[Paraview-developers] strange bug proving hard to track down

Biddiscombe, John A. biddisco at cscs.ch
Thu Jun 27 01:54:39 EDT 2013


Dear Anyone

I have a custom representation which uses cuda from a plugin. When I setup paraview client-server with the remote render threshold set to zero, all works perfectly.
When I set the remote render data size threshold to a few MB so that a small model is transferred to the client for rendering, the client crashes and the servers report "wrong tag received, expected 10, got 1" or similar (from memory. Might have mixed the numbers).

If I first start the client server with remote threshold at zero, view the model, then change the threshold to 3MB destroy the window, create a new one, it works fine - I expected a crash here.

My representation plugin initializes cuda on first call to AddToView(....), when it needs to setup stuff.
If I manually start servers attach gdb, then create a model and render it, just when I expect it to crash, gdb produces the message "Detaching after fork from child process ...pid", and it looks as thought he pvserver tasks are spawning something. This is unexpected.

I'm still looking into this, but my question is ...
Can anyone explain why it might crash in this way when remote render is non zero (ie transfer to client is enabled), but not crash when remote render is used.
And why pvserver is forking something, which I did not expect.

Thanks  -any clues welcome as this is a most annoying bug I want to fix in my plugin.

JB




--
John Biddiscombe,                        email:biddisco @.at.@ cscs.ch
http://www.cscs.ch/
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Trevano 131, 6900 Lugano, Switzerland   | Fax:  +41 (91) 610.82.82

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview-developers/attachments/20130627/3e7a8e57/attachment.htm>


More information about the Paraview-developers mailing list