git@vger.kernel.org list mirror (unofficial, one of many)
 help / color / mirror / Atom feed
* HEAD and namespaces
@ 2019-05-21  2:59 Bryan Turner
  2019-05-21 21:46 ` Jeff King
  0 siblings, 1 reply; 6+ messages in thread
From: Bryan Turner @ 2019-05-21  2:59 UTC (permalink / raw)
  To: Git Users

When using GIT_NAMESPACE, it appears the "symref" added to the
capabilities advertisement doesn't get the namespace stripped. The
namespace is stripped for the advertised refs, including "HEAD", but
not on the "symref".

Lafiel:test bturner$ GIT_NAMESPACE=foo GIT_TRACE_PACKET=1 git ls-remote origin
19:51:54.012696 pkt-line.c:80           packet:          git<
b9acca03606d4c674be8b7e79cd788cedbec957c HEAD\0multi_ack thin-pack
side-band side-band-64k ofs-delta shallow deepen-since deepen-not
deepen-relative no-progress include-tag multi_ack_detailed
symref=refs/namespaces/foo/HEAD:refs/namespaces/foo/refs/heads/master
agent=git/2.20.1
19:51:54.013017 pkt-line.c:80           packet:          git<
b9acca03606d4c674be8b7e79cd788cedbec957c refs/heads/master
19:51:54.013028 pkt-line.c:80           packet:          git< 0000
19:51:54.013035 pkt-line.c:80           packet:          git> 0000
b9acca03606d4c674be8b7e79cd788cedbec957c HEAD
b9acca03606d4c674be8b7e79cd788cedbec957c refs/heads/master

I don't know whether the client somehow does some stripping on the
"symref" to decide that HEAD should be
"refs/namespaces/foo/refs/heads/master", but I'd assume not. Either
way, since the clone can't find the HEAD value (either because it
doesn't see "refs/namespaces/foo/HEAD" as HEAD, which I'd assume is
the case, or because it can't find
"refs/namespaces/foo/refs/heads/master"), it falls back on the old
behavior of picking the first branch with a matching commit. That
brings back all the "wrong branch" issues with cloning that adding
"symref"  fixed.

Is this a bug? An oversight? An intentional decision? How is HEAD
supposed to work when using GIT_NAMESPACE? Perhaps the expectation is
that namespaces won't have their own HEADs? I'd say perhaps it's that
even in the namespace the target ref shouldn't be namespaced, but that
doesn't seem like it could possibly be correct since the namespace
could contain refs that don't exist outside it, so Git would see the
symbolic ref as broken.

Any insight would be greatly appreciated!

Bryan

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: HEAD and namespaces
  2019-05-21  2:59 HEAD and namespaces Bryan Turner
@ 2019-05-21 21:46 ` Jeff King
  2019-05-22  4:31   ` [PATCH] upload-pack: strip namespace from symref data Jeff King
  0 siblings, 1 reply; 6+ messages in thread
From: Jeff King @ 2019-05-21 21:46 UTC (permalink / raw)
  To: Bryan Turner; +Cc: Git Users

On Mon, May 20, 2019 at 07:59:03PM -0700, Bryan Turner wrote:

> When using GIT_NAMESPACE, it appears the "symref" added to the
> capabilities advertisement doesn't get the namespace stripped. The
> namespace is stripped for the advertised refs, including "HEAD", but
> not on the "symref".
> 
> Lafiel:test bturner$ GIT_NAMESPACE=foo GIT_TRACE_PACKET=1 git ls-remote origin
> 19:51:54.012696 pkt-line.c:80           packet:          git<
> b9acca03606d4c674be8b7e79cd788cedbec957c HEAD\0multi_ack thin-pack
> side-band side-band-64k ofs-delta shallow deepen-since deepen-not
> deepen-relative no-progress include-tag multi_ack_detailed
> symref=refs/namespaces/foo/HEAD:refs/namespaces/foo/refs/heads/master

Yeah, that seems totally broken. It's neat that we find HEAD inside the
namespace. But I think there are two bugs here:

  1. We must call it "HEAD" and not the nonsense
     refs/namespaces/foo/HEAD

  2. We need to similarly remove the namespace from its contents. I
     suppose it's possible that somebody stores the already-truncated
     name in the symref, but as you noted then such a symref would be
     totally broken from the perspective of the non-namespaced repo.

