Help me to help

Hi all,

I subscribed three weeks ago, so I'm still an SL Hopeless Newbie (tm), and only because I found that LL provides a client for Linux.

Now, the alpha stage is evident, and I don't complain. It crashes so often that's not even funny, but it was my precise choice to live on the bleeding edge, and I don't mind about blood stains.

Said this, the question: how can we help Tofu and the other Linux devs (if there are some) improve the Linux client? I set both viewers (main grid and beta grid) so they send crash reports every time they need to, either at crash or at restart.

Is this too much? Too little? There are other ways to help the devs? Running the client under a debugger? Providing the kernel net params? I don't know what could help them, but I'm willing to provide them the info they will tell me they'd like to have.

Config as reported by inworld About, and commented where lacking:

CPU: Can't get terse CPU information
  {Athlon XP 2500+, mask CPUslow}
Memory: 2028 MB
  {DDR400}
OS Version: Linux 2.6.20-gentoo #1 PREEMPT Mon Feb 12 12:41:54 CET 2007 i686
  {gentoo ~x86, glibc 2.5.0, gtk+ 2.10.11, Xorg 7.2.0}
Graphics Card Vendor: NVIDIA Corporation
Graphics Card: GeForce 6600 GT/AGP/SSE/3DNOW!
  {128MB, mask Class2}
  {LL_GL_BASICEXT commented out}
  {exported an empty LL_GL_BLACKLIST}
OpenGL Version: 2.1.0 NVIDIA 97.55
LLMozLib Version: 1.1.0 (Mozilla GRE:1.8.0.10_0000000000)

Also, I have monumental packet loss, which at times makes moving around an exercise in applied perseverance. Is there something I can do about that, considering that I access the grid(s) with a 4096/256 kbit/sec DSL, and I have a 180-290ms rtt latency? What's a good address to tracepath, to provide a cogent latency map from me to the grid?

Try messing around with the bandwidth settings on the client. Usually it will try to adjust, but packet loss will happen if it is set too high.

Thanks Kerik,

Maybe it's coincidence, but I throttled down the max connection speed from 1000 to 770 and I had a decisive increase in texture loading performance and - gasp! - in stability. Which, honestly, baffled me to no end. I played for around three hours without crashing. I also managed to QUIT the client on my own free will (as opposed as to its own ).

Very wildly guessing, it seems like the client expects the data at a certain rate, and if the actual rate doesn't meet its expectations, it likes to crash with unreferenced pointers, corrupted double lists, and other amenities like these.

The equally unsettling observation is that the client keeps throttling the net speed (relaxing or tightening) on its own during the execution - around the same values I've always seen in the logs.

My understanding is that the bandwidth setting is to tell the host how fast to pump the data at you and if it is set too high, the data is pumped out faster than your client (or interveining routers) can pick it up. So slower will be faster as less packets are lost.

I have to keep my bandwidth settings quite low as my slow old 1-core CPU can't process incoming packets and SL rendering at the same time very well. I am figuring a dual-core cpu might help here even with SL being single-threaded as SL can do its stuff on one core while the OS does its stuff on the other ?????

Excellent observation. I've throttled my own settings to 700 (back from full blast @ 1500) and have had an uncanny increase in stability and loading performance. You may have stumbled on to something here!

back to the original question i would like to know too, i dont have hardly any problems

it next to never crashes
its more than 2x faster than the windows client
ect ect ect

id like to submit as much data as i can to LL to help them figure out why it works so well for me and not so well for others