Skip to contentSkip to conversation

Allow usage of --gpg-sign flag in commit-tree builtin.

PATCH In review Brandon Richardson 2 participants 2 messages Jan 17, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Allow usage of --gpg-sign flag in commit-tree builtin.

---
Hi,

This is my first contribution, so please bear with me. All feedback is appreciated.

Ran into this issue while writing a signed commit object manually. Here are the steps I followed to replicate the issue:

mkdir test && cd test
git init
echo 'test' > test.txt
git hash-object -w test.txt
git update-index --add --cacheinfo 100644 <blob hash> test.txt git write-tree
git commit-tree --gpg-sign -m 'test commit msg' <tree obj hash>

Thanks,
Brandon

 builtin/commit-tree.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
builtin/commit-tree.c +3 -1
index 9ec36a82b..9a06594f6 100644
@@ -66,7 +66,9 @@ int cmd_commit_tree(int argc, const char **argv, const char *prefix)
 			continue;
 		}
 
-		if (skip_prefix(arg, "-S", &sign_commit))
+		if (skip_prefix(arg, "-S", &sign_commit) ||
+			skip_prefix(arg, "--gpg-sign=", &sign_commit) ||
+			skip_prefix(arg, "--gpg-sign", &sign_commit))
 			continue;
 
 		if (!strcmp(arg, "--no-gpg-sign")) {
Signature · 1 line
2.20.1
#1 lore raw

Hi Brandon,

Welcome to the list! :-)

Quoted 1 line · Brandon Richardson

Subject: [PATCH] Allow usage of --gpg-sign flag in commit-tree builtin.

Good find!

Maybe "commit-tree: handle --gpg-sign" which looks more like what we see in git log --oneline builtin/commit-tree.c.

Quoted 1 line

It could be worthwhile to note in the proposed commit message that this option is actually documented, since 55ca3f99ae ("commit-tree: add and document --no-gpg-sign", 2013-12-13), so it's clearly an omission that it's not implemented.

Quoted 9 lines

Here are the steps I followed to replicate the issue:

mkdir test && cd test
git init
echo 'test' > test.txt
git hash-object -w test.txt
git update-index --add --cacheinfo 100644 <blob hash> test.txt git write-tree
git commit-tree --gpg-sign -m 'test commit msg' <tree obj hash>

Have you looked at turning this into a test in the t/ directory? Grepping a little suggests that t7510-signed-commit.sh might be a good spot. A test would make the patch more obviously correct, and would make sure that this doesn't regress in the future.

builtin/commit-tree.c int cmd_commit_tree(int argc, const char **argv, const char *prefix) · line 72

Quoted 5 lines
builtin/commit-tree.c +3 -1
-               if (skip_prefix(arg, "-S", &sign_commit))
+               if (skip_prefix(arg, "-S", &sign_commit) ||
+                       skip_prefix(arg, "--gpg-sign=", &sign_commit) ||
+                       skip_prefix(arg, "--gpg-sign", &sign_commit))
                        continue;

This would match "--gpg-signfoo", which seems like a bug. The first two skip_prefix() calls look ok. Then maybe

  if (!strcmp(arg, "--gpg-sign")) {
          sign_commit = "";
          continue;
  }

That's entirely untested, though.

Thanks for looking into this, rather than just side-stepping it. ;-)

Martin