From: George Shammas <georgyo@gmail.com>
To: peff@peff.net
Cc: Junio C Hamano <gitster@pobox.com>, l.s.r@web.de, git@vger.kernel.org
Subject: Re: git merge -s subtree seems to be broken.
Date: Tue, 31 Jul 2018 15:52:26 -0400 [thread overview]
Message-ID: <CAF1Ko+FHsvmqzwVHh+fEnk=UGUftNW8VkFwaWTSKu3xYprb+wg@mail.gmail.com> (raw)
In-Reply-To: <20180731190459.GA3372@sigill.intra.peff.net>
[-- Attachment #1: Type: text/plain, Size: 7840 bytes --]
This is the fastest I ever seen an open source project respond to an issue
I reported. Thanks for being awesome!
On Tue, Jul 31, 2018 at 3:05 PM Jeff King <peff@peff.net> wrote:
> On Tue, Jul 31, 2018 at 01:23:04PM -0400, Jeff King wrote:
>
> > On Tue, Jul 31, 2018 at 10:17:15AM -0700, Junio C Hamano wrote:
> >
> > > Jeff King <peff@peff.net> writes:
> > >
> > > > +...
> > > > + } else if (cmp > 0) {
> > > > /* path2 does not appear in one */
> > > > + score += score_missing(two.entry.mode,
> two.entry.path);
> > > > + update_tree_entry(&two);
> > > > + continue;
> > > > + } if (oidcmp(one.entry.oid, two.entry.oid)) {
> > >
> > > As the earlier ones do the "continue at the end of the block", this
> > > does not affect the correctness, but I think you either meant "else if"
> > > or a fresh "if/else" that is disconnected from the previous if/else
> if/...
> > > chain.
> >
> > Yes, thanks. I actually started to write it without the "continue" at
> > all, and a big "else" that checked the "we have both" case. But I backed
> > that out (in favor of a smaller diff), and forgot to add back in the
> > "else if".
>
> So here it is fixed, and with a commit message. I'm not happy to omit a
> regression test, but I actually couldn't come up with a minimal one that
> tickled the problem, because we're playing around with heuristics. So I
> compensated by probably over-explaining in the commit message. But
> clearly this is not a well-tested code path given the length of time
> between introducing and detecting the bug.
>
> -- >8 --
> Subject: [PATCH] score_trees(): fix iteration over trees with missing
> entries
>
> In score_trees(), we walk over two sorted trees to find
> which entries are missing or have different content between
> the two. So if we have two trees with these entries:
>
> one two
> --- ---
> a a
> b c
> c d
>
> we'd expect the loop to:
>
> - compare "a" to "a"
>
> - compare "b" to "c"; because these are sorted lists, we
> know that the second tree does not have "b"
>
> - compare "c" to "c"
>
> - compare "d" to end-of-list; we know that the first tree
> does not have "d"
>
> And prior to d8febde370 (match-trees: simplify score_trees()
> using tree_entry(), 2013-03-24) that worked. But after that
> commit, we mistakenly increment the tree pointers for every
> loop iteration, even when we've processed the entry for only
> one side. As a result, we end up doing this:
>
> - compare "a" to "a"
>
> - compare "b" to "c"; we know that we do not have "b", but
> we still increment both tree pointers; at this point
> we're out of sync and all further comparisons are wrong
>
> - compare "c" to "d" and mistakenly claim that the second
> tree does not have "c"
>
> - exit the loop, mistakenly not realizing that the first
> tree does not have "d"
>
> So contrary to the claim in d8febde370, we really do need to
> manually use update_tree_entry(), because advancing the tree
> pointer depends on the entry comparison.
>
> That means we must stop using tree_entry() to access each
> entry, since it auto-advances the pointer. Instead:
>
> - we'll use tree_desc.size directly to know if there's
> anything left to look at (which is what tree_entry() was
> doing under the hood)
>
> - rather than do an extra struct assignment to "e1" and
> "e2", we can just access the "entry" field of tree_desc
> directly
>
> That makes us a little more intimate with the tree_desc
> code, but that's not uncommon for its callers.
>
> There's no regression test here, as it's a little tricky to
> trigger this with a minimal example. The user-visible effect
> is that the heuristics fail to correlate two trees that
> should be. But in a minimal example, there aren't a lot of
> other trees to match, so we often end up doing the right
> thing anyway.
