git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] Fix branches file configuration

From
Daniel Barkalow <barkalow@iabervon.org>
Date
Mar 26, 2008, 06:51 UTC
Message-ID
<alpine.LNX.1.00.0803260240570.19665@iabervon.org>
In-Reply-To
<7v3aqel5ms.fsf@gitster.siamese.dyndns.org>
On Tue, 25 Mar 2008, Junio C Hamano wrote:
Show 42 quoted lines
> Daniel Barkalow <barkalow@iabervon.org> writes:
> 
> > It turns out that .git/branches/foo should fetch into refs/heads/foo.
> >
> > Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
> > ---
> > We still don't have a test for that bit of the behavior, so I could have 
> > it still doing the wrong thing. But it at least should do what I think 
> > people want.
> >
> > Andrew, could you give this a try, on top of current master (or, for that 
> > matter, any released version that doesn't work), and let me know if it 
> > does the wrong thing?
> 
> To give bystanders a bit of context, Andrew had an old Cogito-style
> 
> 	.git/branches/foo
> 
> that had a URL (and no #frag) in it.  "git fetch foo" should have fetched
> from the 'master' branch of the named repository and stored it in his
> refs/heads/foo, and git up to 1.5.3.X series do so correctly, but we had
> this uncaught regression in 1.5.4 (the support for "branches" file is not
> deprecated, even though I personally do not see a reason to use it in new
> repositories).
> 
> Now, Daniel, I was independently fixing this one and had a question about
> this code.
> 
> What is the deal with this "slash" magic?  It appears that when you are
> fed:
> 
> 	git fetch foo/bar
> 
> you look up ".git/branches/foo" for $URL and #$branch, then use "$URL/bar"
> as the URL and $branch (or 'master' if you did not see '#') as the branch
> to fetch from, and store it in "foo/bar" locally.  Is that really what
> should have happened?  I am reasonably sure about "fetching from $URL/bar"
> part, but I am unsure about "store in foo/bar" part.
> 
> Any Cogito survivers out there who knows how this was supposed to have
> worked?  We can go back and look at git-fetch.sh (in contrib/examples),
> but I'd rather be lazy ;-)

This code actually is from Linus, nearly 2 years ago, converting git-push.sh; I then moved it from builtin-push.c to remote.c and later added support to remote.c for the fetch side aspects while leaving that part of the code alone. Of course, I didn't realize at any point that, as a fetch configuration, a branches configuration ever stored things anywhere, and it looks like it didn't with a slash. (See line 115 of 755225de:git-parse-remote.sh)

So I guess there should be:
> >  	add_url_alias(remote, p);
> > -	add_fetch_refspec(remote, branch);
        if (!slash)
> > +	strbuf_addf(&branch, ":refs/heads/%s", remote->name);
> > +	add_fetch_refspec(remote, strbuf_detach(&branch, 0));
> >  	remote->fetch_tags = 1; /* always auto-follow */
> >  }
	-Daniel
*This .sig left intentionally blank*
Previous: Andrew MortonNext: Junio C Hamano
Message 5 of 8 in “Fix branches file configuration”
  1. Fix branches file configurationDaniel Barkalow, Mar 25, 2008
  2. Andrew MortonMar 25, 2008
  3. Junio C HamanoMar 25, 2008
  4. Andrew MortonMar 26, 2008
  5. Daniel BarkalowMar 26, 2008
  6. Junio C HamanoMar 26, 2008
  7. Junio C HamanoMar 26, 2008
  8. git-fetch test: test tracking fetch results, not just FETCH_HEADJunio C Hamano, Mar 26, 2008

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.