stash: fix "--staged" with binary files

Commit Message

Adam Johnson April 22, 2024, 10:28 a.m. UTC

From: Adam Johnson <[e-mail]>

"git stash --staged" would crash with binary files, after saving the stash.
This behaviour dates back to the addition of the feature in 41a28eb6c1
(stash: implement '--staged' option for 'push' and 'save', 2021-10-18).
Adding the "--binary" option of "diff-tree" fixes this. The "diff-tree" call
in stash_patch() also omits "--binary", but that is fine since binary files
cannot be selected interactively.

Helped-By: Jeff King <[e-mail]>
Helped-By: Randall S. Becker <[e-mail]>
Signed-off-by: Adam Johnson <[e-mail]>
---
    stash: fix "--staged" with binary files

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1722%2Fadamchainz%2Faj%2Fstash-binary-fix-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1722/adamchainz/aj/stash-binary-fix-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/1722

 builtin/stash.c  | 4 ++--
 t/t3903-stash.sh | 9 +++++++++
 2 files changed, 11 insertions(+), 2 deletions(-)


base-commit: ae3196a5ea84a9e88991d576020cf66512487088

Comments

Junio C Hamano April 22, 2024, 7 p.m. UTC | #1

"Adam Johnson via GitGitGadget" <[e-mail]> writes:

> From: Adam Johnson <[e-mail]>
>
> "git stash --staged" would crash with binary files, after saving the stash.

"would crash with" -> "errors out when given", probably.

We would be better off reserving the word "crash" to refer to an
uncontrolled exit with non-zero status of the command, possibly
killed via system signals like SEGV and SIGBUS.  In this case, I
suspect that you would get an "error:" from underlying apply
machinery, which leads the caller to exit with a non-zero status
due to an error.

> This behaviour dates back to the addition of the feature in 41a28eb6c1
> (stash: implement '--staged' option for 'push' and 'save', 2021-10-18).

When you find the commit that introduced the problem you are fixing,
especially when the author of the commit is still active on the
mailing list, it would be nice to give a carbon-copy of the message
to them.

> Adding the "--binary" option of "diff-tree" fixes this. The "diff-tree" call
> in stash_patch() also omits "--binary", but that is fine since binary files
> cannot be selected interactively.

I love seeing an explanation like this in a proposed commit log
message.  It is concise but very clear at the same time.  It is so
clearly written that anybody who is reasonably familiar with the
code does not even need to look at the patch text itself to see what
the actual fix would look like.

Very nicely explained.

> Helped-By: Jeff King <[e-mail]>
> Helped-By: Randall S. Becker <[e-mail]>
> Signed-off-by: Adam Johnson <[e-mail]>
> ---

Will queue.

Thanks.

> diff --git a/builtin/stash.c b/builtin/stash.c
> index 062be1fbc07..7751bca868e 100644
> --- a/builtin/stash.c
> +++ b/builtin/stash.c
> @@ -1205,8 +1205,8 @@ static int stash_staged(struct stash_info *info, struct strbuf *out_patch,
>  	}
>  
>  	cp_diff_tree.git_cmd = 1;
> -	strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "-U1", "HEAD",
> -		     oid_to_hex(&info->w_tree), "--", NULL);
> +	strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "--binary",
> +		     "-U1", "HEAD", oid_to_hex(&info->w_tree), "--", NULL);
>  	if (pipe_command(&cp_diff_tree, NULL, 0, out_patch, 0, NULL, 0)) {
>  		ret = -1;
>  		goto done;
> diff --git a/t/t3903-stash.sh b/t/t3903-stash.sh
> index 00db82fb245..a7f71f8126f 100755
> --- a/t/t3903-stash.sh
> +++ b/t/t3903-stash.sh
> @@ -393,6 +393,15 @@ test_expect_success 'stash --staged' '
>  	test bar,bar4 = $(cat file),$(cat file2)
>  '
>  
> +test_expect_success 'stash --staged with binary file' '
> +	printf "\0" >file &&
> +	git add file &&
> +	git stash --staged &&
> +	git stash pop &&
> +	printf "\0" >expect &&
> +	test_cmp expect file
> +'

Adam Johnson April 22, 2024, 8:48 p.m. UTC | #2

Thank you for the feedback, that was clearly explained. I will try to remember to CC more relevant folks in the future.