> I don't know whether the client somehow does some stripping on the
> "symref" to decide that HEAD should be
> "refs/namespaces/foo/refs/heads/master", but I'd assume not.

I don't think it can. It has no idea that namespaces are in play at all
(and the point of the feature is for it to be invisible to the client).

> Is this a bug? An oversight? An intentional decision? How is HEAD
> supposed to work when using GIT_NAMESPACE? Perhaps the expectation is
> that namespaces won't have their own HEADs? I'd say perhaps it's that
> even in the namespace the target ref shouldn't be namespaced, but that
> doesn't seem like it could possibly be correct since the namespace
> could contain refs that don't exist outside it, so Git would see the
> symbolic ref as broken.

I think it should be doing:

diff --git a/upload-pack.c b/upload-pack.c
index 24298913c0..4d2129e7fc 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1037,8 +1037,8 @@ static int find_symref(const char *refname, const struct object_id *oid,
 	symref_target = resolve_ref_unsafe(refname, 0, NULL, &flag);
 	if (!symref_target || (flag & REF_ISSYMREF) == 0)
 		die("'%s' is a symref but it is not?", refname);
-	item = string_list_append(cb_data, refname);
-	item->util = xstrdup(symref_target);
+	item = string_list_append(cb_data, strip_namespace(refname));
+	item->util = xstrdup(strip_namespace(symref_target));
 	return 0;
 }
 

and this case has just been broken since 7171d8c15f (upload-pack: send
symbolic ref information as capability, 2013-09-17). It mostly works
because of fallback to the old guessing behavior, so nobody really
noticed. (I also suspect namespaces are not widely used at all, but I
could be wrong).

-Peff

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH] upload-pack: strip namespace from symref data
  2019-05-21 21:46 ` Jeff King
@ 2019-05-22  4:31   ` Jeff King
  2019-05-22 10:33     ` Ævar Arnfjörð Bjarmason
  0 siblings, 1 reply; 6+ messages in thread
From: Jeff King @ 2019-05-22  4:31 UTC (permalink / raw)
  To: Bryan Turner; +Cc: Git Users

Since 7171d8c15f (upload-pack: send symbolic ref information as
capability, 2013-09-17), we've sent cloning and fetching clients special
information about which branch HEAD is pointing to, so that they don't
have to guess based on matching up commit ids.

However, this feature has never worked properly with the GIT_NAMESPACE
feature.  Because upload-pack uses head_ref_namespaced(find_symref), we
do find and report on refs/namespaces/foo/HEAD instead of the actual
HEAD of the repo. This makes sense, since the branch pointed to by the
top-level HEAD may not be advertised at all. But we do two things wrong:

  1. We report the full name refs/namespaces/foo/HEAD, instead of just
     HEAD. Meaning no client is going to bother doing anything with that
     symref, since we're not otherwise advertising it.

  2. We report the symref destination using its full name (e.g.,
     refs/namespaces/foo/refs/heads/master). That's similarly useless to
     the client, who only saw "refs/heads/master" in the advertisement.

We should be stripping the namespace prefix off of both places (which
this patch fixes).

Likely nobody noticed because bug (1) means that from the client's
perspective, we did not report on HEAD at all. And thus it uses the
pre-7171d8c15f fallback code to guess the correct HEAD, which is usually
right. It only falls down in ambiguous cases (like the one laid out in
the included test).

This also means that we don't have to worry about breaking anybody who
was putting pre-stripped names into their namespace symrefs when we fix
bug (2). Because of bug (1), nobody would have been using the symref we
advertised in the first place (not to mention that those symrefs would
have appeared broken for any non-namespaced access).

Note that we have separate fixes here for the v0 and v2 protocols. The
symref advertisement moved in v2 to be a part of the ls-refs command.
This actually gets part (1) right, since the symref annotation
piggy-backs on the existing ref advertisement, which is properly
stripped. But it still needs a fix for part (2). The included tests
cover both protocols.

Reported-by: Bryan Turner <bturner@atlassian.com>
Signed-off-by: Jeff King <peff@peff.net>
---
This is the same as my earlier fix, but with the v2 bit added, and of
course tests and a commit message. Thanks (as usual) for a helpful bug
report.

