On Thu, Jul 05, 2007 at 06:30:44PM -0700, chromatic wrote: > In theory, this patch should apply and run cleanly. It doesn't. > > Thus, something somewhere pokes into memory it shouldn't. > > Any ideas? Alternately, any comments on this analysis?
I also get segfaults after applying this patch. However, if I change the patch such that the "size_t sentinel;" line goes at the end of the struct PMC instead of the beginning, then everything appears to compile and run. In fact, if "size_t sentinel;" is placed as the second or third line in the struct (i.e., after "pobj_t obj;" or "VTABLE *vtable;") then things still appear to compile and run. It's only when the sentinel is given as the first line that I get errors. So, I'm guessing there's some code somewhere that assumes that "obj" is always the first component of PMC. Pm