Re: BUG: git log: fatal: internal error in diff-resolve-rename-copy

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Aug 13, 2010, 17:53 UTC
Message-ID: <AANLkTinSUy9pDtsJizmQTTOsMOVEQ=7+U5Ri4=-RZmB-@mail.gmail.com>
In-Reply-To: <7vhbiyl8ji.fsf@alter.siamese.dyndns.org>

On Fri, Aug 13, 2010 at 10:36 AM, Junio C Hamano <gitster@pobox.com> wrote:

Show 13 quoted lines

>
> Another potential "funny" is this (unrelated to the reported issue).
>
> The "--follow" logic is called from diff_tree_sha1() function, but the
> input trees to diff_tree_sha1() are not necessarily the top-level trees
> (compare_tree_entry() calls it while it recursively descends into
> subtrees).  For example, with the example Constantine gave us, the first
> "try-to-follow-renames" call happens with the "base" set to "platform/"
> but the rename source is actually "lang-api/src/com/intellij/..."
> hierarchy, so it is a wasted call.  I think we only want to run the rename
> following at the very top level, i.e. like the attached patch.
>
> Linus, what do you think?  Am I missing something obvious?

I think you're right. You're certainly not missing anythng _obvious_. Iirc the only reason I put that 'follow' hack in diff_tree_sha1() was because that way I didn't need to worry about all the callers, but I didn't even think about the whole recursion issue.

So ack on that.

                Linus