8774 causes Unidata's IDV to crash

[dmetz99]

Updating from 8762 to 8774 causes Unidata's IDV to crash as soon as the map display comes up. (Crash dump from IDV attached.) 8762 works fine. The problem is repeatable: 8762 = OK 8774 = crash.

If you're not familiar with IDV, here's the link:

http://www.unidata.ucar.edu/software/idv/

This is a large java app used to display and manipulate goesciences datasets. This is the first breakage I've encountered in quite a while due to a driver update.

(edited for spelling)

[netllama]

Please generate and post an nvidia-bug-report.log.

Also, does this require a dataset to reproduce, or just running IDV?

Thanks,
Lonni

[dmetz99]

No dataset is required to produced the crash, just starting up IDV. I just reverted back to 8762 so I could post-process the 18z NAM run (in progress). I'll do the bug report after that.

[dmetz99]

OK - bug report is attached. IDV starts, splash screen comes up, then "dashboard" comes up for a split second, then IDV crashes. No data set need be loaded for crash to occur.

[netllama]

I just installed IDV, however running "runIDV" fails immediately with the error:
Failed to load Main-Class mainifest attribute from idv.jar

I'm not clear how to debug that problem, however it seems unrelated to the nvidia X driver.

-Lonni

[dmetz99]

H'mm - I've not run into that problem. That sounds like type of messages you get from using a "wrong" jre. IDV typically installs its own jre. I've always installed IDV in a folder in my home login and it's worked fine.

[netllama]

Thanks. It looks like when I run the symlink that the installed createad into /usr/local/bin, I get the error I noted, however if I run runIDV directory from where it was installed, it starts up fine.

Unfortunately, for you, I'm not seeing a crash with 1.0-8774. The app starts up fine, and I see the green map.

Have you contacted IDV about this problem for their input on the crash?

-Lonni

[dmetz99]

Well poo...

Anyway, I've submitted a bug report to unidata about this problem. It's still reproducable here. Thanks for trying to reproduce it on your end. I'll also poke around in their mailing list archives and see what I can come up with.

[netllama]

I'd be curious whether Unidata can reproduce this. If so, please ask them to provide a bug report from a system where it reproduces, and I'll be happy to take a look. Perhaps seeing bug reports from multiple systems might shed some light on commonalities between the systems.

Thanks,
Lonni

[dmetz99]

I heard back from Unidata - they suggested trying different verision of Java3D, with the though that the problem might be glibc version incompatabilities. (I've also asked them if they can reproduce this problem.) The long & short of it that, here, all versions of Java3d from 1.3.0 to the latest 1.5.0 build 4 work fine with 8762 but crash exactly the same way with 8774. The crash messages appear to be pointing the finger at libGL.so.1 (which is linked to libGL.so.1.0.8762). Any further suggestion??

[netllama]

Was Unidata able to reproduce the crash? If so, I'd like to see a bug report from their system. I'm afraid that I don't have any other suggestions without additional datapoints.

Thanks,
Lonni

[dmetz99]

Unidata hasn't said whether they've been able to reproduce the crash, yet. I'll let you know if they do.