On Saturday 15 March 2008 13:57:41 Peter Gibbs wrote: > Since the valgrind client requests are supposed to be very low overhead > when not running under valgrind, there should be no problem with adding > a configure step to define it (and set the correct library path, which I > just hardcoded), but I just used -DVALGRIND. That patch is not > committed, as I don't know if valgrind is used often enough to justify > it.
Given the cost of Parrot_assert() which we also use, it seems useful enough to enable when building with debugging symbols anyway. -- c
