Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Thu, 14 Dec 2006 13:39:07 -0700

> The strange thing though is that it's reporting an unknown C++
> exception, so there is probably still something there that's able to
> catch whatever it being thrown.

What you are seeing is a message from the last ditch exception handler in
the C++ epicsThread class thread entry point wrapper.

Since we first try, unsuccessfully, to catch std::exception and nearly all
exceptions thrown derive from std::exception I am guessing that this may not
be a true C++ exception, but instead some sort of worse problem like a bus
error that gets mapped into a C++ exception. Whether that occurs or not is
highly dependent on the implementation. The debugger might help with
determining exactly what type of exception is being thrown.

> If we could modify the CAC-UDP task to
> expect that exception we could at least suppress the warning message, or
> convert it to something a little more innocuous.

The problem is that this is coming from general purpose code, and we
actually like to see error messages when code fails so that bugs (even this
one) eventually get fixed. 

Jeff

> -----Original Message-----
> From: Andrew Johnson [mailto:[email protected]]
> Sent: Thursday, December 14, 2006 1:18 PM
> To: Jeff Hill
> Cc: 'EPICS core-talk'
> Subject: Re: C++ Exceptions from CAC-UDP on exit
> 
> Jeff Hill wrote:
> >
> > The root problem here I suspect is that DB CA link code does not have a
> > shutdown procedure and therefore it is not destroying its CA context
> before
> > the process exits. So when the process rundown starts and the resources
> that
> > the UDP thread uses start to cease to exist then the UDP thread crashes.
> 
> If that is the problem the I agree we don't have much chance of fixing
> it.  The strange thing though is that it's reporting an unknown C++
> exception, so there is probably still something there that's able to
> catch whatever it being thrown.  If we could modify the CAC-UDP task to
> expect that exception we could at least suppress the warning message, or
> convert it to something a little more innocuous.
> 
> > What OS do you see this on? We might be able to attach with a debugger
> to
> > get more detail, but I suspect that we will only see that the above
> > suspicions are accurate.
> 
> I've only seen it on linux-x86 so far - I tried it on Solaris but it
> didn't occur there, and I've not run the tests on vxWorks yet.
> 
> - Andrew
> --
> There is considerable overlap between the intelligence of the smartest
> bears and the dumbest tourists -- Yosemite National Park Ranger

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System