# [PATCH] fetch: don't try to update unfetched tracking refs 6 messages from 2013-05-27 to 2013-05-27. Participants: John Keeping, Jeff King. Thread: https://gitlist.dev/t/33948 ## John Keeping, 2013-05-27 11:40 Subject: [PATCH] fetch: don't try to update unfetched tracking refs Message-ID: <9c6fd36368cd6b0fba4855be4d31106781c02f58.1369654825.git.john@keeping.me.uk> URL: https://gitlist.dev/e/9c6fd36368cd6b0fba4855be4d31106781c02f58.1369654825.git.john%40keeping.me.uk ``` Since commit f269048 (fetch: opportunistically update tracking refs, 2013-05-11) we update tracking refs opportunistically when fetching remote branches. However, if a refspec is given on the command line that does not include a configured (non-pattern) refspec a fatal error occurs. Fix this by setting the "missing_ok" flag when calling get_fetch_map. Signed-off-by: John Keeping --- builtin/fetch.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/builtin/fetch.c b/builtin/fetch.c index e41cc0d..d15a734 100644 --- a/builtin/fetch.c +++ b/builtin/fetch.c @@ -183,7 +183,7 @@ static struct ref *get_ref_map(struct transport *transport, old_tail = tail; for (i = 0; i < transport->remote->fetch_refspec_nr; i++) get_fetch_map(ref_map, &transport->remote->fetch[i], - &tail, 0); + &tail, 1); for (rm = *old_tail; rm; rm = rm->next) rm->fetch_head_status = FETCH_HEAD_IGNORE; } else { -- 1.8.3.rc3.438.gb3e4ae3 ``` ## Jeff King, 2013-05-27 15:42 Subject: Re: [PATCH] fetch: don't try to update unfetched tracking refs Message-ID: <20130527154252.GB8250@sigill.intra.peff.net> URL: https://gitlist.dev/e/20130527154252.GB8250%40sigill.intra.peff.net In-Reply-To: <9c6fd36368cd6b0fba4855be4d31106781c02f58.1369654825.git.john@keeping.me.uk> ``` On Mon, May 27, 2013 at 12:40:25PM +0100, John Keeping wrote: > Since commit f269048 (fetch: opportunistically update tracking refs, > 2013-05-11) we update tracking refs opportunistically when fetching > remote branches. However, if a refspec is given on the command line > that does not include a configured (non-pattern) refspec a fatal error > occurs. I'm not sure I understand what the last sentence means. I tried to add a test like: diff --git a/t/t5510-fetch.sh b/t/t5510-fetch.sh index ff43e08..02e30e1 100755 --- a/t/t5510-fetch.sh +++ b/t/t5510-fetch.sh @@ -422,6 +422,16 @@ test_expect_success 'configured fetch updates tracking' ' ) ' +test_expect_success 'non-configured ref does not confuse tracking update' ' + cd "$D" && + git update-ref refs/odd/location HEAD && + ( + cd three && + git fetch origin refs/odd/location && + git rev-parse --verify FETCH_HEAD + ) +' + test_expect_success 'pushing nonexistent branch by mistake should not segv' ' cd "$D" && but it does not fail with the existing code. Can you give an example that fails? -Peff ``` ## John Keeping, 2013-05-27 16:01 Subject: Re: [PATCH] fetch: don't try to update unfetched tracking refs Message-ID: <20130527160129.GN27005@serenity.lan> URL: https://gitlist.dev/e/20130527160129.GN27005%40serenity.lan In-Reply-To: <20130527154252.GB8250@sigill.intra.peff.net> ``` On Mon, May 27, 2013 at 11:42:52AM -0400, Jeff King wrote: > On Mon, May 27, 2013 at 12:40:25PM +0100, John Keeping wrote: > > > Since commit f269048 (fetch: opportunistically update tracking refs, > > 2013-05-11) we update tracking refs opportunistically when fetching > > remote branches. However, if a refspec is given on the command line > > that does not include a configured (non-pattern) refspec a fatal error > > occurs. > > I'm not sure I understand what the last sentence means. I tried to add a > test like: > > diff --git a/t/t5510-fetch.sh b/t/t5510-fetch.sh > index ff43e08..02e30e1 100755 > --- a/t/t5510-fetch.sh > +++ b/t/t5510-fetch.sh > @@ -422,6 +422,16 @@ test_expect_success 'configured fetch updates