I don't know if we have a general philosophy for testing v0 versus v2.
Without specifying the protocol at all, we'd catch the former on a
regular run and the latter under a GIT_TEST_PROTOCOL_VERSION=2 run. So
we _could_ just rely on that, but since I had to do two separate fixes,
it made sense to me to include explicit tests.

 ls-refs.c                        |  3 ++-
 t/t5509-fetch-push-namespaces.sh | 28 ++++++++++++++++++++++++++++
 upload-pack.c                    |  4 ++--
 3 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/ls-refs.c b/ls-refs.c
index 0a7dbc6442..818aef70a0 100644
--- a/ls-refs.c
+++ b/ls-refs.c
@@ -57,7 +57,8 @@ static int send_ref(const char *refname, const struct object_id *oid,
 		if (!symref_target)
 			die("'%s' is a symref but it is not?", refname);
 
-		strbuf_addf(&refline, " symref-target:%s", symref_target);
+		strbuf_addf(&refline, " symref-target:%s",
+			    strip_namespace(symref_target));
 	}
 
 	if (data->peel) {
diff --git a/t/t5509-fetch-push-namespaces.sh b/t/t5509-fetch-push-namespaces.sh
index c88df78c0b..75cbfcc392 100755
--- a/t/t5509-fetch-push-namespaces.sh
+++ b/t/t5509-fetch-push-namespaces.sh
@@ -124,4 +124,32 @@ test_expect_success 'try to update a hidden full ref' '
 	test_must_fail git -C original push pushee-namespaced master
 '
 
+test_expect_success 'set up ambiguous HEAD' '
+	git init ambiguous &&
+	(
+		cd ambiguous &&
+		git commit --allow-empty -m foo &&
+		git update-ref refs/namespaces/ns/refs/heads/one HEAD &&
+		git update-ref refs/namespaces/ns/refs/heads/two HEAD &&
+		git symbolic-ref refs/namespaces/ns/HEAD \
+			refs/namespaces/ns/refs/heads/two
+	)
+'
+
+test_expect_success 'clone chooses correct HEAD (v0)' '
+	GIT_NAMESPACE=ns git -c protocol.version=0 \
+		clone ambiguous ambiguous-v0 &&
+	echo refs/heads/two >expect &&
+	git -C ambiguous-v0 symbolic-ref HEAD >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'clone chooses correct HEAD (v2)' '
+	GIT_NAMESPACE=ns git -c protocol.version=2 \
+		clone ambiguous ambiguous-v2 &&
+	echo refs/heads/two >expect &&
+	git -C ambiguous-v2 symbolic-ref HEAD >actual &&
+	test_cmp expect actual
+'
+
 test_done
diff --git a/upload-pack.c b/upload-pack.c
index 24298913c0..4d2129e7fc 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1037,8 +1037,8 @@ static int find_symref(const char *refname, const struct object_id *oid,
 	symref_target = resolve_ref_unsafe(refname, 0, NULL, &flag);
 	if (!symref_target || (flag & REF_ISSYMREF) == 0)
 		die("'%s' is a symref but it is not?", refname);
-	item = string_list_append(cb_data, refname);
-	item->util = xstrdup(symref_target);
+	item = string_list_append(cb_data, strip_namespace(refname));
+	item->util = xstrdup(strip_namespace(symref_target));
 	return 0;
 }
 
-- 
2.22.0.rc1.539.g7bfcdfe86d




^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH] upload-pack: strip namespace from symref data
  2019-05-22  4:31   ` [PATCH] upload-pack: strip namespace from symref data Jeff King
@ 2019-05-22 10:33     ` Ævar Arnfjörð Bjarmason
  2019-05-23  6:11       ` [PATCH v2] " Jeff King
  0 siblings, 1 reply; 6+ messages in thread
From: Ævar Arnfjörð Bjarmason @ 2019-05-22 10:33 UTC (permalink / raw)
  To: Jeff King; +Cc: Bryan Turner, Git Users


On Wed, May 22 2019, Jeff King wrote:

> Since 7171d8c15f (upload-pack: send symbolic ref information as
> capability, 2013-09-17), we've sent cloning and fetching clients special
> information about which branch HEAD is pointing to, so that they don't
> have to guess based on matching up commit ids.

I'd add a 2nd paragraph here:

    The client will then use the corresponding code added in a45b5f0552
    ("connect: annotate refs with their symref information in
    get_remote_head()", 2013-09-17) to find the intended symref. See
    9907d1359c ("Merge branch 'jc/upload-pack-send-symref'", 2013-10-30)
    for the full set of changes related to this.

...because, see later...

