200373 - OS/2- crash when Chatzilla unable to steal focus from other apps

Status: RESOLVED FIXED

(Core :: DOM: UI Events & Focus Handling, defect)

Description

[Felix Miata]

This is a spinoff of bug 199999 that happens in all OS/2 trunk builds tested,
2003031312 and newer, but not in W32 builds. See bug 199999 for repro.

Comment 1

[Mike Kaply [:mkaply]]

mine

Comment 2

[Mike Kaply [:mkaply]]

Attachment: Hack fix

Check session ID and don't set focus if over 0xFF (non PM session)

Comment 3

[jhp (no longer active)]

Comment on attachment 119215 [details] [diff] [review]
Hack fix

Did you try this with both DOS and Win-OS/2 sessions?  What about a
non-fullscreen DOS session?  What happens then?

If this patch works for all of those, then r=pedemont.

Comment 4

[Mike Kaply [:mkaply]]

I checked all those things including a Win-OS/2 window and everything worked fine.

Comment 5

[Felix Miata]

Fixed when? This is not fixed in 2003040120:

04-03-2003  10:02:32  SYS3171  PID 013f  TID 0001  Slot 0071
H:\MOZILLA\BIN\MOZILLA.EXE
c0000005
1d8988f1
P1=00000002  P2=0003fff4  P3=XXXXXXXX  P4=XXXXXXXX  
EAX=0004003c  EBX=1dc3d5e4  ECX=00ce08e0  EDX=00000000
ESI=00000000  EDI=01c685c0  
DS=0053  DSACC=f0f3  DSLIM=ffffffff  
ES=0053  ESACC=f0f3  ESLIM=ffffffff  
FS=150b  FSACC=00f3  FSLIM=00000030
GS=0000  GSACC=****  GSLIM=********
CS:EIP=005b:1d8988f1  CSACC=f0df  CSLIM=ffffffff
SS:ESP=0053:0003fff8  SSACC=f0f3  SSLIM=ffffffff
EBP=00040000  FLG=00012212

GKLAYOUT.DLL 0001:000b88f1

Comment 6

[Felix Miata]

It happened again in 2003052112.

Comment 7

[Felix Miata]

It happened again in 2004010408.

Comment 8

[Mike Kaply [:mkaply]]

I still think this is a problem with your machine or some enhancement you have
installed.

Comment 9

[Felix Miata]

I have no enhancements installed. I just replaced MCP with eCS 1.1 prior to my
comment 7, so highly unlikely it is the particular OS installation/configuration.

I just tried on my Socket 7 SiS530 and Intel TX chipset boxen, and couldn't
repro there. Since I have an Athlon XP box almost ready to switch this HD into,
I should soon confirm if it's hardware related.

If it is hardware, the known buggy VIA MVP3 motherboard chipset may be the
culprit. I think I know of a couple I can get to try, but it might take a while
(if lucky, within a month) to do it. 

Comment 10

[Mike Kaply [:mkaply]]

Per reporter, this is working now.

Comment 11

[Jason Bassford]

Nothing appears to have been checked in, and no patch or other bug referenced as
the fix.

->WORKSFORME

Comment 12

[Felix Miata]

http://bugzilla.mozilla.org/attachment.cgi?id=119215&action=view

Comment 13

[Felix Miata]

Works in the GCC trunk, but not in VACPP 1.4.2 20040316 w/ CZ 0.9.61. Attempting
to reach http://127.0.0.2 works, but the following is from CZ inability to steal:

03-17-2004  10:11:34  SYS3171  PID 0101  TID 0001  Slot 004a
H:\MOZ14\BIN\MOZILLA.EXE
c0000005
1d6ac238
P1=00000002  P2=0004fff8  P3=XXXXXXXX  P4=XXXXXXXX  
EAX=00b96020  EBX=1da3d710  ECX=000500cc  EDX=174e63bc
ESI=00050178  EDI=00b96020  
DS=0053  DSACC=f0f3  DSLIM=ffffffff  
ES=0053  ESACC=f0f3  ESLIM=ffffffff  
FS=150b  FSACC=00f3  FSLIM=00000030
GS=0000  GSACC=****  GSLIM=********
CS:EIP=005b:1d6ac238  CSACC=f0df  CSLIM=ffffffff
SS:ESP=0053:0004fffc  SSACC=f0f3  SSLIM=ffffffff
EBP=000500a4  FLG=00012202

GKLAYOUT.DLL 0001:0011c238

Comment 14

[Felix Miata]

Not VACPP only after all. From 2004031608 (trunk)

03-17-2004  12:15:37  SYS3171  PID 011b  TID 0001  Slot 008f
H:\MOZILLA\BIN\MOZILLA.EXE
c0000005
1de8e895
P1=00000002  P2=0002ff04  P3=XXXXXXXX  P4=XXXXXXXX  
EAX=00000000  EBX=00a368c0  ECX=00c3bfa0  EDX=17a93d28
ESI=00000004  EDI=00030aec  
DS=0053  DSACC=f0f3  DSLIM=ffffffff  
ES=0053  ESACC=f0f3  ESLIM=ffffffff  
FS=150b  FSACC=00f3  FSLIM=00000030
GS=0000  GSACC=****  GSLIM=********
CS:EIP=005b:1de8e895  CSACC=f0df  CSLIM=ffffffff
SS:ESP=0053:0002fee0  SSACC=f0f3  SSLIM=ffffffff
EBP=00030118  FLG=00012246

GKLAYOUT.DLL 0001:0026e895

I guess I wasn't switching from freshly opened CZ to the DOS session fast enough
before.

Comment 15

[knut st. osmundsen]

I guess that's a stack underflow in thread one.
(Hey guys didn't I tell you to use the SAME stacksize as they do on WINDOWS???)

Comment 16

[Mike Kaply [:mkaply]]

Attachment: Better patch

More thorough patch based on input from OS/2 team.

Still a hack

Comment 17

[Mike Kaply [:mkaply]]

Attachment: Remove hack fix

Noone can recreate this anymore.

I want to remove my original hack fix so that if someone sees this problem, we
can recreate it easier.

Maybe it was fixed in OS/2 and we will never see this again.

After I remove the hack, I am marking this WFM.

Comment 18

[Mike Kaply [:mkaply]]

"fix" removed from trunk.