tracking' ' > ) > ' > > +test_expect_success 'non-configured ref does not confuse tracking update' ' > + cd "$D" && > + git update-ref refs/odd/location HEAD && > + ( > + cd three && > + git fetch origin refs/odd/location && > + git rev-parse --verify FETCH_HEAD > + ) > +' > + > test_expect_success 'pushing nonexistent branch by mistake should not segv' ' > > cd "$D" && > > but it does not fail with the existing code. Can you give an example > that fails? I have this in my .git/config for git.git: [remote "origin"] url = git://github.com/gitster/git fetch = +refs/heads/*:refs/remotes/origin/* fetch = +refs/notes/amlog:refs/notes/amlog Then doing "git fetch origin master" fails because: fatal: Couldn't find remote ref refs/notes/amlog The following test fails for me (and passes with my patch) - note that in "two", remote.one.fetch is configured as "refs/heads/master:refs/heads/one". -- >8 -- diff --git a/t/t5510-fetch.sh b/t/t5510-fetch.sh index ff43e08..c540257 100755 --- a/t/t5510-fetch.sh +++ b/t/t5510-fetch.sh @@ -422,6 +422,19 @@ test_expect_success 'configured fetch updates tracking' ' ) ' +test_expect_success 'configured ref does not confuse tracking' ' + + cd "$D" && + ( + cd one && + git branch -f side + ) && + ( + cd two && + git fetch one side + ) +' + test_expect_success 'pushing nonexistent branch by mistake should not segv' ' cd "$D" && ``` ## Jeff King, 2013-05-27 16:19 Subject: Re: [PATCH] fetch: don't try to update unfetched tracking refs Message-ID: <20130527161934.GB9235@sigill.intra.peff.net> URL: https://gitlist.dev/e/20130527161934.GB9235%40sigill.intra.peff.net In-Reply-To: <20130527160129.GN27005@serenity.lan> ``` On Mon, May 27, 2013 at 05:01:29PM +0100, John Keeping wrote: > > I'm not sure I understand what the last sentence means. I tried to add a > > test like: > [...] > > but it does not fail with the existing code. Can you give an example > > that fails? > > I have this in my .git/config for git.git: > > [remote "origin"] > url = git://github.com/gitster/git > fetch = +refs/heads/*:refs/remotes/origin/* > fetch = +refs/notes/amlog:refs/notes/amlog Ah, I see. It is not "the refspec on the command-line does not match a configured refspec", but rather "there exists a configured non-pattern refspec that does not match what was on the command-line (even if what was on the command-line did match another refspec)". So your fix makes perfect sense. Do you mind squashing in this test below? I think it is a little less subtle than what you posted, as it sets up the situation explicitly in the test. It also checks that the refs we _did_ match still get updated ("master" in this case). diff --git a/t/t5510-fetch.sh b/t/t5510-fetch.sh index ff43e08..fde6891 100755 --- a/t/t5510-fetch.sh +++ b/t/t5510-fetch.sh @@ -422,6 +422,22 @@ test_expect_success 'configured fetch updates tracking' ' ) ' +test_expect_success 'non-matching refspecs do not confuse tracking update' ' + cd "$D" && + git update-ref refs/odd/location HEAD && + ( + cd three && + git update-ref refs/remotes/origin/master base-origin-master && + git config --add remote.origin.fetch \ + refs/odd/location:refs/remotes/origin/odd && + o=$(git rev-parse --verify refs/remotes/origin/master) && + git fetch origin master && + n=$(git rev-parse --verify refs/remotes/origin/master) && + test "$o" != "$n" && + test_must_fail git rev-parse --verify refs/remotes/origin/odd + ) +' + test_expect_success 'pushing nonexistent branch by mistake should not segv' ' cd "$D" && Thanks for the fix. Acked-by: Jeff King -Peff ``` ## John Keeping, 2013-05-27 16:33 Subject: [PATCH v2] fetch: don't try to update unfetched tracking refs Message-ID: <20130527163308.GO27005@serenity.lan> URL: https://gitlist.dev/e/20130527163308.GO27005%40serenity.lan In-Reply-To: <20130527161934.GB9235@sigill.intra.peff.net> ``` Since