> However, this feature has never worked properly with the GIT_NAMESPACE
> feature.  Because upload-pack uses head_ref_namespaced(find_symref), we
> do find and report on refs/namespaces/foo/HEAD instead of the actual
> HEAD of the repo. This makes sense, since the branch pointed to by the
> top-level HEAD may not be advertised at all. But we do two things wrong:
>
>   1. We report the full name refs/namespaces/foo/HEAD, instead of just
>      HEAD. Meaning no client is going to bother doing anything with that
>      symref, since we're not otherwise advertising it.
>
>   2. We report the symref destination using its full name (e.g.,
>      refs/namespaces/foo/refs/heads/master). That's similarly useless to
>      the client, who only saw "refs/heads/master" in the advertisement.
>
> We should be stripping the namespace prefix off of both places (which
> this patch fixes).
>
> Likely nobody noticed because bug (1) means that from the client's
> perspective, we did not report on HEAD at all. And thus it uses the
> pre-7171d8c15f fallback code to guess the correct HEAD, which is usually
> right. It only falls down in ambiguous cases (like the one laid out in
> the included test).

...because here you're talking about "the client's perspective" and "it
uses the pre-7171d8c15f [...] code", but this should say "the
pre-a45b5f0552 code", i.e. mention the commit that changed the *client*
logic.

Well, the client also "uses" the server-side code indirectly, but I
think it's easier to follow along if we note what both sides are (as I
dug up when reviewing this...).

> This also means that we don't have to worry about breaking anybody who
> was putting pre-stripped names into their namespace symrefs when we fix
> bug (2). Because of bug (1), nobody would have been using the symref we
> advertised in the first place (not to mention that those symrefs would
> have appeared broken for any non-namespaced access).
>
> Note that we have separate fixes here for the v0 and v2 protocols. The
> symref advertisement moved in v2 to be a part of the ls-refs command.
> This actually gets part (1) right, since the symref annotation
> piggy-backs on the existing ref advertisement, which is properly
> stripped. But it still needs a fix for part (2). The included tests
> cover both protocols.
>
> Reported-by: Bryan Turner <bturner@atlassian.com>
> Signed-off-by: Jeff King <peff@peff.net>
> ---
> This is the same as my earlier fix, but with the v2 bit added, and of
> course tests and a commit message. Thanks (as usual) for a helpful bug
> report.
>
> I don't know if we have a general philosophy for testing v0 versus v2.
> Without specifying the protocol at all, we'd catch the former on a
> regular run and the latter under a GIT_TEST_PROTOCOL_VERSION=2 run. So
> we _could_ just rely on that, but since I had to do two separate fixes,
> it made sense to me to include explicit tests.

Yeah, we shouldn't rely on the GIT_TEST_* stuff for coverage, it should
just be used to find things that fall between the cracks.

