Re: [brlcad-devel] Coverty and comgeom-g

[Christopher S. M. <br...@ma...>]

On Apr 25, 2012, at 6:50 AM, Tom Browder <tom...@gm...> wrote:

> I wonder why coverity is not finding all the failures to release
> memory in src/conv/comgeom/solid.c?  Does it not check for that?

Coverity has a variety of memory allocation checks, but there's only a few instances where you can be sure it's not a false positive.  Wiping out the only pointer to dynamically allocated memory, c++ classes that allocate but never free, even allocating suspicious sizes within malloc, and more.  It just doesn't/can't know when you're done with something that has a pointer referencing it. 

There are some really good memory checker tools (like valgrind), but most analyze runtime behavior instead of static analysis.

Cheers!