Looks OK to me. I know it's not likely, but what if a reloc really does happen at address zero? > * dlltool.c (flush_page): Do not munge padding entries in the > .reloc section.