On Thu, Jun 16, 2016 at 4:47 PM, Charles Bailey <charles@xxxxxxxxxxxxx> wrote: > On Thu, Jun 16, 2016 at 02:47:09PM +0700, Duy Nguyen wrote: >> I don't think revert is right. It rather needs a re-fix like below. >> Basically we want grep_file() to run as normal, but grep_sha1() >> (i.e. git grep --cached) should ignore i-t-a entries, because empty >> SHA-1 is not the right content to grep. It does not matter in positive >> matching, sure, but it may in -v cache. > > You don't think the revert is correct or you don't think the revert is > sufficient? (I wasn't able to find a test case which proved that the > change to line 399 was necessary, so perhaps I don't understand.) OK insufficient. > I would have thought that grepping the empty SHA-1 would be correct for > with or without -v. An "intent to add" file has no content in the index > so I would expect it to have zero matching and zero non-matching lines > for any grep --cached query? > > Or is this an efficiency and not a correctness concern? "git grep --cached" searches file content that will be committed by "git commit" (no -a). An i-t-a entry will not be committed (you would need "git add" first, or do "git commit -a"). So if I say "search among the to-be-committed file content, list files that do not match abc" (git grep -l -v --cached abc), the i-t-a entry will show up because its fake content is empty (i.e. not contain "abc"), even though it's not in the "to-be-committed" list. So yeah, correctness issue. -- Duy -- 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