These are addressing some very weird cases. They should be an improvement, but I wonder if we can't do better.
> void xprintf (const char *fmt, ...) > { > va_list ap; >- size_t size; >+ size_t size = 0; > DWORD written; It would make more sense, IMO, to update the vstrfmtmake function to also set *lenp to 0, if lenp is set, when it runs out of memory. Winetest uses that function indirectly in other places, and one has a similar problem. (Of course, if we did run out of memory, it's a pretty dismal situation, and the returned string will be NULL, but at least xprintf will silently fail instead of crashing.) > extract_test (struct wine_test *test, const char *dir, LPTSTR res_name) > { > BYTE* code; >- DWORD size; >+ DWORD size = 0; For this to matter, the extract_rcdata call has to fail. This is, again, a pretty dismal situation. In this case, we'll later be writing data starting at NULL, and of an undefined, or with this change 0, length. I don't know if that actually works, and it may be better to give up early. Vincent Povirk