git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[RFC/PATCH] add-patch: handle splitting hunks with diff.suppressBlankEmpty

From
Jeff King <peff@peff.net>
Date
Jul 10, 2024, 09:36 UTC
Message-ID
<20240710093610.GA2076910@coredump.intra.peff.net>
In-Reply-To
<xmqq4j937pyf.fsf@gitster.g>
On Fri, Jul 05, 2024 at 09:39:52AM -0700, Junio C Hamano wrote:
> As to the "commit -p" issue, I think the patch parser is in the
> wrong and needs to be corrected, period.  As long as the patches
> given as input are well-formed, we should be prepared to grok
> them (we even allow manual editing of patches, right?).
Maybe this?
-- >8 --
Subject: add-patch: handle splitting hunks with diff.suppressBlankEmpty

When "add -p" parses diffs, it looks for context lines starting with a single space. But when diff.suppressBlankEmpty is in effect, an empty context line will omit the space, giving us a true empty line. This confuses the parser, which is unable to split based on such a line.

It's tempting to say that we should just make sure that we generate a diff without that option. But we may parse diffs not only generated by Git, but ones that users have manually edited. And POSIX calls the decision of whether to print the space here "implementation-defined".

So let's handle both cases: a context line either starts with a space or consists of a totally empty line.

Reported-by: Ilya Tumaykin <itumaykin@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
I'm a little worried that this creates ambiguities, since I don't think
we are careful about following the hunk header's line counts. Imagine
you had an input like this:
  @@ -1,2 +1,2 @@
  -old
  +new
   stuff
  some garbage

We obviously know that "some garbage" is not a context line and is just trailing junk, because it does not begin with "-", "+" or space. But what about the blank line in between? It looks like an empty context line, but we can only know that it is not by respecting the counts in the hunk header.

I don't think we'd ever generate this ourselves, but could somebody manually edit a hunk into this shape? When I tried it in practice, it looks like we fail to apply the result even before my patch, though. I'm not sure why that is. If I put "some garbage" without the blank line, we correctly realize it should be discarded. It's possible I'm just holding it wrong.

 add-patch.c                |  8 ++++----
 t/t3701-add-interactive.sh | 32 ++++++++++++++++++++++++++++++++
 2 files changed, 36 insertions(+), 4 deletions(-)
diff --git a/add-patch.c b/add-patch.c
index 6e176cd21a..7beead1d0a 100644
--- a/add-patch.c
+++ b/add-patch.c
@@ -588,7 +588,7 @@ static int parse_diff(struct add_p_state *s, const struct pathspec *ps)
 			    (int)(eol - (plain->buf + file_diff->head.start)),
 			    plain->buf + file_diff->head.start);
 
-		if ((marker == '-' || marker == '+') && *p == ' ')
+		if ((marker == '-' || marker == '+') && (*p == ' ' || *p == '\n'))
 			hunk->splittable_into++;
 		if (marker && *p != '\\')
 			marker = *p;
@@ -964,7 +964,7 @@ static int split_hunk(struct add_p_state *s, struct file_diff *file_diff,
 		 * Is this the first context line after a chain of +/- lines?
 		 * Then record the start of the next split hunk.
 		 */
-		if ((marker == '-' || marker == '+') && ch == ' ') {
+		if ((marker == '-' || marker == '+') && (ch == ' ' || ch == '\n')) {
 			first = 0;
 			hunk[1].start = current;
 			if (colored)
@@ -979,14 +979,14 @@ static int split_hunk(struct add_p_state *s, struct file_diff *file_diff,
 		 * Then just increment the appropriate counter and continue
 		 * with the next line.
 		 */
-		if (marker != ' ' || (ch != '-' && ch != '+')) {
+		if ((marker != ' ' && marker != '\n') || (ch != '-' && ch != '+')) {
 next_hunk_line:
 			/* Comment lines are attached to the previous line */
 			if (ch == '\\')
 				ch = marker ? marker : ' ';
 
 			/* current hunk not done yet */
-			if (ch == ' ')
+			if (ch == ' ' || ch == '\n')
 				context_line_count++;
 			else if (ch == '-')
 				header->old_count++;
diff --git a/t/t3701-add-interactive.sh b/t/t3701-add-interactive.sh
index 5d78868ac1..92c8e6dc8c 100755
--- a/t/t3701-add-interactive.sh
+++ b/t/t3701-add-interactive.sh
@@ -1164,4 +1164,36 @@ test_expect_success 'reset -p with unmerged files' '
 	test_must_be_empty staged
 '
 
+test_expect_success 'splitting handles diff.suppressBlankEmpty' '
+	test_when_finished "git reset --hard" &&
+	cat >file <<-\EOF &&
+	1
+	2
+
+	3
+	4
+	EOF
+	git add file &&
+
+	cat >file <<-\EOF &&
+	one
+	two
+
+	three
+	four
+	EOF
+	test_write_lines s n y |
+	git -c diff.suppressBlankEmpty=true add -p &&
+
+	git cat-file blob :file >actual &&
+	cat >expect <<-\EOF &&
+	1
+	2
+
+	three
+	four
+	EOF
+	test_cmp expect actual
+'
+
 test_done
-- 
2.45.2.1249.gb036353db5
Previous: Junio C HamanoNext: Phillip Wood
Message 4 of 12 in “git crashes in `git commit --patch` with diff.suppressBlankEmpty = true”
  1. Ilya TumaykinJul 3, 2024
  2. Phillip WoodJul 4, 2024
  3. Junio C HamanoJul 5, 2024
  4. add-patch: handle splitting hunks with diff.suppressBlankEmptyJeff King, Jul 10, 2024
  5. Phillip WoodJul 10, 2024
  6. Jeff KingJul 11, 2024
  7. phillip.wood123@gmail.comJul 13, 2024
  8. Jeff KingJul 13, 2024
  9. Phillip WoodJul 18, 2024
  10. Phillip WoodJul 18, 2024
  11. Junio C HamanoJul 10, 2024
  12. Jeff KingJul 11, 2024

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.