[PATCH 1/2] Move setup_reference to after writing the config file

[PATCH 1/2] Move setup_reference to after writing the config file

[Sverre Rabbelier]

Currently setup_reference calls remote_get before the remote is
added to the config file. This will result in an improperly
configured remote (in memory) if later on remote_get is called
with an argument that is not equal to the initial remote_get call
in setup_reference. Fix this by delaying the remote_get call until
after the remote has been added to the config file.
---

  It is also possible to split setup_reference so that the part that
  calls remote_get is where the entire setup_referene call is moved
  to now, this also works, and results in a far simpler patch.

  This change is required for the next patch.

 builtin/clone.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index 05f8fb4..f878563 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -470,9 +470,6 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 	 */
 	unsetenv(CONFIG_ENVIRONMENT);
 
-	if (option_reference)
-		setup_reference(git_dir);
-
 	git_config(git_default_config, NULL);
 
 	if (option_bare) {
@@ -504,6 +501,9 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 		strbuf_reset(&key);
 	}
 
+	if (option_reference)
+		setup_reference(git_dir);
+
 	fetch_pattern = value.buf;
 	refspec = parse_fetch_refspec(1, &fetch_pattern);
 
-- 
1.7.0.2.325.g634e5

[PATCH 2/2] clone: pass the remote name to remote_get

[Sverre Rabbelier]

Currently when using a remote helper to clone a repository, the
remote helper will be passed the url of the target repository as
first argument (which represents the name of the remote). This name
is extracted from transport->remote->name, which is set by
builtin/clone.c when it calls remote_get with argv[0] as argument.

Fix this by passing the name remote will be set up as instead.
---

  I've spent quite a few hours trying to figure out why this broke
  t5700, so I would appreciate feedback from those with more
  knowledge about hardlinking and clone/fetch in general as to
  whether this couild cause any problems.

 builtin/clone.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index f878563..068d61f 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -513,7 +513,7 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 		refs = clone_local(path, git_dir);
 		mapped_refs = wanted_peer_refs(refs, refspec);
 	} else {
-		struct remote *remote = remote_get(argv[0]);
+		struct remote *remote = remote_get(option_origin);
 		transport = transport_get(remote, remote->url[0]);
 
 		if (!transport->get_refs_list || !transport->fetch)
-- 
1.7.0.2.325.g634e5

Re: [PATCH 2/2] clone: pass the remote name to remote_get

[Sverre Rabbelier]

Heya,

On Sun, Mar 21, 2010 at 21:19, Sverre Rabbelier <srabbelier@gmail.com> wrote:
> Currently when using a remote helper to clone a repository, the
> remote helper will be passed the url of the target repository as
> first argument (which represents the name of the remote). This name
> is extracted from transport->remote->name, which is set by
> builtin/clone.c when it calls remote_get with argv[0] as argument.
>
> Fix this by passing the name remote will be set up as instead.

Ah, I forgot a sign-off it seems. Also, these two patches were a bit
more involved before (I didn't arrive at their current form for quite
a few hours), so I guess it could makes sense to just squash them into
one patch. Whichever is preferable.

-- 
Cheers,

Sverre Rabbelier