>  ls-refs.c                        |  3 ++-
>  t/t5509-fetch-push-namespaces.sh | 28 ++++++++++++++++++++++++++++
>  upload-pack.c                    |  4 ++--
>  3 files changed, 32 insertions(+), 3 deletions(-)
>
> diff --git a/ls-refs.c b/ls-refs.c
> index 0a7dbc6442..818aef70a0 100644
> --- a/ls-refs.c
> +++ b/ls-refs.c
> @@ -57,7 +57,8 @@ static int send_ref(const char *refname, const struct object_id *oid,
>  		if (!symref_target)
>  			die("'%s' is a symref but it is not?", refname);
>
> -		strbuf_addf(&refline, " symref-target:%s", symref_target);
> +		strbuf_addf(&refline, " symref-target:%s",
> +			    strip_namespace(symref_target));
>  	}
>
>  	if (data->peel) {
> diff --git a/t/t5509-fetch-push-namespaces.sh b/t/t5509-fetch-push-namespaces.sh
> index c88df78c0b..75cbfcc392 100755
> --- a/t/t5509-fetch-push-namespaces.sh
> +++ b/t/t5509-fetch-push-namespaces.sh
> @@ -124,4 +124,32 @@ test_expect_success 'try to update a hidden full ref' '
>  	test_must_fail git -C original push pushee-namespaced master
>  '
>
> +test_expect_success 'set up ambiguous HEAD' '
> +	git init ambiguous &&
> +	(
> +		cd ambiguous &&
> +		git commit --allow-empty -m foo &&
> +		git update-ref refs/namespaces/ns/refs/heads/one HEAD &&
> +		git update-ref refs/namespaces/ns/refs/heads/two HEAD &&
> +		git symbolic-ref refs/namespaces/ns/HEAD \
> +			refs/namespaces/ns/refs/heads/two
> +	)
> +'
> +
> +test_expect_success 'clone chooses correct HEAD (v0)' '
> +	GIT_NAMESPACE=ns git -c protocol.version=0 \
> +		clone ambiguous ambiguous-v0 &&
> +	echo refs/heads/two >expect &&
> +	git -C ambiguous-v0 symbolic-ref HEAD >actual &&
> +	test_cmp expect actual
> +'
> +
> +test_expect_success 'clone chooses correct HEAD (v2)' '
> +	GIT_NAMESPACE=ns git -c protocol.version=2 \
> +		clone ambiguous ambiguous-v2 &&
> +	echo refs/heads/two >expect &&
> +	git -C ambiguous-v2 symbolic-ref HEAD >actual &&
> +	test_cmp expect actual
> +'
> +
>  test_done
> diff --git a/upload-pack.c b/upload-pack.c
> index 24298913c0..4d2129e7fc 100644
> --- a/upload-pack.c
> +++ b/upload-pack.c
> @@ -1037,8 +1037,8 @@ static int find_symref(const char *refname, const struct object_id *oid,
>  	symref_target = resolve_ref_unsafe(refname, 0, NULL, &flag);
>  	if (!symref_target || (flag & REF_ISSYMREF) == 0)
>  		die("'%s' is a symref but it is not?", refname);
> -	item = string_list_append(cb_data, refname);
> -	item->util = xstrdup(symref_target);
> +	item = string_list_append(cb_data, strip_namespace(refname));
> +	item->util = xstrdup(strip_namespace(symref_target));
>  	return 0;
>  }

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH v2] upload-pack: strip namespace from symref data
  2019-05-22 10:33     ` Ævar Arnfjörð Bjarmason
@ 2019-05-23  6:11       ` Jeff King
  2019-05-28 17:01         ` Junio C Hamano
  0 siblings, 1 reply; 6+ messages in thread
From: Jeff King @ 2019-05-23  6:11 UTC (permalink / raw)
  To: Ævar Arnfjörð Bjarmason; +Cc: Bryan Turner, Git Users

On Wed, May 22, 2019 at 12:33:56PM +0200, Ævar Arnfjörð Bjarmason wrote:

> > Likely nobody noticed because bug (1) means that from the client's
> > perspective, we did not report on HEAD at all. And thus it uses the
> > pre-7171d8c15f fallback code to guess the correct HEAD, which is usually
> > right. It only falls down in ambiguous cases (like the one laid out in
> > the included test).
> 
> ...because here you're talking about "the client's perspective" and "it
> uses the pre-7171d8c15f [...] code", but this should say "the
> pre-a45b5f0552 code", i.e. mention the commit that changed the *client*
> logic.

Heh. I actually found this split, too, while digging in the history, but
thought it might be adding distracting details to the commit message.
But hey, if you're gonna read them _carefully_, I guess I can be more
precise. ;)

The fallback code isn't even from a45b5f0552; it's much older than that.
We started looking at ref annotations at all in fbb074c253 (remote: make
guess_remote_head() use exact HEAD lookup if it is available,
2009-02-27). The guessing code that falls back to comes from 8434c2f1af
(Build in clone, 2008-04-27).  I didn't bother finding the true origin
in the shell script. ;)

So instead of mentioning it at the top (because the bug really is just
on the server side), I re-wrote the paragraph about the client a little
more clearly and precisely:

    Likely nobody noticed because we tend to do the right thing anyway. Bug
    (1) means that we said nothing about HEAD (just refs/namespace/foo/HEAD).
    And so the client half of the code, from a45b5f0552 (connect: annotate
    refs with their symref information in get_remote_head(), 2013-09-17),
    does not annotate HEAD, and we use the fallback in guess_remote_head(),
    matching refs by object id. Which is usually right. It only falls down
    in ambiguous cases, like the one laid out in the included test.

Revised patch below.

-- >8 --
Subject: [PATCH] upload-pack: strip namespace from symref data

Since 7171d8c15f (upload-pack: send symbolic ref information as
capability, 2013-09-17), we've sent cloning and fetching clients special
information about which branch HEAD is pointing to, so that they don't
have to guess based on matching up commit ids.

