[tree-ssa mudflap] Add more verbosity

[Eyal Lebedinsky]

When debugging a wrapper I often hit a problem when I register
an objects incorrectly. This almost always leads to a failure at
unregister time in mf-runtime.c in
		assert (old_obj->data.type >= 0 &&
			old_obj->data.type <= __MF_TYPE_MAX_CEM);

I find it usefull to add a violation message at this point (just
before the assert, using the same condition) like:
	if (!(old_obj->data.type >= 0 &&
		old_obj->data.type <= __MF_TYPE_MAX_CEM)) {
		fprintf (stderr,
			"*******\n"
			"mudflap assert: bad object type\n");
		__mf_describe_object (&old_obj->data);
	}

Maybe this should depend of some runtime option? Still, it is
highly effective for me. Will this sort of patch be acceptable?

What is the current procedure for me submitting a patch -
gcc-patches list? gcc list? or should I directly email someone?

--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>