains CSS
They used to load fine untill recently. A sample of a page that crashes is found
at http://home.c2i.net/dark/documents.html

Comment 4

[sujay]

I just tried this with random non-CSS pages in my recent file menu list...no
crash. works fine..

then I tried with your page with CSS...still no crash...I can load
that page from my Recent File menu..

Comment 5

[sujay]

but wait, I 'm using yeterdays build....lemme try todays...

Comment 6

[Daniel Glazman (:glazou) (not active in Mozilla any more)]

I just tried this with random non-CSS pages in my recent file menu list...no
crash. works fine..

then I tried with your page with CSS...still no crash...I can load
that page from my Recent File menu..

Comment 7

[Daniel Glazman (:glazou) (not active in Mozilla any more)]

No crash with yesterday's linux build on http://home.c2i.net/dark/documents.html

Very nice page design btw, clear and readable.

Comment 8

[R.K.Aa.]

thanks - it's one in a series of good intentions - WIP.
The build that crashes origins from a pull at around 22:55 bonsai time yesterday
(after the bidi checkins). Clobbered before build.

.mozconfig:
ac_add_options --disable-debug
ac_add_options --disable-tests
ac_add_options --enable-optimize=-O2
ac_add_options --disable-logging
ac_add_options --enable-crypto

RH7.1 gcc-2.96-81.

I'll do some tests with the optimize flags over the next builds - see what happens.

Comment 9

[sujay]

I'll also try this out in a current build...today's 5/10 build
crashes for me...

Comment 10

[sujay]

tried on today's 5/10 build...works fine...

Comment 11

[R.K.Aa.]

Hmm. I loaded it in browser and tried to open it with "open web location" to see
if that changed anything, but nothing ever opened. Odd. I'll repull and build
with -O1 now.

Comment 12

[R.K.Aa.]

This may not be related to CSS at all, allthough it makes it easyer to trigger:

 I now got the same crash once by merely opening composer. And on one occation
the document loaded in composer after i first had clicked in the composer page
(i noticed toolbar icons shifted some at that point), but a second after having
closed composer the same crash occured.

Rebuilding with -O2 first without clobbering. This is so guaranteed "crashy"
here perhaps i missed something :/

Comment 13

[rubydoo123]

assigning to kin for review and initial debug

Comment 14

[R.K.Aa.]

Since the crash seems to involve "SetDataWithLength":
I notice some warnings during build - of many alike are these:

In file included from nsPrefService.cpp:28:
nsPrefBranch.h:74:3: warning: pasting "->" and "GetData" does not give a valid
preprocessing token
nsPrefBranch.h:74:3: warning: pasting "->" and "SetData" does not give a valid
preprocessing token
nsPrefBranch.h:74:3: warning: pasting "->" and "ToString" does not give a valid
preprocessing token
nsPrefBranch.h:74:3: warning: pasting "->" and "SetDataWithLength" does not give
a valid preprocessing token
nsPrefBranch.h:74:3: warning: pasting "->" and "AdoptData" does not give a valid
preprocessing token
nsPrefBranch.h:74:3: warning: pasting "->" and "AdoptDataWithLength" does not
give a valid preprocessing token
nsPrefBranch.h:75:3: warning: pasting "->" and "GetType" does not give a valid
preprocessing token

Comment 15

[R.K.Aa.]

that last comment may not be significant at all. I compared some of the warnings
i get with the ones tinderbox reports - i probably get several thousand more
warnings than are listed there, whereof a thousand are those "does not give a
valid preprocessing token".

Comment 16

[R.K.Aa.]

no luck with this rebuild - the crash is with me.

Comment 17

[R.K.Aa.]

The crash is real enough but my compiler is not the official Mozilla one. Hyatt
just wrote in n.p.m.builds that users are to downgrade to egcs again. Since
others don't see this bug, it's obviously a problem only with my own compiler -
thus resolving as invalid.

Comment 18

[R.K.Aa.]

four letter words...five letter words.. that was LEAF writing in the newsgroup,
not at all Hyatt.

Comment 19

[sujay]

verified.

Comment 20

[R.K.Aa.]

turns out this is bug 80097. I crash there all the time, with the latest nightly
also when starting browser or adding comment in bugzilla.