However, this feature has never worked properly with the GIT_NAMESPACE
feature.  Because upload-pack uses head_ref_namespaced(find_symref), we
do find and report on refs/namespaces/foo/HEAD instead of the actual
HEAD of the repo. This makes sense, since the branch pointed to by the
top-level HEAD may not be advertised at all. But we do two things wrong:

  1. We report the full name refs/namespaces/foo/HEAD, instead of just
     HEAD. Meaning no client is going to bother doing anything with that
     symref, since we're not otherwise advertising it.

  2. We report the symref destination using its full name (e.g.,
     refs/namespaces/foo/refs/heads/master). That's similarly useless to
     the client, who only saw "refs/heads/master" in the advertisement.

We should be stripping the namespace prefix off of both places (which
this patch fixes).

Likely nobody noticed because we tend to do the right thing anyway. Bug
(1) means that we said nothing about HEAD (just refs/namespace/foo/HEAD).
And so the client half of the code, from a45b5f0552 (connect: annotate
refs with their symref information in get_remote_head(), 2013-09-17),
does not annotate HEAD, and we use the fallback in guess_remote_head(),
matching refs by object id. Which is usually right. It only falls down
in ambiguous cases, like the one laid out in the included test.

This also means that we don't have to worry about breaking anybody who
was putting pre-stripped names into their namespace symrefs when we fix
bug (2). Because of bug (1), nobody would have been using the symref we
advertised in the first place (not to mention that those symrefs would
have appeared broken for any non-namespaced access).

Note that we have separate fixes here for the v0 and v2 protocols. The
symref advertisement moved in v2 to be a part of the ls-refs command.
This actually gets part (1) right, since the symref annotation
piggy-backs on the existing ref advertisement, which is properly
stripped. But it still needs a fix for part (2). The included tests
cover both protocols.

Reported-by: Bryan Turner <bturner@atlassian.com>
Signed-off-by: Jeff King <peff@peff.net>
---
 ls-refs.c                        |  3 ++-
 t/t5509-fetch-push-namespaces.sh | 28 ++++++++++++++++++++++++++++
 upload-pack.c                    |  4 ++--
 3 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/ls-refs.c b/ls-refs.c
index 0a7dbc6442..818aef70a0 100644
--- a/ls-refs.c
+++ b/ls-refs.c
@@ -57,7 +57,8 @@ static int send_ref(const char *refname, const struct object_id *oid,
 		if (!symref_target)
 			die("'%s' is a symref but it is not?", refname);
 
-		strbuf_addf(&refline, " symref-target:%s", symref_target);
+		strbuf_addf(&refline, " symref-target:%s",
+			    strip_namespace(symref_target));
 	}
 
 	if (data->peel) {
diff --git a/t/t5509-fetch-push-namespaces.sh b/t/t5509-fetch-push-namespaces.sh
index c88df78c0b..75cbfcc392 100755
--- a/t/t5509-fetch-push-namespaces.sh
+++ b/t/t5509-fetch-push-namespaces.sh
@@ -124,4 +124,32 @@ test_expect_success 'try to update a hidden full ref' '
 	test_must_fail git -C original push pushee-namespaced master
 '
 
+test_expect_success 'set up ambiguous HEAD' '
+	git init ambiguous &&
+	(
+		cd ambiguous &&
+		git commit --allow-empty -m foo &&
+		git update-ref refs/namespaces/ns/refs/heads/one HEAD &&
+		git update-ref refs/namespaces/ns/refs/heads/two HEAD &&
+		git symbolic-ref refs/namespaces/ns/HEAD \
+			refs/namespaces/ns/refs/heads/two
+	)
+'
+
+test_expect_success 'clone chooses correct HEAD (v0)' '
+	GIT_NAMESPACE=ns git -c protocol.version=0 \
+		clone ambiguous ambiguous-v0 &&
+	echo refs/heads/two >expect &&
+	git -C ambiguous-v0 symbolic-ref HEAD >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'clone chooses correct HEAD (v2)' '
+	GIT_NAMESPACE=ns git -c protocol.version=2 \
+		clone ambiguous ambiguous-v2 &&
+	echo refs/heads/two >expect &&
+	git -C ambiguous-v2 symbolic-ref HEAD >actual &&
+	test_cmp expect actual
+'
+
 test_done
diff --git a/upload-pack.c b/upload-pack.c
index 24298913c0..4d2129e7fc 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1037,8 +1037,8 @@ static int find_symref(const char *refname, const struct object_id *oid,
 	symref_target = resolve_ref_unsafe(refname, 0, NULL, &flag);
 	if (!symref_target || (flag & REF_ISSYMREF) == 0)
 		die("'%s' is a symref but it is not?", refname);
-	item = string_list_append(cb_data, refname);
-	item->util = xstrdup(symref_target);
+	item = string_list_append(cb_data, strip_namespace(refname));
+	item->util = xstrdup(strip_namespace(symref_target));
 	return 0;
 }
 
