t7400 broken on pu (Mac OS X)

DISCUSSION Torsten Bögershausen 3 participants 6 messages Jun 15, 2016

lore

Share & tools

lore

Torsten Bögershausen

author

2016-06-15 22:55 UTC #0 lore raw

The current pu fails on Mac OS, case insensitive FS.

Bisecting points out
commit 3f28e4fafc046284657945798d71c57608bee479
[snip]
Date: Sun Jan 6 13:21:07 2013 +0700

Convert add_files_to_cache to take struct pathspec

And I veryfied that the preceeding commit 05647d2d8a5dc456d1f4ef73 is OK.

It fails here:
not ok 38 - gracefully add submodule with a trailing slash

A run of a modified t7400 looks like this:
Is there anything more, that I can do to debug this?

[snip]
ok 37 - do not add files from a submodule

expecting success:

	git reset --hard &&
	echo 1 >&2 &&
	git commit -m "commit subproject" init &&
	echo 2 >&2 &&
	(cd init &&
	echo 3 >&2 &&
	 echo b > a) &&
	echo 4 >&2 &&
	git add init/ &&
	echo 5 >&2 &&
	git diff --exit-code --cached init &&
	echo 6 >&2 &&
	commit=$(cd init &&
	echo 7 >&2 &&
	 git commit -m update a >/dev/null &&
	echo 8 >&2 &&
	 git rev-parse HEAD) &&
	echo 9 >&2 &&
	git add init/ &&
	echo 10 >&2 &&
	test_must_fail git diff --exit-code --cached init &&
	echo 11 >&2 &&
	test $commit = $(git ls-files --stage |
		sed -n "s/^160000 \([^ ]*\).*/\1/p")

HEAD is now at 57f2622 super commit 1
1
[second 1b8c63f] commit subproject
Author: A U Thor <[email protected]>

 1 file changed, 1 insertion(+), 1 deletion(-)

2
3
4
5
6
7
8
9
10
test_must_fail: command succeeded: git diff --exit-code --cached init not ok 38 - gracefully add submodule with a trailing slash

Junio C Hamano

2016-06-15 22:55 UTC #2 lore raw

Quoted 9 lines · Duy Nguyen

Quoted 8 lines · Torsten Bögershausen
The current pu fails on Mac OS, case insensitive FS.
Bisecting points out
commit 3f28e4fafc046284657945798d71c57608bee479
[snip]
Date: Sun Jan 6 13:21:07 2013 +0700

Quoted 6 lines

Quoted 2 lines
Convert add_files_to_cache to take struct pathspec
I can reproduce it by setting core.ignorecase to true. There is a bug that I overlooked. Can you verify if this throw-away patch fixes it for you? A proper fix will be in the reroll later.

I can see why it is wrong to let pathspec.raw be rewritten without making matching change to the containing pathspec, but I find it strange why it matters only on case-insensitive codepath.

I agree with the "Hack" comment that the canonicalization should be done at a higher level upfront. Then ls-files does not need its own strip_trailing_slash_from_submodules(), and check_path_for_gitlink() can (and should---the callers of "check_anything" would not expect the function to change things) stop rewriting its parameter.

Thanks for a quick response.

Quoted 60 lines

-- 8< --

builtin/add.c +17 -6

index 641037f..61cb8bd 100644
@@ -155,12 +155,13 @@ static char *prune_directory(struct dir_struct *dir, const char **pathspec, int
 	return seen;
 }
 
-static void treat_gitlinks(const char **pathspec)
+static int treat_gitlinks(const char **pathspec)
 {
 	int i;
+	int modified = 0;
 
 	if (!pathspec || !*pathspec)
-		return;
+		return modified;
 
 	for (i = 0; i < active_nr; i++) {
 		struct cache_entry *ce = active_cache[i];

@@ -171,15 +172,17 @@ static void treat_gitlinks(const char **pathspec)
 				if (len2 <= len || pathspec[j][len] != '/' ||
 				    memcmp(ce->name, pathspec[j], len))
 					continue;
-				if (len2 == len + 1)
+				if (len2 == len + 1) {
 					/* strip trailing slash */
 					pathspec[j] = xstrndup(ce->name, len);
-				else
+					modified = 1;
+				} else
 					die (_("Path '%s' is in submodule '%.*s'"),
 						pathspec[j], len, ce->name);
 			}
 		}
 	}
+	return modified;
 }
 
 static void refresh(int verbose, const struct pathspec *pathspec)

@@ -418,7 +421,16 @@ int cmd_add(int argc, const char **argv, const char *prefix)
 
 	if (read_cache() < 0)
 		die(_("index file corrupt"));
-	treat_gitlinks(pathspec.raw);
+	if (treat_gitlinks(pathspec.raw))
+		/*
+		 * HACK: treat_gitlinks strips the trailing slashes
+		 * out of submodule entries but it only affects
+		 * raw[]. Everything in pathspec.items is not touched.
+		 * Re-init it to propagate the change. Long term, this
+		 * function should be moved to pathspec.c and update
+		 * everything in a consistent way.
+		 */
+		init_pathspec(&pathspec, pathspec.raw);
 
 	if (add_new_files) {
 		int baselen;
-- 8< --

Torsten Bögershausen

author

2016-06-15 22:55 UTC #3 lore raw

Quoted 20 lines · Junio C Hamano

Quoted 9 lines · Duy Nguyen
Quoted 8 lines · Torsten Bögershausen
The current pu fails on Mac OS, case insensitive FS.
Bisecting points out
commit 3f28e4fafc046284657945798d71c57608bee479
[snip]
Date: Sun Jan 6 13:21:07 2013 +0700
Quoted 5 lines
Quoted 2 lines
Convert add_files_to_cache to take struct pathspec
I can reproduce it by setting core.ignorecase to true. There is a bug that I overlooked. Can you verify if this throw-away patch fixes it for you? A proper fix will be in the reroll later.
I can see why it is wrong to let pathspec.raw be rewritten without making matching change to the containing pathspec, but I find it strange why it matters only on case-insensitive codepath.

Quoted 8 lines

I agree with the "Hack" comment that the canonicalization should be done at a higher level upfront. Then ls-files does not need its own strip_trailing_slash_from_submodules(), and check_path_for_gitlink() can (and should---the callers of "check_anything" would not expect the function to change things) stop rewriting its parameter.
Thanks for a quick response.

The patch fixes t7400.
Thanks from my side as well
/Torsten

Duy Nguyen

2016-06-15 22:55 UTC #4 lore raw

In reply to Junio C Hamano

Quoted 3 lines · Junio C Hamano

I can see why it is wrong to let pathspec.raw be rewritten without making matching change to the containing pathspec, but I find it strange why it matters only on case-insensitive codepath.

Yeah, I don't get it either. I can see that core.ignorecase exercises some more code, but still fail to see the link. I should get to the bottom of this and write some tests to for core.ignorecase-only code.

Signature · 1 line

Duy

Junio C Hamano

2016-06-15 22:55 UTC #5 lore raw

In reply to Torsten Bögershausen

Quoted 7 lines · Torsten Bögershausen

The current pu fails on Mac OS, case insensitive FS.
Bisecting points out
commit 3f28e4fafc046284657945798d71c57608bee479
[snip]
Date: Sun Jan 6 13:21:07 2013 +0700

Next time do not [snip] but please find the author address there, and Cc such a report.

I think this topic is planned to be rerolled anyway, and your report would be a valuable input while doing so.

Thanks.