threads / patch / 13764

patchrebase --interactive: Compute upstream SHA1 before switching branches

Subject: [PATCH] rebase --interactive: Compute upstream SHA1 before switching branches

## tl;dr

4 messages between Jun 2, 2008 and Jun 3, 2008. Diffs are folded; open one to read it.

replies: 3people: 3as markdown or json

Johannes Sixt· Jun 2, 2008, 14:01 UTC · lore

From: Johannes Sixt <johannes.sixt@telecom.at>

If the upstream argument to rebase (the first argument) was relative to HEAD and the name of the branch to rebase (the second argument) was given, the upstream would have been interpreted relative to the second argument. In particular, this command

    git rebase -i HEAD topic

would always finish with "Nothing to do". (a1bf91e fixed the same issue for non-interactive rebase.)

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
---

I made this with -U5 so that you can see the checkout in the context.

BTW, methinks, this checkout is unnecessary, since before the rebase begins, there is a 'git checkout $ONTO', and the branch switching is certainly not needed to compute the todo list...

-- Hannes

 git-rebase--interactive.sh |    7 +++----
 1 files changed, 3 insertions(+), 4 deletions(-)

Show changes to git-rebase--interactive.sh +3 −4

diff --git a/git-rebase--interactive.sh b/git-rebase--interactive.sh
index 8ee08ff..0ca986f 100755
--- a/git-rebase--interactive.sh
+++ b/git-rebase--interactive.sh
@@ -473,25 +473,24 @@ do
 			;;
 		esac

 		require_clean_work_tree

+		UPSTREAM=$(git rev-parse --verify "$1") || die "Invalid base"
+		test -z "$ONTO" && ONTO=$UPSTREAM
+
 		if test ! -z "$2"
 		then
 			output git show-ref --verify --quiet "refs/heads/$2" ||
 				die "Invalid branchname: $2"
 			output git checkout "$2" ||
 				die "Could not checkout $2"
 		fi

 		HEAD=$(git rev-parse --verify HEAD) || die "No HEAD?"
-		UPSTREAM=$(git rev-parse --verify "$1") || die "Invalid base"
-
 		mkdir "$DOTEST" || die "Could not create temporary $DOTEST"

-		test -z "$ONTO" && ONTO=$UPSTREAM
-
 		: > "$DOTEST"/interactive || die "Could not mark as interactive"
 		git symbolic-ref HEAD > "$DOTEST"/head-name 2> /dev/null ||
 			echo "detached HEAD" > "$DOTEST"/head-name

 		echo $HEAD > "$DOTEST"/head

-- 
1.5.6.rc0.885.gdf17.dirty

Johannes Schindelin· Jun 2, 2008, 15:29 UTC · re: Johannes Sixt · lore

Re: [PATCH] rebase --interactive: Compute upstream SHA1 before switching branches

Hi,

On Mon, 2 Jun 2008, Johannes Sixt wrote:

Show 13 quoted lines

> From: Johannes Sixt <johannes.sixt@telecom.at>
> 
> If the upstream argument to rebase (the first argument) was relative to
> HEAD and the name of the branch to rebase (the second argument) was given,
> the upstream would have been interpreted relative to the second argument.
> In particular, this command
> 
>     git rebase -i HEAD topic
> 
> would always finish with "Nothing to do". (a1bf91e fixed the same issue
> for non-interactive rebase.)
> 
> Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>

Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>

Show 5 quoted lines

> I made this with -U5 so that you can see the checkout in the context.
> 
> BTW, methinks, this checkout is unnecessary, since before the rebase 
> begins, there is a 'git checkout $ONTO', and the branch switching is 
> certainly not needed to compute the todo list...

It is needed to determine which branch to update after finish.

Ciao, Dscho

Junio C Hamano· Jun 3, 2008, 02:28 UTC · re: Johannes Schindelin · lore

Re: [PATCH] rebase --interactive: Compute upstream SHA1 before switching branches

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Show 5 quoted lines

>> BTW, methinks, this checkout is unnecessary, since before the rebase 
>> begins, there is a 'git checkout $ONTO', and the branch switching is 
>> certainly not needed to compute the todo list...
>
> It is needed to determine which branch to update after finish.

I'll apply Hannes's patch for now as I do not want to leave an obvious fix out of -rc1.

I think Hannes is referring to the same issue as the one dealt with with 0cb0664 (rebase [--onto O] A B: omit needless checkout, 2008-03-15) for the non-interactive codepath.

You certainly "need to determine which branch to update after finish", and you do need to remember what branch you were on (if you were not called with the <branch> parameter), or what branch were given from the command line (if you got one as <branch> parameter) in order to do so, but the way to remember these values does not have to involve an extra checkout.

By doing an extra checkout of "the original HEAD", before switching to a detached HEAD state at the $ONTO commit to "clean the slate" to apply the sequence on top, more work tree entries will be smudged, forcinging more recompilations than necessary if the tracked contents are "source" files.

Johannes Schindelin· Jun 3, 2008, 13:55 UTC · re: Junio C Hamano · lore

Re: [PATCH] rebase --interactive: Compute upstream SHA1 before switching branches

Hi,

On Mon, 2 Jun 2008, Junio C Hamano wrote:

Show 27 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> >> BTW, methinks, this checkout is unnecessary, since before the rebase 
> >> begins, there is a 'git checkout $ONTO', and the branch switching is 
> >> certainly not needed to compute the todo list...
> >
> > It is needed to determine which branch to update after finish.
> 
> I'll apply Hannes's patch for now as I do not want to leave an obvious fix
> out of -rc1.
> 
> I think Hannes is referring to the same issue as the one dealt with with
> 0cb0664 (rebase [--onto O] A B: omit needless checkout, 2008-03-15) for
> the non-interactive codepath.
> 
> You certainly "need to determine which branch to update after finish", 
> and you do need to remember what branch you were on (if you were not 
> called with the <branch> parameter), or what branch were given from the 
> command line (if you got one as <branch> parameter) in order to do so, 
> but the way to remember these values does not have to involve an extra 
> checkout.
> 
> By doing an extra checkout of "the original HEAD", before switching to a 
> detached HEAD state at the $ONTO commit to "clean the slate" to apply 
> the sequence on top, more work tree entries will be smudged, forcinging 
> more recompilations than necessary if the tracked contents are "source" 
> files.

Right. I failed to follow the given hint to the other commit, but it makes perfect sense now.

Thanks, Dscho