[PATCH] unpack-trees: do not abort when overwriting an existing file with the same content

["Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 The "git reset" as illustrated in the test case is one case people
 may be hit by this. We can also do the same for "uptodate" check, but
 I'm not sure how that could happen.

 t/t1011-read-tree-sparse-checkout.sh |  3 ++-
 t/t2021-checkout-overwrite.sh        |  8 ++++++++
 unpack-trees.c                       | 27 +++++++++++++++++++++++++++
 3 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/t/t1011-read-tree-sparse-checkout.sh b/t/t1011-read-tree-sparse-checkout.sh
index 5c0053a..38f9899 100755
--- a/t/t1011-read-tree-sparse-checkout.sh
+++ b/t/t1011-read-tree-sparse-checkout.sh
@@ -238,7 +238,8 @@ test_expect_success 'print errors when failed to update worktree' '
 	echo sub >.git/info/sparse-checkout &&
 	git checkout -f init &&
 	mkdir sub &&
-	touch sub/added sub/addedtoo &&
+	echo modified >sub/added &&
+	echo modified >sub/addedtoo &&
 	test_must_fail git checkout top 2>actual &&
 	cat >expected <<\EOF &&
 error: The following untracked working tree files would be overwritten by checkout:
diff --git a/t/t2021-checkout-overwrite.sh b/t/t2021-checkout-overwrite.sh
index 5da63e9..4163449 100755
--- a/t/t2021-checkout-overwrite.sh
+++ b/t/t2021-checkout-overwrite.sh
@@ -47,4 +47,12 @@ test_expect_success SYMLINKS 'checkout commit with dir must not remove untracked
 	test -h a/b
 '
 
+test_expect_success 'do not abort on overwriting an existing file with the same content' '
+	echo abc >bar &&
+	git add bar &&
+	git commit -m "new file" &&
+	git reset HEAD^ &&
+	git checkout HEAD@{1}
+'
+
 test_done
diff --git a/unpack-trees.c b/unpack-trees.c
index 0e1a196..16adc03 100644
--- a/unpack-trees.c
+++ b/unpack-trees.c
@@ -1363,6 +1363,7 @@ static int check_ok_to_remove(const char *name, int len, int dtype,
 			      struct unpack_trees_options *o)
 {
 	struct cache_entry *result;
+	unsigned long ce_size;
 
 	/*
 	 * It may be that the 'lstat()' succeeded even though
@@ -1405,6 +1406,32 @@ static int check_ok_to_remove(const char *name, int len, int dtype,
 			return 0;
 	}
 
+	/*
+	 * If it has the same content that we are going to write down,
+	 * there's no point in complaining. We still overwrite it in the
+	 * end though. Permission is not checked so it may be lost.
+	 */
+	if (ce &&
+	    S_ISREG(st->st_mode) && S_ISREG(ce->ce_mode) &&
+	    st->st_size < 1024 * 1024 && /* should be configurable */
+	    sha1_object_info(ce->sha1, &ce_size) == OBJ_BLOB &&
+	    ce_size == st->st_size) {
+		void *buffer = NULL;
+		unsigned long size;
+		enum object_type type;
+		struct strbuf sb = STRBUF_INIT;
+		int matched =
+			strbuf_read_file(&sb, ce->name, ce_size) == ce_size &&
+			(buffer = read_sha1_file(ce->sha1, &type, &size)) != NULL &&
+			type == OBJ_BLOB &&
+			size == ce_size &&
+			!memcmp(buffer, sb.buf, size);
+		free(buffer);
+		strbuf_release(&sb);
+		if (matched)
+			return 0;
+	}
+
 	return o->gently ? -1 :
 		add_rejected_path(o, error_type, name);
 }
-- 
1.8.0.rc2.23.g1fb49df