274281 - [BeOS] BeOS lets events stay in port through the 'restarts' in LaunchChild causing crashes on faulty events.

Status: RESOLVED FIXED

(Toolkit :: Startup and Profile System, defect)

Description

[tqh]

User-Agent:       Mozilla/5.0 (BeOS; U; BeOS BePC; en-US; rv:1.7.5) Gecko/20041212 Firefox/1.0
Build Identifier: Mozilla/5.0 (BeOS; U; BeOS BePC; en-US; rv:1.7.5) Gecko/20041212 Firefox/1.0

In plevent.c as well as in nsAppShell BeOS always tries to reuse a port for
eventqueues if it's open. As the port survives the 'restarts' in LaunchChild it
may (and does) contain events with faulty pointers. This causes Firefox to crash
after profile creation.

Reproducible: Always
Steps to Reproduce:
1. Remove any current profile for Firefox
2. Launch Firefox, and see it crash.


Actual Results:  
Firefox crashes after profile creation.

Expected Results:  
Firefox should have launched.

MethodInfo::Invoke(void):
Invoke__10MethodInfo:
+001e ef0cd19a: * 388b movl (%eax), %edi
firefox-bin:sc
frame retaddr
fcffbe5c ef0cbdd2 nsAppShell::Run(void) + 0000013e
fcffbecc ee6cf361 nsAppShellService::Run(void) + 00000051
fcffbf0c 80020822 xre_main(int, char **, nsXREAppData const *) + 00001f32

Comment 1

[tqh]

Attachment: First part: plEvent patch

Better way to handle the port in plevent.c. BeOS only changes, as it's inside
ifdefs.

Comment 2

[tqh]

Attachment: Second part: nsAppShell patch

Better way of setting upp the port in nsAppShell.

Comment 3

[Sergei Dolgov]

Tqh, what do you think about fact that plevent and nsAppShel do use atm port
with different capacities?

Comment 4

[tqh]

(In reply to comment #3)
> Tqh, what do you think about fact that plevent and nsAppShel do use atm port
> with different capacities?

Don't they use the same port? It looks so to me.

Comment 5

[tqh]

Attachment: plEvent patch w. comments and smaller size port

Comment 6

[tqh]

Attachment: nsAppShell patch w. comments and smaller port size

Comment 7

[tqh]

Comment on attachment 168537 [details] [diff] [review]
First part: plEvent patch

removing review on old patches.

Comment 8

[tqh]

Attachment: plEvent patch with better formatting and comments

Comment 9

[tqh]

Attachment: nsAppShell patch w better formatting and comments

Comment 10

[Sergei Dolgov]

Comment on attachment 168546 [details] [diff] [review]
plEvent patch with better formatting and comments

r=[e-mail]

I think we need here either sr and approval too, inspite BeOS-only character of
changes, as this file is outside BeOS folders

Comment 11

[Sergei Dolgov]

Comment on attachment 168548 [details] [diff] [review]
nsAppShell patch w better formatting and comments

r=[e-mail]

Pure BeOS-port patch.
No need for sr.

Comment 12

[tqh]

Comment on attachment 168546 [details] [diff] [review]
plEvent patch with better formatting and comments

sr?

Comment 13

[Mike Shaver (:shaver emeritus)]

Comment on attachment 168546 [details] [diff] [review]
plEvent patch with better formatting and comments

>+    	if(self->eventport < 0) { 
>+              self->eventport = create_port(200, portname);
>+            }

Indentation here needs to be made sane.  Otherwise, this looks
like it can only affect BeOS, so I will happily stamp on the
basis of sergei's review.

Comment 14

[tqh]

Attachment: plEvent.c Better formatting otherwise same as SR'd

Comment 15

[Christian :Biesinger (don't email me, ping me on IRC)]

Checking in widget/src/beos/nsAppShell.cpp;
/cvsroot/mozilla/widget/src/beos/nsAppShell.cpp,v  <--  nsAppShell.cpp
new revision: 1.27; previous revision: 1.26
done
Checking in xpcom/threads/plevent.c;
/cvsroot/mozilla/xpcom/threads/plevent.c,v  <--  plevent.c
new revision: 1.45; previous revision: 1.44
done


please mark fixed if it is.