230764 - go to weather.com, get flash ad ( without flash plugin ), see mozilla die

Status: RESOLVED INVALID

(Core Graveyard :: GFX: Gtk, defect)

Description

[Drew "DanteAliegri" Ogle]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.5) Gecko/20031225 Firebird/0.7
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7a) Gecko/20040110

Flash is killing mozilla,
it is when a page first loads,
I'm not sure if it is when going somewhere after being at a place with flash, or
going into a place with flash.

Reproducible: Always

Steps to Reproduce:
1.goto weather.com without flash plugin installed.
2.reload a few times.
3.see mozilla die.

Actual Results:  
mozilla crashes.

Expected Results:  
mozilla not dead, with a page blissfuly absent of flash.

May be limited to linux gtk2 ( optimized? ) builds. haven't tested.

Comment 1

[Drew "DanteAliegri" Ogle]

Attachment: Backtrace of crash

Comment 2

[Olivier Cahagne]

do you also crash with a GTK1 build ?
http://ftp.mozilla.org/pub/mozilla.org/firebird/nightly/latest-trunk/

Comment 3

[Drew "DanteAliegri" Ogle]

No, I do not crash with a gtk1 build.
I had be using my own build, but I also nabbed a gtk2 FB build when getting the
gtk1 tar, and that crashes as well, so it is not my build options.

Comment 4

[Christopher Blizzard (:blizzard)]

Yuck.  Could be flash, could be us.  Hard to say.

Comment 5

[Andrew Schultz]

> Flash is killing mozilla,

> 1.goto weather.com without flash plugin installed.

to be perfectly clear... you don't have the flash plugin installed.  When you
visit weather.com, you get the plugin puzzle piece instead of a flash animation,
right?

Comment 6

[Drew "DanteAliegri" Ogle]

Correct. I load weather.com. I get the "Plugin not installed Press OK to get it".
I press either one, weather.com loads with the puzzle piece.

I was flourishing a bit by saying "Flash is killing mozilla" .. what I should
have said was "the use of flash in webpages seems to be killing mozilla".

Comment 7

[Drew "DanteAliegri" Ogle]

on a hunch, I downgraded gtk/glib/pango to 2.2, and the crash no longer happens.

Comment 8

[Andrew Schultz]

worksforme with linux turnk CVS with gtk2/xft
can you build with symbols and grab a stack trace?

Comment 9

[Drew "DanteAliegri" Ogle]

Attachment: Backtrace of crash, with debugging symbols

Comment 10

[Andrew Schultz]

> on a hunch, I downgraded gtk/glib/pango to 2.2, and the crash no longer happens.

what had you been using before?  2.4?

Comment 11

[Drew "DanteAliegri" Ogle]

well, there is no gtk 2.4 yet -- I was using 2.3, if that is what you meant.

Comment 12

[otaylor]

The problem here is that the logic of when to call XSync() in gtk2xtbin.c
is not correct, so changes in exact sequences of operation or timing with
GTK+-2.3 cause it to crash.

XSync should be called:

 - When a resource has been created in GTK+, and the code needs to make
   sure that it exists on the server before making an Xt call on a different 
   display (call XSync on the GTK+ display by calling gdk_flush())

 - When a resource has been created in Xt, and the code needs to make
   sure that it exists on the server before making a GDK call on a different
   display (call XSync on the Xt display)

 - Before removing an error trap when X calls have been made that are
   not round trip. (I don't believe the code actually needs error 
   traps when the first two rules are followed, however.)

Comment 13

[Stephanie Daugherty [:sdaugherty]]

This may be a dupe of bug 169322

Comment 14

[Andrew Schultz]

Drew: Are you still seeing this? (WFM with gtk 2.10)

Comment 15

[Drew "DanteAliegri" Ogle]

No, I haven't been able to recreate this since mid-2005.

Comment 16

[Stephanie Daugherty [:sdaugherty]]

No longer reproduceable. Please comment or reopen if this problem returns.
-> INVALID