5490 - multiple BROWSER windows garbled, cause eventual crash

Status: VERIFIED FIXED

(Core Graveyard :: Tracking, defect, P1)

Description

[cpratt]

build id: 1999042608
bug occurs on windows nt 4 sp 4, mac os 8.5.1
not tested on linux (yet)

to reproduce this:
launch apprunner. open a local file. open a second local file.
result: crash before the app finished rendering the second opened page.
expected result: no crash.

talkback tracking ID: DZK45LUQ

this happens regardless of whether or not the second file | open selection took
place in the first or subsequent window opened by apprunner.

Comment 1

[don]

Assigned to law.

Comment 2

[Bill Law]

I think this is yet another "shared XUL document" problem.  I'll debug it
further tomorrow and decide what to do with it.

Comment 3

[don]

Re-assigned to trudelle.

Peter, Bill says this looks to be caused by the "shared XUL document" problem
that Hyatt and Waterson are working on fixing.

Comment 4

[Peter Trudelle]

reassigning to hyatt as p1 for m6, cc waterson

Comment 5

[Bill Law]

*** Bug 5535 has been marked as a duplicate of this bug. ***

Comment 6

[David Hyatt]

I don't see how this could be related to the shared XUL document problem.  The
crash happens even if two files are opened in the same apprunner window. (Two
HTML documents, but both loaded inside the same XUL document.)  Am I missing
something?

Comment 7

[chris hofmann]

I'm not sure but the stack trace might be in this talkback report.

http://cyclone/reports/incidenttemplate.CFM?reportID=124&style=0&tc=26&cp=3&ck1=
SUser+email+address&cd1=%25cpratt%40netscape%2Ecom%25&co1=like&bbid=6729450


  raptorwidget.dll + 0x6f40 (0x01836f40)
  raptorwidget.dll + 0x5a7e (0x01835a7e)
  raptorwidget.dll + 0xa98c (0x0183a98c)
  USER32.dll + 0x131f (0x77e
  USER32.dll + 0x1519d (0x77e8519d)
  ntdll.dll + 0x1637b (0x77f7637b)
  raptorwidget.dll + 0xa32f (0x0183a32f)
  nsappshell.dll + 0x1593 (0x013a1593)
  apprunner.exe + 0x1bff (0x00401bff)
  KERNEL32.dll + 0x1ba3c (0x77f1ba3c)

Comment 8

[David Hyatt]

Here's the crash.  It's in nsViewManager.cpp.

nsViewManager::RenderViews(nsIView * 0x051f8500, nsIRenderingContext & {...},
const nsRect & {...}, int & 0) line 1199 + 30 bytes
nsViewManager::Refresh(nsIView * 0x051f8500, nsIRenderingContext * 0x052ab610,
const nsRect * 0x0084f540, unsigned int 1) line 521
nsViewManager::DispatchEvent(nsViewManager * const 0x051f8ec0, nsGUIEvent *
0x0084f5c8, nsEventStatus & nsEventStatus_eIgnore) line 1643
HandleEvent(nsGUIEvent * 0x0084f5c8) line 67
nsWindow::DispatchEvent(nsWindow * const 0x051f85d4, nsGUIEvent * 0x0084f5c8,
nsEventStatus & nsEventStatus_eIgnore) line 414 + 10 bytes
nsWindow::DispatchWindowEvent(nsGUIEvent * 0x0084f5c8) line 435
nsWindow::OnPaint() line 2811 + 24 bytes
nsWindow::ProcessMessage(unsigned int 15, unsigned int 0, long 0, long *
0x0084f798) line 2181 + 17 bytes
nsWindow::WindowProc(void * 0x000007a0, unsigned int 15, unsigned int 0, long 0)
line 478 + 27 bytes
KERNEL32! bff7363b()
KERNEL32! bff942e7()

Kipp, are you the current view expert now that michaelp is gone?

The code in nsViewManager.cpp looks dangerous to me.  I can see why it's
crashing.  On line 1141, the following test is made...

if ((localrect.width > gBlendWidth) || (localrect.height > gBlendHeight))

If that condition isn't met, then the mOffscreenCX isn't initialized, so that
might be a way to end up with a null context.

Anyway, this window problem (with garbling in the additional windows) began
showing up recently.

Comment 9

[David Hyatt]

This also only occurs with the browser window.... messenger and composer
work with multiple windows just fine... which I find very strange.

Comment 10

[David Hyatt]

*** Bug 5499 has been marked as a duplicate of this bug. ***

Comment 11

[Chris Waterson]

I have a fix for this bug in hand. We are adding the RDF local store to each
RDF composite datasource, which causes assertions to be shared across _all_
windows. I want to remove this so that each window's content model will be
private.

Comment 12

[Chris Waterson]

Fix checked in, a=chofmann

Comment 13

[leger]

Putting on QA Blocker radar in case fix not in across all platforms.

Comment 14

[cpratt]

not crashing using the 1999042908 build on nt or linux.
there are problems on the mac os, but they are (apparently) not related to this
issue.
marking this as verified fixed, opening new bugs for the mac problem and for not
drawing anything in the content are of new windows.

Comment 15

[leger]

Moving all Apprunner bugs past and present to Other component temporarily whilst
don and I set correct component.  Apprunner component will be deleted/retired
shortly.