R devel - Nov 2008 - Diagnosing Crashes on garbage collection

I'm having problems with an access violation error on some compiled code
that I'm writing and I was hoping that some one had insights as to what
might be causing the access violation and possibly how to remedy it.

Here's the situation.  The compiled code performs an EM algorithm on the
data for given penalty parameters. The penalty parameters are estimated
through optim on the cross validated log likelihood.  The program works most
of the time. When the EM Algorithm fails to converge  the program crashes on
garbage collection.  Although this makes little sense, as there is no
special computations when the algorithm fails to converge.

The compiled code can either free it's used memory or let R's garbage
collection take care of it, it makes no differeence, the program still
crashes.

Any help on trying to either fix or even help on figuring it out, like tools
to use, would be a great help.

Thanks,
Andrew

	[[alternative HTML version deleted]]

On Sun, 16 Nov 2008, Andrew Redd wrote:
> I'm having problems with an access violation error on some compiled
code
> that I'm writing and I was hoping that some one had insights as to what
> might be causing the access violation and possibly how to remedy it.
If you are using (or can use) Linux, then valgrind is an excellent tool 
for this. There's some documentation in 'Writing R Extensions'.
Valgrind
runs your code in a virtual machine and tracks all memory accesses, so it 
often will even find bugs in C code that are hard to reproduce.

 	-thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle