Jeff King <peff@xxxxxxxx> writes: > ... OTOH it is not really > solving the more fundamental problem, which is that p->parent[i].path is > only sometimes useful (we do not fill it in if it would just be the same > as p->path, so the patch only changes it from uninitialized memory into > an empty strbuf). > > And that is probably not something we want to change, as allocating > duplicates of each path may be expensive. Nicely said. I reached the same conclusion after looking at the existing code, even though I have to admit that I am not a huge fan of the more recent part of combine-diff.c and its data structures. Thanks.
