173234 - Mozilla crashes [@ GDI32.DLL][@ nsDrawingSurfaceWin::Lock]

Status: RESOLVED FIXED

(Core :: Print Preview, defect, P1)

Description

[Peter Thomassen]

Hi,

go to http://www.riemenschneider-gymnasium.de/, watch the Print Preview and look
what happens.
I think it's caused by the overflow property of one of the divs.

Peter

Comment 1

[Olivier Cahagne]

please mention build ID, Talkback ID for this crash
(bin/mozilla/bin/components/talkback.exe) and attach reduced testcase (if possible).

Comment 2

[Peter Thomassen]

Build ID: 2002091014
Talkback ID: TB12241606Y

Sorry, no time for a reduced testcase.

Comment 3

[Tuomo Tikkanen]

I can confirm the same behaviour (crash) with:

Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:1.1) Gecko/20020827

Last few lines from "truss" output on that platform.

19334:  read(5, "\0\0\001\003\001\0\0\0\0".., 2688)     = 1232
19334:  read(5, "\0\0\0\0\0\0\0\0\0\0\0\0".., 1456)     = 1456
19334:      Incurred fault #6, FLTBOUNDS  %pc = 0xFF351A58
19334:        siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000
19334:      Received signal #11, SIGSEGV [caught]
19334:        siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000
19334:  sigprocmask(SIG_SETMASK, 0xFEAEEFF0, 0x00000000) = 0
19334:  unlink("/home/..../lock") = 0
19334:  _exit(11)

Comment 4

[Peter 'vt100' Schwindt]

TB ID TB12252579K, Build ID 2002100804, Win2k SP2

Comment 5

[greer]

Attachment: Stack from 12252579

Comment 6

[Roland Mainz]

Tuomo Tikkanen wrote:
> I can confirm the same behaviour (crash) with:
> Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:1.1) Gecko/20020827
> Last few lines from "truss" output on that platform.

"truss" does not help much here - but using "/usr/proc/bin/pstack core" on the
resulting "core" file would deliver a detailed stack trace...

Comment 7

[Olivier Cahagne]

not crashing using debug build 20020920 on Linux (trunk) but here're assertions
I get when loading/print previewing page:

###!!! ASSERTION: nsBlender::Blend() called with nsnull aSrc: 'aSrc', file
nsBlender.cpp, line 139
Break: at file nsBlender.cpp, line 139
###!!! ASSERTION: nsBlender::Blend() called with nsnull aDst: 'aDst', file
nsBlender.cpp, line 140
Break: at file nsBlender.cpp, line 140
WARNING: NS_ENSURE_TRUE(aSrc) failed, file nsBlender.cpp, line 141
WARNING: Blend failed!, file nsViewManager.cpp, line 1258

Comment 8

[rods (gone)]

don, a fun one for you.

Comment 9

[Roland Mainz]

Dumb question:
|NS_ENSURE_TRUE()| is working even in non-debug builds, right ?

Comment 10

[dcone (gone)]

Attachment: Checks for a valid bitmap before proceeding with the lock.

This will fix the crash, there is still and issue with support (or not
supporting) the blending code for print preview because the printers do not
support offscreen drawing surfaces.. this can cause a problem.	Printing does
not use this code path, but the print preview still does.

Comment 11

[rods (gone)]

Comment on attachment 102574 [details] [diff] [review]
Checks for a valid bitmap before proceeding with the lock.

r=rods

Comment 12

[Roland Mainz]

I thought I fixed this issue long ago.

The only possible issue may be that NS_ENSURE_TRUE() is a NO-OP in non-debug
builds... - can anyone confirm this ?

Comment 13

[Roland Mainz]

The bug I was thinking of is bug 161364 ("Print preview crashes when page has
transparent content") - but it looks the macro is working even for non-debug
builds... weired...

Comment 14

[kinmoz]

Comment on attachment 102574 [details] [diff] [review]
Checks for a valid bitmap before proceeding with the lock.

==== Should we be checking for valid |mDIBits| and |mBitmapInfo| pointers
before we use them after this call?

+	     mBitmapInfo = CreateBitmapInfo(mBitmap.bmWidth, mBitmap.bmHeight,
mBitmap.bmBitsPixel, (void **)&mDIBits);

Comment 15

[dcone (gone)]

Attachment: Updated patch

Checkin for success of CreateBitmapInfo()

Comment 16

[rods (gone)]

Comment on attachment 103179 [details] [diff] [review]
Updated patch

r=rods

Comment 17

[kinmoz]

Comment on attachment 103179 [details] [diff] [review]
Updated patch

sr=[e-mail]

Comment 18

[Olivier Cahagne]

patch ready to be checked in ?

Comment 19

[dcone (gone)]

Fixed