Cool VL Viewer forum

I'm getting frequent crashes out of nowhere, just the client disappearing. It does seem to occur in more populated areas than others.

All the logs attached

This seems related to these lines in the log:
Try disabling the "Multi-Threaded Curl" option in "Advanced" -> "Network" (once done, you wil need to restart the viewer for it to take effect). Let me know if it cures your crashing issue (if it does, I'll make this option disabled by default).

It's not crashing due to that error anymore, but it still does for some other reason. Sometimes just for nothing, other times while zooming or moving with the camera. Occasionally there will be a driver crash on the video.

CPU: Intel(R) Core(TM)2 Duo CPU T9300 @ 2.50GHz (2493.75 MHz)
Memory: 4031 MB
OS Version: Microsoft Windows 7 64-bit Service Pack 1 (Build 7601)
Graphics Card Vendor: NVIDIA Corporation
Graphics Card: Quadro FX 570M/PCI/SSE2
Windows Graphics Driver Version: 8.17.0012.7061
OpenGL Version: 3.3.0

libcurl Version: libcurl/7.20.1 OpenSSL/0.9.8j zlib/1.2.3
J2C Decoder Version: KDU
Audio Driver Version: FMOD version 3.750000
Qt Webkit Version: 4.7.1

These sound like video driver related crashes... Current Nvidia drivers are providing OpenGL v4, yours is at v3 only. I suggest you try and update your video card driver. The logs alas do not provide any useful info (too bad you are not running it under Linux: we would get the "stack_trace.log" file that pinpoints where exactly the crash occurred, including when in video drivers, or you could run the viewer under gdb...).

UPDATE: I just found, in the former session (Secondlife.old) log file the cause for a crash bug (unrelated to graphics however, so you still may have had graphic driver issues for the last session), caused by a corrupted sound (UUID: 37eff99e-555a-ba37-60e0-db859f38a9fd), and could reproduce that bug by standing here. It caused the viewer to try and decode that corrupted sound forever and also caused a severe memory leak that would end up in the viewer exhausting the free memory and therefore bailing out ("crashing") in the end... That bug (that could very well affect many other viewers) has now been fixed. The fix will be part of v1.26.0.19 and v1.26.1.6, to be released this week.

I know I'm not the OP, but reporting in that I tried this myself on 1.26.0 (18) and *knocks on wood* so far it seems to be working without hourly crashes.

So you keep saying, and far be it from me to doubt you. I have, however, observed numerous cases where stack_trace.log was a mere 4 bytes in size and essentially contained a single-digit number and nothing else --- last entry in the log file is (obviously) the "Opening stack trace file" message. Any idea what might be the cause for that? The system is otherwise rock-solid (Ubuntu 10.04 LTS, latest Nvidia drivers and plenty of RAM, i.e. no OOM kills or the like).

I would certainly provide more crash reports if I didn't know that not having the stack trace essentially makes things way harder for you.

An almost 100% certain way to torture the viewer (and myself, but that's another story ) into this situation is simply standing in Damsel at the sim border looking over towards the Bondage Ranch gate. No other interaction, just watching the virtual world pass by. Sometimes the viewer (and this has been with any version of Cool for a long time now) will last 15 minutes, sometimes 30. The longer I have been on before I go there, the shorter it tends to last.

I know, no proper report, but maybe it gives you some ideas.

Love,
Lia.

That's why I mentioned gdb... Sometimes, the memory gets so badly messed up that even the stack logger fails. But gdb never fails.

To run the viewer under gdb, proceed as follow:

• edit the "snowglobe" wrapper script,
• uncomment the "#export LL_WRAPPER='gdb --args'" (removing the hash at the start to read: "export LL_WRAPPER='gdb --args'").
• From a terminal (xterm, gnome-terminal, kterm, etc), change to the viewer installation directory and type ./snowglobe
• Once gdb got the viewer loaded, type "run" and then use the viewer as usual till it crashes (the display will "freeze" forever when it does).
• Go back to the terminal and type "bt" to get the stack trace.

I've updated to the latest latest driver (a beta just released) but it appears NVidia does not support OpenGL 4 on my chipset, which is a G84.

I spent 1H10 idling there (with over 50 avatars around): no crash.

v1.26.1.6 (and v1.26.0.19) have been released. Give it a try: it got some crash bugs fixed that could well solve your issues (idled again with it in Damsel over 1H30 yesterday, without a crash).