Manta Interactive Ray Tracer Development Mailing List

Text archives Help


Re: [MANTA] Exiting manta...


Chronological Thread 
  • From: James Bigler <bigler@cs.utah.edu>
  • Cc: "'manta@sci.utah.edu'" <manta@sci.utah.edu>
  • Subject: Re: [MANTA] Exiting manta...
  • Date: Mon, 11 Jul 2005 16:30:51 -0600

The action taken by the signals are being registered in the install_signal_handlers() function. Looking at the man page for sigaction, we can cache the old function for later. You can add a new function to the thread library that reinstates the old functions. You could then call this from main.

James

James Bigler wrote:
I don't believe you can. The place where that variable is queried is called in Thread::initialize(). This is the first thing called by any of the constructors in the Thread library. If we have a mutex or other thing as a static or global variable it will be called before main().

James

Abe Stephens wrote:


Hi,

After taking out a prism extreme five times in one hour, I have discovered a problem when manta is run with certain versions of vizserver. The signal catching code in Thread_pthreads.cc appears to cause a crash to kdb when the renderer exits. Setting THREAD_NO_CATCH_SIGNALS eliminates the problem, however it doesn't seem that I can set this environment variable from inside my main function.
Should I just be able to place: "setenv("THREAD_NO_CATCH_SIGNALS","1",false);" in main?

I have three questions:

1.) Is the SCIRun code doing anything out of the ordinary when it catches these signals.
2.) Which platforms actually require this functionality for correct operation. It appears that it is not necessary on altix.
3.) If we elimitate this code (might be easier then fixing the linux kernel or vizserver) from the IA-64 manta build will it affect any other parts of the renderer?

Thanks--
Abe







Archive powered by MHonArc 2.6.16.

Top of page