-- 
2.22.0.rc1.549.gadb183c4cb


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v2] upload-pack: strip namespace from symref data
  2019-05-23  6:11       ` [PATCH v2] " Jeff King
@ 2019-05-28 17:01         ` Junio C Hamano
  0 siblings, 0 replies; 6+ messages in thread
From: Junio C Hamano @ 2019-05-28 17:01 UTC (permalink / raw)
  To: Jeff King; +Cc: Ævar Arnfjörð Bjarmason, Bryan Turner, Git Users

Jeff King <peff@peff.net> writes:

> Revised patch below.

Thanks.  I never took the "separate namespaces" thing as a serious
feature, so it is understandable that I ignored it completely when I
did 7171d8c15f.

The explanation and the code both look good.  Thanks.


> -- >8 --
> Subject: [PATCH] upload-pack: strip namespace from symref data
>
> Since 7171d8c15f (upload-pack: send symbolic ref information as
> capability, 2013-09-17), we've sent cloning and fetching clients special
> information about which branch HEAD is pointing to, so that they don't
> have to guess based on matching up commit ids.
>
> However, this feature has never worked properly with the GIT_NAMESPACE
> feature.  Because upload-pack uses head_ref_namespaced(find_symref), we
> do find and report on refs/namespaces/foo/HEAD instead of the actual
> HEAD of the repo. This makes sense, since the branch pointed to by the
> top-level HEAD may not be advertised at all. But we do two things wrong:
>
>   1. We report the full name refs/namespaces/foo/HEAD, instead of just
>      HEAD. Meaning no client is going to bother doing anything with that
>      symref, since we're not otherwise advertising it.
>
>   2. We report the symref destination using its full name (e.g.,
>      refs/namespaces/foo/refs/heads/master). That's similarly useless to
>      the client, who only saw "refs/heads/master" in the advertisement.
>
> We should be stripping the namespace prefix off of both places (which
> this patch fixes).
>
> Likely nobody noticed because we tend to do the right thing anyway. Bug
> (1) means that we said nothing about HEAD (just refs/namespace/foo/HEAD).
> And so the client half of the code, from a45b5f0552 (connect: annotate
> refs with their symref information in get_remote_head(), 2013-09-17),
> does not annotate HEAD, and we use the fallback in guess_remote_head(),
> matching refs by object id. Which is usually right. It only falls down
> in ambiguous cases, like the one laid out in the included test.
>
> This also means that we don't have to worry about breaking anybody who
> was putting pre-stripped names into their namespace symrefs when we fix
> bug (2). Because of bug (1), nobody would have been using the symref we
> advertised in the first place (not to mention that those symrefs would
> have appeared broken for any non-namespaced access).
>
> Note that we have separate fixes here for the v0 and v2 protocols. The
> symref advertisement moved in v2 to be a part of the ls-refs command.
> This actually gets part (1) right, since the symref annotation
> piggy-backs on the existing ref advertisement, which is properly
> stripped. But it still needs a fix for part (2). The included tests
> cover both protocols.
>
> Reported-by: Bryan Turner <bturner@atlassian.com>
> Signed-off-by: Jeff King <peff@peff.net>
> ---
>  ls-refs.c                        |  3 ++-
>  t/t5509-fetch-push-namespaces.sh | 28 ++++++++++++++++++++++++++++
>  upload-pack.c                    |  4 ++--
>  3 files changed, 32 insertions(+), 3 deletions(-)
>
> diff --git a/ls-refs.c b/ls-refs.c
> index 0a7dbc6442..818aef70a0 100644
> --- a/ls-refs.c
> +++ b/ls-refs.c
> @@ -57,7 +57,8 @@ static int send_ref(const char *refname, const struct object_id *oid,
>  		if (!symref_target)
>  			die("'%s' is a symref but it is not?", refname);
>  
> -		strbuf_addf(&refline, " symref-target:%s", symref_target);
> +		strbuf_addf(&refline, " symref-target:%s",
> +			    strip_namespace(symref_target));
>  	}
>  
>  	if (data->peel) {
> diff --git a/t/t5509-fetch-push-namespaces.sh b/t/t5509-fetch-push-namespaces.sh
> index c88df78c0b..75cbfcc392 100755
> --- a/t/t5509-fetch-push-namespaces.sh
> +++ b/t/t5509-fetch-push-namespaces.sh
> @@ -124,4 +124,32 @@ test_expect_success 'try to update a hidden full ref' '
>  	test_must_fail git -C original push pushee-namespaced master
>  '
>  
> +test_expect_success 'set up ambiguous HEAD' '
> +	git init ambiguous &&
> +	(
> +		cd ambiguous &&
> +		git commit --allow-empty -m foo &&
> +		git update-ref refs/namespaces/ns/refs/heads/one HEAD &&
> +		git update-ref refs/namespaces/ns/refs/heads/two HEAD &&
> +		git symbolic-ref refs/namespaces/ns/HEAD \
> +			refs/namespaces/ns/refs/heads/two
> +	)
> +'
> +
> +test_expect_success 'clone chooses correct HEAD (v0)' '
> +	GIT_NAMESPACE=ns git -c protocol.version=0 \
> +		clone ambiguous ambiguous-v0 &&
> +	echo refs/heads/two >expect &&
> +	git -C ambiguous-v0 symbolic-ref HEAD >actual &&
> +	test_cmp expect actual
> +'
> +
> +test_expect_success 'clone chooses correct HEAD (v2)' '
> +	GIT_NAMESPACE=ns git -c protocol.version=2 \
> +		clone ambiguous ambiguous-v2 &&
> +	echo refs/heads/two >expect &&
> +	git -C ambiguous-v2 symbolic-ref HEAD >actual &&
> +	test_cmp expect actual
> +'
> +
>  test_done
> diff --git a/upload-pack.c b/upload-pack.c
> index 24298913c0..4d2129e7fc 100644
> --- a/upload-pack.c
> +++ b/upload-pack.c
> @@ -1037,8 +1037,8 @@ static int find_symref(const char *refname, const struct object_id *oid,
>  	symref_target = resolve_ref_unsafe(refname, 0, NULL, &flag);
>  	if (!symref_target || (flag & REF_ISSYMREF) == 0)
>  		die("'%s' is a symref but it is not?", refname);
> -	item = string_list_append(cb_data, refname);
> -	item->util = xstrdup(symref_target);
> +	item = string_list_append(cb_data, strip_namespace(refname));
> +	item->util = xstrdup(strip_namespace(symref_target));
>  	return 0;
>  }