commit f269048 (fetch: opportunistically update tracking refs, 2013-05-11) we update tracking refs opportunistically when fetching remote branches. However, if there is a configured non-pattern refspec that does not match any of the refspecs given on the command line then a fatal error occurs. Fix this by setting the "missing_ok" flag when calling get_fetch_map. Test-added-by: Jeff King Signed-off-by: John Keeping Acked-by: Jeff King --- On Mon, May 27, 2013 at 12:19:34PM -0400, Jeff King wrote: > On Mon, May 27, 2013 at 05:01:29PM +0100, John Keeping wrote: > > > > I'm not sure I understand what the last sentence means. I tried to add a > > > test like: > > [...] > > > but it does not fail with the existing code. Can you give an example > > > that fails? > > > > I have this in my .git/config for git.git: > > > > [remote "origin"] > > url = git://github.com/gitster/git > > fetch = +refs/heads/*:refs/remotes/origin/* > > fetch = +refs/notes/amlog:refs/notes/amlog > > Ah, I see. It is not "the refspec on the command-line does not match a > configured refspec", but rather "there exists a configured non-pattern > refspec that does not match what was on the command-line (even if what > was on the command-line did match another refspec)". Exactly. I've changed the commit message to (hopefully) make this clearer. > So your fix makes perfect sense. Do you mind squashing in this test > below? I think it is a little less subtle than what you posted, as it > sets up the situation explicitly in the test. It also checks that the > refs we _did_ match still get updated ("master" in this case). Done. builtin/fetch.c | 2 +- t/t5510-fetch.sh | 16 ++++++++++++++++ 2 files changed, 17 insertions(+), 1 deletion(-) diff --git a/builtin/fetch.c b/builtin/fetch.c index e41cc0d..d15a734 100644 --- a/builtin/fetch.c +++ b/builtin/fetch.c @@ -183,7 +183,7 @@ static struct ref *get_ref_map(struct transport *transport, old_tail = tail; for (i = 0; i < transport->remote->fetch_refspec_nr; i++) get_fetch_map(ref_map, &transport->remote->fetch[i], - &tail, 0); + &tail, 1); for (rm = *old_tail; rm; rm = rm->next) rm->fetch_head_status = FETCH_HEAD_IGNORE; } else { diff --git a/t/t5510-fetch.sh b/t/t5510-fetch.sh index ff43e08..fde6891 100755 --- a/t/t5510-fetch.sh +++ b/t/t5510-fetch.sh @@ -422,6 +422,22 @@ test_expect_success 'configured fetch updates tracking' ' ) ' +test_expect_success 'non-matching refspecs do not confuse tracking update' ' + cd "$D" && + git update-ref refs/odd/location HEAD && + ( + cd three && + git update-ref refs/remotes/origin/master base-origin-master && + git config --add remote.origin.fetch \ + refs/odd/location:refs/remotes/origin/odd && + o=$(git rev-parse --verify refs/remotes/origin/master) && + git fetch origin master && + n=$(git rev-parse --verify refs/remotes/origin/master) && + test "$o" != "$n" && + test_must_fail git rev-parse --verify refs/remotes/origin/odd + ) +' + test_expect_success 'pushing nonexistent branch by mistake should not segv' ' cd "$D" && -- 1.8.3.rc3.438.gb3e4ae3 ``` ## Jeff King, 2013-05-27 16:35 Subject: Re: [PATCH v2] fetch: don't try to update unfetched tracking refs Message-ID: <20130527163500.GA18405@sigill.intra.peff.net> URL: https://gitlist.dev/e/20130527163500.GA18405%40sigill.intra.peff.net In-Reply-To: <20130527163308.GO27005@serenity.lan> ``` On Mon, May 27, 2013 at 05:33:09PM +0100, John Keeping wrote: > > Ah, I see. It is not "the refspec on the command-line does not match a > > configured refspec", but rather "there exists a configured non-pattern > > refspec that does not match what was on the command-line (even if what > > was on the command-line did match another refspec)". > > Exactly. I've changed the commit message to (hopefully) make this > clearer. It does. To me, anyway. :) Thanks again. -Peff ```