Apache OpenOffice (AOO) Bugzilla – Issue 6472
Writer crashes GDI on a multimon Win2k
Last modified: 2003-09-08 16:56:16 UTC
From Philippe Auphelle, pauphelle@irci.fr (Do not hesitate to email for further details, experiment requests, or any help I could provide in reproducing / searching this issue) Running the French edition of OpenOffice 1.0 for Windows, under Win2k sp2. Random but recurring problem, that I have seen happening within a few minutes to 30 minutes of working with OpenOffice 1.0 The hardware configuration is a very stable, three years old Dual PIII 600, multi-headed configuration (multi-monitor, or "MultiMon" in Microsoft / MSDN / Technet parlance). The video board is a Matrox 450 DualHead PCI, using the latest stable driver. The primary screen is the one on the left. The problem described only happens with OpenOffice. Never seen it before. Description of the problem: While working in OpenOffice, the whole desktop (both screens) become all at once totally garbled, mainly black. OpenOffice displays a dialog box titles "Fatal Application Exit" that says "Application Error". When pushing the OK button, OpenOffice exits. Simultaneously to the occurence of the problem, it happens that some other application pop up a "resources low" dialog box (never seen that before under Win2k). At that point, the desktop is completely garbled (both screens): other apps, windows, icons. The GDI is unstable and "GDI error" messages start to appear occasionnally, until the machine is rebooted. It seems that the Windows subsystem is partially lost in its coordinates system (some windows objects are repainted at the wrong places). Login off / on ain't enough to recover. There is still plenty of RAM still available in the machine (task manager reports 350 Mb of available memory out of 512), but at this point, the GDI is corrupted / unstable. A collegue of mine is seeing a very similar problem while running Win98 SE, in a multimon config too, but using two non-Matrox single monitor PCI boards. We have never seen this problem here on non-multimon setups, thus or suspicion about the multimon environment. So far, I have seen this crash happening while I was working in various OpenOffice configuration screen rather than on the document windows. Latest occurence, I was configuring something in the "Bullets / numbering " screen. Finally and incidentally, it appears that OpenOffice is not "multimon aware": when I have dragged the main window of OpenOffice on the left (secondary) screen and I pull down a menu, the pulled down window appears on the left of the primary (right) screen. This is frequent behavior for non-multimon-aware applications, but it normally doesn't create any further problem. In other words, Windows is designed to handle non-multimon-aware apps (lotsa them) properly, in an acceptably degraded way. And *officially*, according Microsoft, a regular non-multimon aware Windows program can't crash anything on a multimon system!... FWIW, Windows Multimon programming information in MSDN: http://msdn.microsoft.com/library/default.asp?url=/library/en-us/gdi/monitor_53sj.asp P.S.: No, the machine ain't haunted nor virused! (up-to-date Norton AntiVirus).
Erratum / addendum: The primary screen is the one on the RIGHT, and quite logically, the secondary screen is the one on the LEFT.
Philippe, thank you for using and supporting OOo. One problem per issue please. In regards to the menu issue you mentioned, please see issue 3428. Does this problem still exist with the latest version of OOo?
Yes, I had it just the same using the French version of 1.0.1 Sorry for the other (menu on dual-headed) issue. I just mentionned it because I thought it could be related somehow, but I swear I won't do that again 8-) It sure could help if the "application error" dialog box that OpenOffice opens before dying could be associated with things like a stack trace, a snapshot dump or something. But since OOo actually internally traps its own error at about the time it garbles the GDI, I can't even get a DrWatson or some equivalent. If there is anything I can do to help you track the problem, just tell me, and I'll do my best. Thanks for your help and great work! /Ph.
If possible check to see if this problem still exists in developer build 643. You may also want to ensure you have the latest drivers for your video card and multi-head display system.
The Matrox driver version is 5.84.23.0, which until 23Sep02 was the latest and greatest non-beta version.. But I did have the problem just the same with previous versions. I will 1) update my driver to the brand new "latest", i.e. 5.86.032. 2) Download developer build 643, install it, check, and keep you posted.
Just a quick note Philippe. Developer builds are unstable and meant for testing only.
Sure, but since using my current setup, the regular build crashes my station within less than 20 minutes, I'm not sure the developer build can be much more unstable that this!
I installed the OOo 1.0.3 US version, and I have been using it for several hours on each of the two strictly identical machines that were having the problem. The GDI crash hasn't occured a single time. Looks real good to me, since with previous releases, I would never run more than 20 minutes without a crash. So I would tend to think that you have somehow swated this one! For the record: - In the end, I didn't change the Matrox driver, still using 5.84.23.0 - But I did find out that the Matrox G450 (as well as other Gxx series) have a firmware BIOS, and I did upgrade the firmware from its original 1.6 to the latest greatest 2.x. I doubt that this was it that killed the problem, but I wanted it to be recorded, just in case... Thanks to the whole for your great work. I can't wait to put my hands on 1.1!
This task is fixed or worked in OOo 1.1 beta2.
closed ...