git-checkout and SUBDIRECTORY_OK='Yes'

Jon Nelson· Dec 23, 2005, 04:20 UTC · lore

Is it safe to set SUBDIRECTORY_OK='Yes' in git-checkout.sh? What about the other *.sh that don't already set it?

-- Jon Nelson <jnelson-git@jamponi.net>

Junio C Hamano· Dec 23, 2005, 04:58 UTC · re: Jon Nelson · lore

Re: git-checkout and SUBDIRECTORY_OK='Yes'

Jon Nelson <jnelson-git@jamponi.net> writes:

> Is it safe to set SUBDIRECTORY_OK='Yes' in git-checkout.sh?

Personally, I do not like this kind of question. If it were known to be safe we would have set it so already.

Ah, I must be tired.  Sorry.  Let's try again.

git-checkout does two very different things, and what they should do when run from subdirectory are probably quite different.

It does not make any sense to run the one that switches the current head from anywhere other than the toplevel:

	git-checkout [-f] <branch>
        git-checkout [-b <branch>] <committish>

We could of course chdir to top and do the whole-tree checkout in git-checkout, but the point is the operation does not make sense on a partial tree.

The other form is to update the index file and working tree file selectively:

	git-checkout <treeish> <file>... ;# out of tree to index and file
        git-checkout -- <file>...	 ;# out of index to file

This form _does_ make sense to run from subdirectory; and I myself often wish we supported this.

Before SUBDIRECTORY_OK was introduced, git-*.sh commands were carefully inspected and the ones that do not make much sense to be run in subdirectories were marked with sh-setup. So if you do not see SUBDIRECTORY_OK, it is very likely that you need some real work (like the analysis above to think what should happen when run from a subdirectory, and actually coding it).

This patch *might* work, or it may not.  Please let us know.

diff --git a/git-checkout.sh b/git-checkout.sh
index 36308d2..1219ea0 100755
--- a/git-checkout.sh
+++ b/git-checkout.sh
@@ -1,6 +1,7 @@
 #!/bin/sh
 
 USAGE='[-f] [-b <new_branch>] [<branch>] [<paths>...]'
+SUBDIRECTORY_OK=Sometimes
 . git-sh-setup
 
 old=$(git-rev-parse HEAD)
@@ -95,6 +96,14 @@ else
 	fi
 fi
 
+# We are switching branches and checking out trees, so
+# we *NEED* to be at the toplevel.
+cdup=$(git-rev-parse --show-cdup)
+if test ! -z "$cdup"
+then
+	cd "$cdup"
+fi
+
 [ -z "$new" ] && new=$old
 
 # If we don't have an old branch that we're switching to,
diff --git a/rev-parse.c b/rev-parse.c
index bb4949a..0c951af 100644
--- a/rev-parse.c
+++ b/rev-parse.c
@@ -216,6 +216,18 @@ int main(int argc, char **argv)
 					puts(prefix);
 				continue;
 			}
+			if (!strcmp(arg, "--show-cdup")) {
+				const char *pfx = prefix;
+				while (pfx) {
+					pfx = strchr(pfx, '/');
+					if (pfx) {
+						pfx++;
+						printf("../");
+					}
+				}
+				putchar('\n');
+				continue;
+			}
 			if (!strcmp(arg, "--git-dir")) {
 				const char *gitdir = getenv(GIT_DIR_ENVIRONMENT);
 				static char cwd[PATH_MAX];

Junio C Hamano· Dec 23, 2005, 17:56 UTC · re: Junio C Hamano · lore

Re: git-checkout and SUBDIRECTORY_OK='Yes'

Junio C Hamano <junkio@cox.net> writes:

> Jon Nelson <jnelson-git@jamponi.net> writes:
>
>> Is it safe to set SUBDIRECTORY_OK='Yes' in git-checkout.sh?

> This patch *might* work, or it may not...

I've done some tests and it seems to work (it is in the proposed updates branch).

In git.git working tree, here is what happens:

	$ git checkout pu			;# start from pu
        $ cd Documentation
        $ echo >>git-daemon.txt			;# smudge
        $ git checkout -- git-daemon.txt	;# restore
	$ git diff -r --name-status		;# no change
        $ git checkout master git-daemon.txt	;# pick from tree
	$ git diff master -r --name-status	;# (1)
	$ git diff -r --name-status		;# (2)
	$ git diff pu -r --name-status          ;# (3)
        M	Documentation/git-daemon.txt

    (1) does not show any because it did take the file out of master
        branch, and git-daemon.txt is the only file different under
        Documentation/ between master and pu.

    (2) does not show any because "checkout master git-daemon.txt"
        updated the path both in index and working tree.

    (3) shows the path indeed is different between master and pu.

The change relies on ls-tree showing full paths regardless of where it gets started (because the output is fed to --index-info form of git-update-index). If we are going to take the "ls-tree prefix truncation" patch from Linus, either we also need to update "git-update-index --index-info" to match this.