On Sat, Nov 14, 2009 at 6:41 AM, Allan McRae <email@example.com> wrote:
> Why would we provide instructions on how to use valgrind? *That seems to *be
> a job for the valgrind authors...
I am always going back to Dan's blog to find how to run valgrind
> Anyway, I believe the pactest suite is set up to use valgrind so you would
> be better using that to detect leaks. *Also, *we have a valgrind.supp file
> to suppress a lot of the false positives.
Actually both are useful, sometimes we want to run valgrind on a
pactest to test a more specific codepath, sometimes we just want to
check a normal operation. Or also an operation which is not available,
Dan's article is still more complete, since it gives instructions for
both cases, and for valgrind/massif/callgrind.
But I think having some of that in HACKING is perfectly fine.