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2019-05-28 17:01 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-21  2:59 HEAD and namespaces Bryan Turner
2019-05-21 21:46 ` Jeff King
2019-05-22  4:31   ` [PATCH] upload-pack: strip namespace from symref data Jeff King
2019-05-22 10:33     ` Ævar Arnfjörð Bjarmason
2019-05-23  6:11       ` [PATCH v2] " Jeff King
2019-05-28 17:01         ` Junio C Hamano

git@vger.kernel.org list mirror (unofficial, one of many)

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://public-inbox.org/git
	git clone --mirror http://ou63pmih66umazou.onion/git
	git clone --mirror http://czquwvybam4bgbro.onion/git
	git clone --mirror http://hjrcffqmbrq6wope.onion/git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V1 git git/ https://public-inbox.org/git \
		git@vger.kernel.org
	public-inbox-index git

Example config snippet for mirrors.
Newsgroups are available over NNTP:
	nntp://news.public-inbox.org/inbox.comp.version-control.git
	nntp://ou63pmih66umazou.onion/inbox.comp.version-control.git
	nntp://czquwvybam4bgbro.onion/inbox.comp.version-control.git
	nntp://hjrcffqmbrq6wope.onion/inbox.comp.version-control.git
	nntp://news.gmane.io/gmane.comp.version-control.git
 note: .onion URLs require Tor: https://www.torproject.org/

code repositories for the project(s) associated with this inbox:

	https://80x24.org/mirrors/git.git

AGPL code for this site: git clone https://public-inbox.org/public-inbox.git