On Mon, 22 Apr 2024, at 20:00, Junio C Hamano wrote:
> "Adam Johnson via GitGitGadget" <[e-mail]> writes:
> 
> > From: Adam Johnson <[e-mail]>
> >
> > "git stash --staged" would crash with binary files, after saving the stash.
> 
> "would crash with" -> "errors out when given", probably.
> 
> We would be better off reserving the word "crash" to refer to an
> uncontrolled exit with non-zero status of the command, possibly
> killed via system signals like SEGV and SIGBUS.  In this case, I
> suspect that you would get an "error:" from underlying apply
> machinery, which leads the caller to exit with a non-zero status
> due to an error.
> 
> > This behaviour dates back to the addition of the feature in 41a28eb6c1
> > (stash: implement '--staged' option for 'push' and 'save', 2021-10-18).
> 
> When you find the commit that introduced the problem you are fixing,
> especially when the author of the commit is still active on the
> mailing list, it would be nice to give a carbon-copy of the message
> to them.
> 
> > Adding the "--binary" option of "diff-tree" fixes this. The "diff-tree" call
> > in stash_patch() also omits "--binary", but that is fine since binary files
> > cannot be selected interactively.
> 
> I love seeing an explanation like this in a proposed commit log
> message.  It is concise but very clear at the same time.  It is so
> clearly written that anybody who is reasonably familiar with the
> code does not even need to look at the patch text itself to see what
> the actual fix would look like.
> 
> Very nicely explained.
> 
> > Helped-By: Jeff King <[e-mail]>
> > Helped-By: Randall S. Becker <[e-mail]>
> > Signed-off-by: Adam Johnson <[e-mail]>
> > ---
> 
> Will queue.
> 
> Thanks.
> 
> > diff --git a/builtin/stash.c b/builtin/stash.c
> > index 062be1fbc07..7751bca868e 100644
> > --- a/builtin/stash.c
> > +++ b/builtin/stash.c
> > @@ -1205,8 +1205,8 @@ static int stash_staged(struct stash_info *info, struct strbuf *out_patch,
> >  }
> >  
> >  cp_diff_tree.git_cmd = 1;
> > - strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "-U1", "HEAD",
> > -      oid_to_hex(&info->w_tree), "--", NULL);
> > + strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "--binary",
> > +      "-U1", "HEAD", oid_to_hex(&info->w_tree), "--", NULL);
> >  if (pipe_command(&cp_diff_tree, NULL, 0, out_patch, 0, NULL, 0)) {
> >  ret = -1;
> >  goto done;
> > diff --git a/t/t3903-stash.sh b/t/t3903-stash.sh
> > index 00db82fb245..a7f71f8126f 100755
> > --- a/t/t3903-stash.sh
> > +++ b/t/t3903-stash.sh
> > @@ -393,6 +393,15 @@ test_expect_success 'stash --staged' '
> >  test bar,bar4 = $(cat file),$(cat file2)
> >  '
> >  
> > +test_expect_success 'stash --staged with binary file' '
> > + printf "\0" >file &&
> > + git add file &&
> > + git stash --staged &&
> > + git stash pop &&
> > + printf "\0" >expect &&
> > + test_cmp expect file
> > +'
> 
>

Jeff King April 23, 2024, 9:44 p.m. UTC | #3

On Mon, Apr 22, 2024 at 10:28:14AM +0000, Adam Johnson via GitGitGadget wrote:

> From: Adam Johnson <[e-mail]>
> 
> "git stash --staged" would crash with binary files, after saving the stash.
> This behaviour dates back to the addition of the feature in 41a28eb6c1
> (stash: implement '--staged' option for 'push' and 'save', 2021-10-18).
> Adding the "--binary" option of "diff-tree" fixes this. The "diff-tree" call
> in stash_patch() also omits "--binary", but that is fine since binary files
> cannot be selected interactively.
> 
> Helped-By: Jeff King <[e-mail]>
> Helped-By: Randall S. Becker <[e-mail]>
> Signed-off-by: Adam Johnson <[e-mail]>

I had to dig in the archive to remember what I might have helped with. ;)

The patch looks good to me, though one thing I noticed:

> +test_expect_success 'stash --staged with binary file' '
> +	printf "\0" >file &&
> +	git add file &&
> +	git stash --staged &&
> +	git stash pop &&
> +	printf "\0" >expect &&
> +	test_cmp expect file
> +'

I wonder if test_cmp would ever have problems with binary files on any
platforms (and yes, I was the one who suggested using it). We use "diff"
on most platforms, but a few (like old SunOS) set GIT_TEST_CMP to "cmp"
because they don't have "diff -u" at all. I'm content to leave it as-is
until somebody turns up a platform that complains, and then the escape
hatch is "OK, so set GIT_TEST_CMP=cmp".

-Peff

Patch

diff --git a/builtin/stash.c b/builtin/stash.c
index 062be1fbc07..7751bca868e 100644
--- a/builtin/stash.c
+++ b/builtin/stash.c
@@ -1205,8 +1205,8 @@  static int stash_staged(struct stash_info *info, struct strbuf *out_patch,
 	}
 
 	cp_diff_tree.git_cmd = 1;
-	strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "-U1", "HEAD",
-		     oid_to_hex(&info->w_tree), "--", NULL);
+	strvec_pushl(&cp_diff_tree.args, "diff-tree", "-p", "--binary",
+		     "-U1", "HEAD", oid_to_hex(&info->w_tree), "--", NULL);
 	if (pipe_command(&cp_diff_tree, NULL, 0, out_patch, 0, NULL, 0)) {
 		ret = -1;
 		goto done;
diff --git a/t/t3903-stash.sh b/t/t3903-stash.sh
index 00db82fb245..a7f71f8126f 100755
--- a/t/t3903-stash.sh
+++ b/t/t3903-stash.sh
@@ -393,6 +393,15 @@  test_expect_success 'stash --staged' '
 	test bar,bar4 = $(cat file),$(cat file2)
 '
 
+test_expect_success 'stash --staged with binary file' '
+	printf "\0" >file &&
+	git add file &&
+	git stash --staged &&
+	git stash pop &&
+	printf "\0" >expect &&
+	test_cmp expect file
+'
+
 test_expect_success 'dont assume push with non-option args' '
 	test_must_fail git stash -q drop 2>err &&
 	test_grep -e "subcommand wasn'\''t specified; '\''push'\'' can'\''t be assumed due to unexpected token '\''drop'\''" err