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

[PATCH] apply: when -R, also reverse list of sections

From
JTJonathan Tan <jonathantanmy@google.com>
Date
Sep 28, 2020, 21:20 UTC
Message-ID
<20200928212038.1625698-1-jonathantanmy@google.com>

A patch changing a symlink into a file is written with 2 sections (in the code, represented as "struct patch"): firstly, the deletion of the symlink, and secondly, the creation of the file. When applying that patch with -R, the sections are reversed, so we get:

 (1) creation of a symlink, then
 (2) deletion of a file.

This causes an issue when the "deletion of a file" section is checked, because Git observes that the so-called file is not a file but a symlink, resulting in a "wrong type" error message.

What we want is:
 (1) deletion of a file, then
 (2) creation of a symlink.

In the code, this is reflected in the behavior of previous_patch() when invoked from check_preimage() when the deletion is checked. Creation then deletion means that when the deletion is checked, previous_patch() returns the creation section, triggering a mode conflict resulting in the "wrong type" error message. But deletion then creation means that when the deletion is checked, previous_patch() returns NULL, so the deletion mode is checked against lstat, which is what we want.

Therefore, when building the list of sections, build them in reverse order (by adding to the front of the list instead of the back) when -R is passed.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
---
I traced the different behavior in previous_patch() to what the
invocation of to_be_deleted() returns, but I couldn't figure out why it
returns false in the current-but-wrong-order case but true in the
future-and-correct-order case. I see that prepare_fn_table() sets
PATH_TO_BE_DELETED for deletions, but couldn't figure out where and when
it is set to something else. Further compounding my confusion,
conceptually, at the point of checking the deletion, both patches are
(in theory) "to be deleted". Any help in this is appreciated.
---
 apply.c                     | 9 +++++++--
 t/t4114-apply-typechange.sh | 7 +++++++
 2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/apply.c b/apply.c
index 76dba93c97..359ceb632c 100644
--- a/apply.c
+++ b/apply.c
@@ -4699,8 +4699,13 @@ static int apply_patch(struct apply_state *state,
 			reverse_patches(patch);
 		if (use_patch(state, patch)) {
 			patch_stats(state, patch);
-			*listp = patch;
-			listp = &patch->next;
+			if (!list || !state->apply_in_reverse) {
+				*listp = patch;
+				listp = &patch->next;
+			} else {
+				patch->next = list;
+				list = patch;
+			}
 
 			if ((patch->new_name &&
 			     ends_with_path_components(patch->new_name,
diff --git a/t/t4114-apply-typechange.sh b/t/t4114-apply-typechange.sh
index ebadbc347f..da3e64f811 100755
--- a/t/t4114-apply-typechange.sh
+++ b/t/t4114-apply-typechange.sh
@@ -88,6 +88,13 @@ test_expect_success 'symlink becomes file' '
 	'
 test_debug 'cat patch'
 
+test_expect_success 'symlink becomes file, in reverse' '
+	git checkout -f foo-symlinked-to-bar &&
+	git diff-tree -p HEAD foo-back-to-file > patch &&
+	git checkout foo-back-to-file &&
+	git apply -R --index < patch
+	'
+
 test_expect_success 'binary file becomes symlink' '
 	git checkout -f foo-becomes-binary &&
 	git diff-tree -p --binary HEAD foo-symlinked-to-bar > patch &&
-- 
2.28.0.709.gb0816b6eb0-goog
Next: Junio C Hamano
Message 1 of 8 in “apply: when -R, also reverse list of sections”
  1. apply: when -R, also reverse list of sectionsJonathan Tan, Sep 28, 2020
  2. Junio C HamanoSep 28, 2020
  3. Jonathan TanOct 20, 2020
  4. Junio C HamanoOct 20, 2020
  5. Junio C HamanoOct 20, 2020
  6. Jonathan TanOct 20, 2020
  7. Junio C HamanoOct 20, 2020
  8. apply: when -R, also reverse list of sectionsJonathan Tan, Oct 20, 2020

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.