contrib/subtree: make it respect spaces in a repository path

Some repositories may have spaces in their paths. Currently git-subtree raises an error in such cases.
Also, git-subtree currently does not have tests for its 'push' command. Following patches are to fix these statements.

Alexey Shumkin (2):
t7900-subtree: test the "space in a subdirectory name" case contrib/subtree: respect spaces in a repository path

 contrib/subtree/git-subtree.sh     |   4 +-
 contrib/subtree/t/t7900-subtree.sh | 194 +++++++++++++++++++++++--------------
 2 files changed, 124 insertions(+), 74 deletions(-)

2.4.1-21

In common case there can be spaces in a subdirectory name. Change tests accorgingly to this statement.

Also, as far as a call to the rejoin_msg function (in cmd_split) does not take into account such a case this patch fixes commit message when --rejoin option is set .

Besides, as fixnl and multiline functions did not take into account the "new" tested "space in a subdirectory name" case they become unused and redundant, so they are removed.

---
 contrib/subtree/git-subtree.sh     |   2 +-
 contrib/subtree/t/t7900-subtree.sh | 147 +++++++++++++++++++------------------
 2 files changed, 76 insertions(+), 73 deletions(-)

index 9f06571..72a20c0 100755
@@ -648,7 +648,7 @@ cmd_split()
 		debug "Merging split branch into HEAD..."
 		latest_old=$(cache_get latest_old)
 		git merge -s ours \
-			-m "$(rejoin_msg $dir $latest_old $latest_new)" \
+			-m "$(rejoin_msg "$dir" $latest_old $latest_new)" \
 			$latest_new >&2 || exit $?
 	fi
 	if [ -n "$branch" ]; then

2.4.1-21

Remote repository may have spaces in its path, so take it into account.

Also, as far as there are no tests for the push command, add them.

---
 contrib/subtree/git-subtree.sh     |  2 +-
 contrib/subtree/t/t7900-subtree.sh | 47 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)

index 72a20c0..308b777 100755
@@ -735,7 +735,7 @@ cmd_push()
 	    refspec=$2
 	    echo "git push using: " $repository $refspec
 	    localrev=$(git subtree split --prefix="$prefix") || die
-	    git push $repository $localrev:refs/heads/$refspec
+	    git push "$repository" $localrev:refs/heads/$refspec
 	else
 	    die "'$dir' must already exist. Try 'git subtree add'."
 	fi

2.4.1-21

Doesn't this order break bisection? It seems that you turn "subdir" to "sub dir" in existing tests, and I understand that the whole point of this series is that such a change will expose that the tool is broken, making tests fail.

Also, if you feel up to it, it might be a good idea to clean t7900 test up to the current best practice before doing any other changes as a pure preparatory clean-up patch.

Namely, using cd outside a subshell of the tests to move around is a bad thing to do, and you are adding more instance of it in this series. If one test with such a cd to go down fails before it has a chance to come back up (or go up and then fail to come back down), the later tests will be left in an unexpected place.

Thanks.

It seems I have to reword commit messages to avoid such an interpretation. Because, the first commit does not break anything. It is to change the tests for git-subtree "to show" that git-subtrees already tested commands (almost) work correctly if there are spaces in paths (except --rejoin-msg issue).
And the second commit adds missing tests and the fix.
Should I add/commit the breaking test first and then commit the fix?

I understand this issue with "cd" (I've just followed the existing t7900 tests "code style").

Alexey Shumkin
E-mail: [email protected]
ICQ: 118001447
Jabber (GoogleTalk): [email protected]
Skype: crezoff