>
> A real-world example (from the original bug report) is:
>
> -- >8 --
> git init repo
> cd repo
>
> echo init >file
> git add file
> git commit -m init
>
> git remote add tig https://github.com/jonas/tig.git
> git fetch tig
> git merge -s ours --no-commit --allow-unrelated-histories tig-2.3.0
> git read-tree --prefix=src/ -u tig-2.3.0
> git commit -m 'get upstream tig-2.3.0'
>
> echo update >file
> git commit -a -m update
>
> git merge -s subtree tig-2.4.0
> -- 8< --
>
> Before this patch, we fail to realize that the tig-2.4.0
> content should go into the "src" directory.
>
> Signed-off-by: Jeff King <peff@peff.net>
> ---
> match-trees.c | 43 ++++++++++++++++++++++++++-----------------
> 1 file changed, 26 insertions(+), 17 deletions(-)
>
> diff --git a/match-trees.c b/match-trees.c
> index 4cdeff53e1..37653308d3 100644
> --- a/match-trees.c
> +++ b/match-trees.c
> @@ -83,34 +83,43 @@ static int score_trees(const struct object_id *hash1,
> const struct object_id *ha
> int score = 0;
>
> for (;;) {
> - struct name_entry e1, e2;
> - int got_entry_from_one = tree_entry(&one, &e1);
> - int got_entry_from_two = tree_entry(&two, &e2);
> int cmp;
>
> - if (got_entry_from_one && got_entry_from_two)
> - cmp = base_name_entries_compare(&e1, &e2);
> - else if (got_entry_from_one)
> + if (one.size && two.size)
> + cmp = base_name_entries_compare(&one.entry,
> &two.entry);
> + else if (one.size)
> /* two lacks this entry */
> cmp = -1;
> - else if (got_entry_from_two)
> + else if (two.size)
> /* two has more entries */
> cmp = 1;
> else
> break;
>
> - if (cmp < 0)
> + if (cmp < 0) {
> /* path1 does not appear in two */
> - score += score_missing(e1.mode, e1.path);
> - else if (cmp > 0)
> + score += score_missing(one.entry.mode,
> one.entry.path);
> + update_tree_entry(&one);
> + } else if (cmp > 0) {
> /* path2 does not appear in one */
> - score += score_missing(e2.mode, e2.path);
> - else if (oidcmp(e1.oid, e2.oid))
> - /* they are different */
> - score += score_differs(e1.mode, e2.mode, e1.path);
> - else
> - /* same subtree or blob */
> - score += score_matches(e1.mode, e2.mode, e1.path);
> + score += score_missing(two.entry.mode,
> two.entry.path);
> + update_tree_entry(&two);
> + } else {
> + /* path appears in both */
> + if (oidcmp(one.entry.oid, two.entry.oid)) {
> + /* they are different */
> + score += score_differs(one.entry.mode,
> + two.entry.mode,
> + one.entry.path);
> + } else {
> + /* same subtree or blob */
> + score += score_matches(one.entry.mode,
> + two.entry.mode,
> + one.entry.path);
> + }
> + update_tree_entry(&one);
> + update_tree_entry(&two);
> + }
> }
> free(one_buf);
> free(two_buf);
> --
> 2.18.0.796.g4bfd63b683
>
>
[-- Attachment #2: Type: text/html, Size: 10318 bytes --]
next prev parent reply other threads:[~2018-07-31 19:52 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-07-31 14:09 git merge -s subtree seems to be broken George Shammas
2018-07-31 15:03 ` George Shammas
2018-07-31 15:50 ` Jeff King
2018-07-31 16:08 ` Junio C Hamano
2018-08-01 0:58 ` René Scharfe
2018-07-31 15:53 ` Junio C Hamano
2018-07-31 15:56 ` George Shammas
2018-07-31 16:15 ` Jeff King
2018-07-31 17:17 ` Junio C Hamano
2018-07-31 17:23 ` Jeff King
2018-07-31 19:04 ` Jeff King
2018-07-31 19:52 ` George Shammas [this message]
2018-07-31 20:40 ` Jeff King
2018-07-31 21:06 ` Junio C Hamano
2018-08-01 0:58 ` René Scharfe
2018-08-02 18:58 ` Jeff King
2018-08-02 18:45 ` Jeff King
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAF1Ko+FHsvmqzwVHh+fEnk=UGUftNW8VkFwaWTSKu3xYprb+wg@mail.gmail.com' \
--to=georgyo@gmail.com \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=l.s.r@web.de \
--cc=peff@peff.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).