Make reference lookups apply precomposeunicode

Before these changes, looking up a reference would return the
same precomposed or decomposed form of the reference name that
was used to look it up, so on MacOS which ignores the difference
between the two, a single reference could be looked up either way
and git_reference_name would return the form of the name that was
used to look it up!  This change makes lookup always return the
precomposed name if core.precomposeunicode is set regardless of
which version was used to look it up.  The reference iterator was
already returning the precomposed form from earlier work.

This also updates the CMakeLists.txt rules for enabling iconv
usage because the clar tests for this code were actually not being
activated properly with the old version.

Finally, this moves git_repository_reset_filesystem from include/
git2/repository.h to include/git2/sys/repository.h since it is not
really a function that normal library users should have to think
about very often.

3 files changed, 46 insertions, 28 deletions

diff --git a/src/path.h b/src/path.h
index 175756938..eaf94d486 100644
--- a/src/path.h
+++ b/src/path.h
@@ -365,11 +365,6 @@ extern int git_path_set_error(
 /* check if non-ascii characters are present in filename */
 extern bool git_path_has_non_ascii(const char *path, size_t pathlen);
 
-/* only enable iconv on Mac for now */
-#ifdef __APPLE__
-#define GIT_USE_ICONV 1
-#endif
-
 #define GIT_PATH_REPO_ENCODING "UTF-8"
 
 #ifdef __APPLE__
diff --git a/src/refs.c b/src/refs.c
index c045ab9dc..0da02a666 100644
--- a/src/refs.c
+++ b/src/refs.c
@@ -138,6 +138,22 @@ int git_reference_name_to_id(
 	return 0;
 }
 
+static int reference_normalize_for_repo(
+	char *out,
+	size_t out_size,
+	git_repository *repo,
+	const char *name)
+{
+	int precompose;
+	unsigned int flags = GIT_REF_FORMAT_ALLOW_ONELEVEL;
+
+	if (!git_repository__cvar(&precompose, repo, GIT_CVAR_PRECOMPOSE) &&
+		precompose)
+		flags |= GIT_REF_FORMAT__PRECOMPOSE_UNICODE;
+
+	return git_reference_normalize_name(out, out_size, name, flags);
+}
+
 int git_reference_lookup_resolved(
 	git_reference **ref_out,
 	git_repository *repo,
@@ -159,13 +175,13 @@ int git_reference_lookup_resolved(
 	else if (max_nesting < 0)
 		max_nesting = DEFAULT_NESTING_LEVEL;
 
-	strncpy(scan_name, name, GIT_REFNAME_MAX);
 	scan_type = GIT_REF_SYMBOLIC;
 
-	if ((error = git_repository_refdb__weakptr(&refdb, repo)) < 0)
-		return -1;
+	if ((error = reference_normalize_for_repo(
+			scan_name, sizeof(scan_name), repo, name)) < 0)
+		return error;
 
-	if ((error = git_reference__normalize_name_lax(scan_name, GIT_REFNAME_MAX, name)) < 0)
+	if ((error = git_repository_refdb__weakptr(&refdb, repo)) < 0)
 		return error;
 
 	for (nesting = max_nesting;
@@ -173,7 +189,7 @@ int git_reference_lookup_resolved(
 		 nesting--)
 	{
 		if (nesting != max_nesting) {
-			strncpy(scan_name, ref->target.symbolic, GIT_REFNAME_MAX);
+			strncpy(scan_name, ref->target.symbolic, sizeof(scan_name));
 			git_reference_free(ref);
 		}
 
@@ -711,17 +727,18 @@ static bool is_all_caps_and_underscore(const char *name, size_t len)
 	return true;
 }
 
+/* Inspired from https://github.com/git/git/blob/f06d47e7e0d9db709ee204ed13a8a7486149f494/refs.c#L36-100 */
 int git_reference__normalize_name(
 	git_buf *buf,
 	const char *name,
 	unsigned int flags)
 {
-	// Inspired from https://github.com/git/git/blob/f06d47e7e0d9db709ee204ed13a8a7486149f494/refs.c#L36-100
-
 	char *current;
 	int segment_len, segments_count = 0, error = GIT_EINVALIDSPEC;
 	unsigned int process_flags;
 	bool normalize = (buf != NULL);
+	git_path_iconv_t ic = GIT_PATH_ICONV_INIT;
+
 	assert(name);
 
 	process_flags = flags;
@@ -733,6 +750,13 @@ int git_reference__normalize_name(
 	if (normalize)
 		git_buf_clear(buf);
 
+	if ((flags & GIT_REF_FORMAT__PRECOMPOSE_UNICODE) != 0) {
+		size_t namelen = strlen(current);
+		if ((error = git_path_iconv_init_precompose(&ic)) < 0 ||
+			(error = git_path_iconv(&ic, &current, &namelen)) < 0)
+			goto cleanup;
+	}
+
 	while (true) {
 		segment_len = ensure_segment_validity(current);
 		if (segment_len < 0) {
@@ -809,6 +833,8 @@ cleanup:
 	if (error && normalize)
 		git_buf_free(buf);
 
+	git_path_iconv_clear(&ic);
+
 	return error;
 }
 
@@ -983,9 +1009,9 @@ static int peel_error(int error, git_reference *ref, const char* msg)
 }
 
 int git_reference_peel(
-		git_object **peeled,
-		git_reference *ref,
-		git_otype target_type)
+	git_object **peeled,
+	git_reference *ref,
+	git_otype target_type)
 {
 	git_reference *resolved = NULL;
 	git_object *target = NULL;
@@ -1027,24 +1053,19 @@ cleanup:
 	return error;
 }
 
-int git_reference__is_valid_name(
-	const char *refname,
-	unsigned int flags)
+int git_reference__is_valid_name(const char *refname, unsigned int flags)
 {
-	int error;
-
-	error = git_reference__normalize_name(NULL, refname, flags) == 0;
-	giterr_clear();
+	if (git_reference__normalize_name(NULL, refname, flags) < 0) {
+		giterr_clear();
+		return false;
+	}
 
-	return error;
+	return true;
 }
 
-int git_reference_is_valid_name(
-	const char *refname)
+int git_reference_is_valid_name(const char *refname)
 {
-	return git_reference__is_valid_name(
-		refname,
-		GIT_REF_FORMAT_ALLOW_ONELEVEL);
+	return git_reference__is_valid_name(refname, GIT_REF_FORMAT_ALLOW_ONELEVEL);
 }
 
 const char *git_reference_shorthand(git_reference *ref)
diff --git a/src/refs.h b/src/refs.h
index 4b91c25e8..80c7703fc 100644
--- a/src/refs.h
+++ b/src/refs.h
@@ -46,6 +46,8 @@
 #define GIT_STASH_FILE "stash"
 #define GIT_REFS_STASH_FILE GIT_REFS_DIR GIT_STASH_FILE
 
+#define GIT_REF_FORMAT__PRECOMPOSE_UNICODE	(1u << 16)
+
 #define GIT_REFNAME_MAX 1024
 
 struct git_reference {