Re: A simple script to do the reverse of git-push

2005-08-11 Thread Petr Baudis
Dear diary, on Tue, Aug 09, 2005 at 12:42:36AM CEST, I got a letter
where Junio C Hamano <[EMAIL PROTECTED]> told me that...
> Johannes Schindelin <[EMAIL PROTECTED]> writes:
> 
> > BTW, if you are lazy, like me, you just pull from Junio once in a while 
> > and do a "make test". Turns out there is a missing dependency though:
> >
> > peek-remote.o: cache.h
> >
> > which in my case lead to a git-peek-remote program which was unable to 
> > peek any ref.
> 
> You are right.  Thanks for noticing.
> 
> $ (make clean ; make ) >/dev/null 2>/dev/null
> $ touch cache.h
> $ make 2>&1 | grep peek-remote
> cc -g -O2 -Wall  '-DSHA1_HEADER=' -o git-peek-remote 
> peek-remote.o libgit.a -lz -lcrypto
> 
> I think recent "depend on object files" Makefile patch broke
> some things.

Indeed. I took care to make the new dependencies a superset of previous
situation when removing the explicit dependencies list, but before,
rebuilding of libgit.a caused rebuilt from source of all the commands,
which wouldn't happen anymore after adding the object files, which this
way sneakily removed an implicit dependency of the command sources on
$(LIB_H). Eek.

-- 
Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
If you want the holes in your knowledge showing up try teaching
someone.  -- Alan Cox
-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Re: A simple script to do the reverse of git-push

2005-08-08 Thread Johannes Schindelin
Hi,

On Mon, 8 Aug 2005, Junio C Hamano wrote:

> We would need something like this.
>
> [...]
>
> +$(patsubst git-%,%.o,$(PROG)): $(LIB_H)

A short

for i in git-*; do \
c=$(echo $i|sed "s/git-\(.*\)/\1.c/g")
if [ -e $c ]; then \
if ! grep cache\\.h $c; then \
echo $c does not depend on cache.h \
fi \
fi \
done

(actually tested) reveals that only get-tar-commit-id.c and stripspace.c 
do not depend on cache.h or rev-cache.h. Thus, your patch is the most 
elegant solution to my problem :-)

Ciao,
Dscho

-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Re: A simple script to do the reverse of git-push

2005-08-08 Thread Junio C Hamano
Johannes Schindelin <[EMAIL PROTECTED]> writes:

> BTW, if you are lazy, like me, you just pull from Junio once in a while 
> and do a "make test". Turns out there is a missing dependency though:
>
> peek-remote.o: cache.h
>
> which in my case lead to a git-peek-remote program which was unable to 
> peek any ref.

You are right.  Thanks for noticing.

$ (make clean ; make ) >/dev/null 2>/dev/null
$ touch cache.h
$ make 2>&1 | grep peek-remote
cc -g -O2 -Wall  '-DSHA1_HEADER=' -o git-peek-remote 
peek-remote.o libgit.a -lz -lcrypto

I think recent "depend on object files" Makefile patch broke
some things.

We would need something like this.
---
# - pu: ls-remote: drop storing operation and add documentation.
# + (working tree)
diff --git a/Makefile b/Makefile
--- a/Makefile
+++ b/Makefile
@@ -172,6 +172,7 @@ init-db.o: init-db.c
$(CC) -c $(CFLAGS) -DDEFAULT_GIT_TEMPLATE_DIR='"$(template_dir)"' $*.c
 
 $(LIB_OBJS): $(LIB_H)
+$(patsubst git-%,%.o,$(PROG)): $(LIB_H)
 $(DIFF_OBJS): diffcore.h
 
 $(LIB_FILE): $(LIB_OBJS)





-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html


A simple script to do the reverse of git-push

2005-08-08 Thread Johannes Schindelin
Hi list,

I mentioned in another mail that I needed the opposite of git-push, namely 
getting all heads, and if they are strict parents of the local refs, just 
update them.

Well, Junio pointed out that it's easy using the available tools, and he 
was right. The result is attached (and tested...).

This script makes it easy to work with a central repository, multiple 
heads and users, in a CVS style: to start your day, you call the attached 
script, and after you have committed something, you can do a push on the 
repository. Sometimes, the push fails, then you have to pull that branch 
in order to merge it before trying a push again.

BTW, if you are lazy, like me, you just pull from Junio once in a while 
and do a "make test". Turns out there is a missing dependency though:

peek-remote.o: cache.h

which in my case lead to a git-peek-remote program which was unable to 
peek any ref.

Ciao,
Dscho
#!/bin/sh

. git-sh-setup-script || die "Not a git archive"

. git-parse-remote "$@"
repo="$_remote_repo"

git-ls-remote-script --heads "$repo" | while read sha1 refname; do
shortname=$(basename $refname)
if [ -e $GIT_DIR/$refname ]; then
if [ $sha1 = $(cat $GIT_DIR/$refname) ]; then
echo $shortname up to date. 1>&2
continue
fi
fi

git-cat-file commit $sha1 >/dev/null 2>&1 ||
git-fetch-script $repo $shortname

if [ ! -e $GIT_DIR/$refname ]; then
echo Got new head $shortname 1>&2
echo $sha1 > $GIT_DIR/$refname
else
origsha1=$(cat $GIT_DIR/$refname)
if [ -z "$(git-rev-list $origsha1 ^$sha1)" ]; then
echo Updated head $shortname 1>&2
echo $sha1 > $GIT_DIR/$refname
else
saveref=$GIT_DIR/FETCH_HEAD_$shortname
echo Error: $shortname not a child of remote head 1>&2
echo "  Saved remote head to $saveref" 1>&2
echo $sha1 > $saveref
fi
fi
done