On Sun, Feb 24, 2008 at 07:08:52PM -0800, Junio C Hamano wrote: > Martin Koegler <mkoegler@xxxxxxxxxxxxxxxxx> writes: > As far as I can tell, the new test is not testing the commit > object we are looking at from the object database. What it is > testing is if the code that parsed and prepared the information > in "struct commit" found the same number of parents an extra > check we are doing here by hand (if not grafted --- but > presumably whoever gave the struct commit we are handling here > would have obtained that information by doing the same parsing), > or the parsing of the graft file (when grafted --- but > presumably whoever gave the struct commit we are handling here > would have obtained that information by calling the same > llokup_commit_graft()). > > So I am not sure what problems in the repository objects these > new checks are designed to catch. > > This needs a lot of explanation than what's in your commit log > message. If we have already parsed a non-commit object and parse_commit_buffer hits such a sha1 in a parent line, it simply drops it (same for grafts). I hope that you agree with me, that fsck should catch such an error. As you don't want to tighten parse_commit_buffer, it must be in fsck. mfg Martin Kögler - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
