Text editor blows up the 8774 x86_64 driver. I didn't believe it at first either.

[actinic]

The system:

x86_64 Fedora Core 5 (kernel 2.6.17-1.2174_FC5)
AMD64 Athlon X2 4200+
Nvidia GeForce 6600 GT PCI-E

Steps to reproduce:

1. Install the Scite text editor (yum install scite)
2. Create one modestly large text file. The file I'm using is 22megs big. Each line in the file is 3000 to 4000 characters long.
3. Open the file in Scite.
4. Doom and disaster. If I start X from runlevel 3, the keyboard locks up and I have to log in over the network. If I run in runlevel 5, gdm restarts and reappears on the next virtual terminal. i.e. if X was virtual terminal 7, it's now on 8. If it was 8, it's now on 9, etc.

Observations:

1. Opening the same text file in other GUI text editors (gedit, kate) does not produce this behaviour
2. Opening the same text file in scite while using the nv driver does not produce this behaviour. I'm using the nv driver for the time being.
3. All manner of other applications whether they be 2D or 3D work with no issue. Google Earth, web browsers, email clients, etc. I have not seen any other sign system instability.
4. Smaller files open without issue in scite. However, I have some 1meg files which also generate the problem.
5. I see the same behaviour in the 8762 release of the drivers.

Whatever the scintilla text editing component of scite is doing, it's causing the nvidia driver to break. Can anyone else reproduce this behaviour? Does it also occur on 32bit systems?

[zander]

Please generate and attach an nvidia-bug-report.log file; please also see http://www.nvnews.net/vbulletin/showthread.php?t=46678 and http://www.nvnews.net/vbulletin/showthread.php?t=58498. Does the problem reproduce if you disable RENDER acceleration with Option "RenderAccel" "0"?

[actinic]

See attached log. With RenderAccel 0 the problem does not occur.

[AaronP]

Hi actinic,
This sounds like a known bug which will be resolved in the first 9xxx driver release. Was the attached bug report captured after X had crashed, before it was started again? The backtrace that X writes to its log when it crashes would be illuminating.

[actinic]

It was captured after X has crashed but before I rebooted the system. i.e. the system was still in the crashed state. Should I crash it, reboot to to run level 3, then generate the log?

[AaronP]

No, that should have been fine. I don't know why X didn't capture a backtrace.

[actinic]

I've tried the same steps using Scite on another system:

Fedora Core 5 i386
AMD Athlon XP 2800+
AGP GeForce FX5200 (driver version 8762)

And I get the same behaviour. I generated a bug report log on that system and it does not include any backtrace either. At least the problem seems easily reproducable on a range of systems.

[mrn]

this is the same problem as mine here:
http://www.nvnews.net/vbulletin/showthread.php?p=980668

And I can reproduce this with gedit on big files, turning off renderaccel solves the problem.

[netllama]

This bug was fixed in 1.0-9625 (last month):
http://www.nzone.com/object/nzone_do...etadriver.html