Re: [RFC PATCH 06/19] Factor ref updating out of fetch_with_import

[Sverre Rabbelier <srabbelier@gmail.com>]

Heya,

On Fri, Oct 30, 2009 at 17:04, Daniel Barkalow <barkalow@iabervon.org> wrote:
> I think you have the sides backwards

Yup, sorry, got them the wrong way around for some reason.

> On the other hand, I think it would make sense to use the same style of
> refspec between the helper and transport-helper.c such that the helper
> reports something like:
>
> refs/svn/origin/trunk:refs/heads/trunkr
> refs/svn/origin/branches/*:refs/heads/*
> refs/svn/origin/tags/*:refs/tags/*
>
> "list" gives:
>
> 000000...000 refs/heads/trunkr
>
> "import refs/heads/trunkr" imports the objects, but the refspecs have to
> be consulted by transport-helper.c in order to determine what ref to read
> to get the value of refs/heads/trunkr. Instead of getting the value with
> read_ref("refs/heads/trunkr", ...) like it does now, it would do
> read_ref("refs/svn/origin/trunk", ...). And systems like p4 that don't
> have a useful standard just wouldn't support the "refspec" command and
> people would have to do site-specific configuration to get anything
> useful.

Yes, that sounds very reasonable, and I think that's the right way to
go. This leaves us with only one thing, we need a remote HEAD for 'git
clone hg::/path/to/repo' to work and have it check out a branch, I
think a seperate 'head' command might be appropriate? If supported it
returns the which local symref (e.g. 'refs/heads/trunkr' in the svn
case) should be pointed at by HEAD. If not supported we can just not
set it and clone will give the default 'no remote HEAD, nothing
checked out' message, which would probably be best for p4?

-- 
Cheers,

Sverre Rabbelier