Git development
 help / color / mirror / Atom feed
* repack behaves unexpectedly if called in a bare _subrepo_
@ 2007-06-17 15:38 Alex Riesen
  2007-06-17 17:22 ` Junio C Hamano
  0 siblings, 1 reply; 3+ messages in thread
From: Alex Riesen @ 2007-06-17 15:38 UTC (permalink / raw)
  To: git

Just a heads-up. This took me by surprise. And I suppose it will be
the same even if the subrepo is a real subproject.

To illustrate:

(
    mkdir super && cd super
    git init
    echo 0 > a; git add a; git commit -m0
    echo 2 >> a; git commit -a -m1
    echo 3 >> a; git commit -a -m2
    cp -a .git sub
    cd sub
    git --bare config --bool core.bare true
    git repack -a -d
)

Unexpectedly, the "super" repo is repacked, even though repack is
called from the "sub". To actually repack the repo in the current
directory if it is bare, use "git --bare repack ...".

This kind of makes core.bare pointless: can't rely on it anyway.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: repack behaves unexpectedly if called in a bare _subrepo_
  2007-06-17 15:38 repack behaves unexpectedly if called in a bare _subrepo_ Alex Riesen
@ 2007-06-17 17:22 ` Junio C Hamano
  2007-06-17 19:27   ` Alex Riesen
  0 siblings, 1 reply; 3+ messages in thread
From: Junio C Hamano @ 2007-06-17 17:22 UTC (permalink / raw)
  To: Alex Riesen; +Cc: git

Alex Riesen <raa.lkml@gmail.com> writes:

> Just a heads-up. This took me by surprise. And I suppose it will be
> the same even if the subrepo is a real subproject.
>
> To illustrate:
>
> (
>     mkdir super && cd super
>     git init
>     echo 0 > a; git add a; git commit -m0
>     echo 2 >> a; git commit -a -m1
>     echo 3 >> a; git commit -a -m2
>     cp -a .git sub
>     cd sub
>     git --bare config --bool core.bare true
>     git repack -a -d
> )

I think what is happening is that in that bare repository 'sub',
the sequence we find where a git repository is is defined so
that the current directory is checked way after "one of my
parents with .git/ subdirectory" check.

If super (a directory) has a subproject sub as its subproject,
then you would be copying .git to sub/.git, not sub/, wouldn't
you?  What are you trying to achieve?

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: repack behaves unexpectedly if called in a bare _subrepo_
  2007-06-17 17:22 ` Junio C Hamano
@ 2007-06-17 19:27   ` Alex Riesen
  0 siblings, 0 replies; 3+ messages in thread
From: Alex Riesen @ 2007-06-17 19:27 UTC (permalink / raw)
  To: Junio C Hamano; +Cc: git

Junio C Hamano, Sun, Jun 17, 2007 19:22:39 +0200:
> > Just a heads-up. This took me by surprise. And I suppose it will be
> > the same even if the subrepo is a real subproject.
> >
> > To illustrate:
> >
> > (
> >     mkdir super && cd super
> >     git init
> >     echo 0 > a; git add a; git commit -m0
> >     echo 2 >> a; git commit -a -m1
> >     echo 3 >> a; git commit -a -m2
> >     cp -a .git sub
> >     cd sub
> >     git --bare config --bool core.bare true
> >     git repack -a -d
> > )
> 
> I think what is happening is that in that bare repository 'sub',
> the sequence we find where a git repository is is defined so
> that the current directory is checked way after "one of my
> parents with .git/ subdirectory" check.

Yes, that's what I found. Couldn't convince myself it could be wrong,
feels right to me.

> If super (a directory) has a subproject sub as its subproject,
> then you would be copying .git to sub/.git, not sub/, wouldn't
> you?

Yes, very likely. But it is not a super project in subproject sense,
and why? Is there a reason I can't have a bare subproject?

> What are you trying to achieve?

I mirror some projects I am interested in. There is a directory where
all the mirrors are: ~/src. It contains the git/git-svn/git-cvs
repositories _and_ the scripts which do the mirroring:

$ ls -a
./                  gitk/          motion.gitsvn/  uclibc.gitsvn/
../                 gnulib.git/    moto4lin/       udev/
busybox.gitsvn/     gpsbabel.git/  mutt/           udhcp.gitsvn/
coreutils.git/      historic/      rt2500/         update-gits*
count-git-objects*  .htaccess      sparse/         use-src/
dclock/             iproute2.git/  sunset/         wireless-2.6.git/
git/                klibc/         tig/
.git/               linux.git/     torsmo/
.gitignore          mimesplit3/    tuner/

The super in my example is this ~/src, and it keeps track of changes
to the scripts (some mirrors can be complicated, so I wanted to be
prepared). Felt kind of natural to do it this way. Until I tried to
repack the mirrored repos, and it didn't work as I expected. The repos
are not tracked as subprojects (in fact, they're .gitignored).
There is no point for me to have working directories for the mirrors:
it is my home server, if I do something, I do it elsewhere.

It is not exactly a "problem", and anyway there is a couple of
solutions (pass --bare, create <mirror>/.git, for instance), I just
fall into it and though it'd be a good idea to let others know.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2007-06-17 19:27 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-06-17 15:38 repack behaves unexpectedly if called in a bare _subrepo_ Alex Riesen
2007-06-17 17:22 ` Junio C Hamano
2007-06-17 19:27   ` Alex Riesen

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox