Rubén Justo <rjusto@xxxxxxxxx> writes: >> So the leak existed before that commit. Indeed it looks >> like the leak predates the introduction of struct image in b94f2eda99 >> (builtin-apply.c: make it more line oriented, 2008-01-26) and when >> the patch does not apply we have been leaking the buffer passed to >> apply_fragments() since the beginning of the builtin apply added in >> ac6245e31a3 (Builtin git-apply., 2006-05-23) > > You are very right. > > I saw that commit. I reviewed again my notes and I recorded the hash of > the previous hop; the one that my message, incorrectly, refers to. > I'll reroll to fix it. FWIW, the very first version of apply_one_fragment() in 3cca928d (git-apply: first cut at actually checking fragment data, 2005-06-05) already had leak; it "returns -1" upon an error without freeing old/new buffers. Over the years, the shape of the code changed, but the leak survived.
