Merge pull request #3711 from joshtriplett/git_repository_discover_default

Add GIT_REPOSITORY_OPEN_FROM_ENV flag to respect $GIT_* environment vars

CHANGELOG.md

@@ -3,6 +3,13 @@ v0.24 + 1
 
 ### Changes or improvements
 
+* Fix repository discovery with `git_repository_discover` and
+  `git_repository_open_ext` to match git's handling of a ceiling
+  directory at the current directory. git only checks ceiling
+  directories when its search ascends to a parent directory.  A ceiling
+  directory matching the starting directory will not prevent git from
+  finding a repository in the starting directory or a parent directory.
+
 ### API additions
 
 * `git_commit_create_buffer()` creates a commit and writes it into a
@@ -13,6 +20,25 @@ v0.24 + 1
   writing into a stream. Useful when you do not know the final size or
   want to copy the contents from another stream.
 
+* New flags for `git_repository_open_ext`:
+
+    * `GIT_REPOSITORY_OPEN_NO_DOTGIT` - Do not check for a repository by
+      appending `/.git` to the `start_path`; only open the repository if
+      `start_path` itself points to the git directory.
+    * `GIT_REPOSITORY_OPEN_FROM_ENV` - Find and open a git repository,
+      respecting the environment variables used by the git command-line
+      tools. If set, `git_repository_open_ext` will ignore the other
+      flags and the `ceiling_dirs` argument, and will allow a NULL
+      `path` to use `GIT_DIR` or search from the current directory. The
+      search for a repository will respect `$GIT_CEILING_DIRECTORIES`
+      and `$GIT_DISCOVERY_ACROSS_FILESYSTEM`.  The opened repository
+      will respect `$GIT_INDEX_FILE`, `$GIT_NAMESPACE`,
+      `$GIT_OBJECT_DIRECTORY`, and `$GIT_ALTERNATE_OBJECT_DIRECTORIES`.
+      In the future, this flag will also cause `git_repository_open_ext`
+      to respect `$GIT_WORK_TREE` and `$GIT_COMMON_DIR`; currently,
+      `git_repository_open_ext` with this flag will error out if either
+      `$GIT_WORK_TREE` or `$GIT_COMMON_DIR` is set.
+
 ### API removals
 
 * `git_blob_create_fromchunks()` has been removed in favour of

include/git2/repository.h

@@ -95,11 +95,29 @@ GIT_EXTERN(int) git_repository_discover(
  * * GIT_REPOSITORY_OPEN_BARE - Open repository as a bare repo regardless
  *   of core.bare config, and defer loading config file for faster setup.
  *   Unlike `git_repository_open_bare`, this can follow gitlinks.
+ * * GIT_REPOSITORY_OPEN_NO_DOTGIT - Do not check for a repository by
+ *   appending /.git to the start_path; only open the repository if
+ *   start_path itself points to the git directory.
+ * * GIT_REPOSITORY_OPEN_FROM_ENV - Find and open a git repository,
+ *   respecting the environment variables used by the git command-line
+ *   tools. If set, `git_repository_open_ext` will ignore the other
+ *   flags and the `ceiling_dirs` argument, and will allow a NULL `path`
+ *   to use `GIT_DIR` or search from the current directory. The search
+ *   for a repository will respect $GIT_CEILING_DIRECTORIES and
+ *   $GIT_DISCOVERY_ACROSS_FILESYSTEM.  The opened repository will
+ *   respect $GIT_INDEX_FILE, $GIT_NAMESPACE, $GIT_OBJECT_DIRECTORY, and
+ *   $GIT_ALTERNATE_OBJECT_DIRECTORIES.  In the future, this flag will
+ *   also cause `git_repository_open_ext` to respect $GIT_WORK_TREE and
+ *   $GIT_COMMON_DIR; currently, `git_repository_open_ext` with this
+ *   flag will error out if either $GIT_WORK_TREE or $GIT_COMMON_DIR is
+ *   set.
  */
 typedef enum {
 	GIT_REPOSITORY_OPEN_NO_SEARCH = (1 << 0),
 	GIT_REPOSITORY_OPEN_CROSS_FS  = (1 << 1),
 	GIT_REPOSITORY_OPEN_BARE      = (1 << 2),
+	GIT_REPOSITORY_OPEN_NO_DOTGIT = (1 << 3),
+	GIT_REPOSITORY_OPEN_FROM_ENV  = (1 << 4),
 } git_repository_open_flag_t;
 
 /**
@@ -110,7 +128,8 @@ typedef enum {
  *        see if a repo at this path could be opened.
  * @param path Path to open as git repository.  If the flags
  *        permit "searching", then this can be a path to a subdirectory
- *        inside the working directory of the repository.
+ *        inside the working directory of the repository. May be NULL if
+ *        flags is GIT_REPOSITORY_OPEN_FROM_ENV.
  * @param flags A combination of the GIT_REPOSITORY_OPEN flags above.
  * @param ceiling_dirs A GIT_PATH_LIST_SEPARATOR delimited list of path
  *        prefixes at which the search for a containing repository should

src/repository.c

@@ -359,7 +359,8 @@ static int find_repo(
 	git_buf path = GIT_BUF_INIT;
 	struct stat st;
 	dev_t initial_device = 0;
-	bool try_with_dot_git = ((flags & GIT_REPOSITORY_OPEN_BARE) != 0);
+	int min_iterations;
+	bool in_dot_git;
 	int ceiling_offset;
 
 	git_buf_free(repo_path);
@@ -367,13 +368,30 @@ static int find_repo(
 	if ((error = git_path_prettify(&path, start_path, NULL)) < 0)
 		return error;
 
-	ceiling_offset = find_ceiling_dir_offset(path.ptr, ceiling_dirs);
+	/* in_dot_git toggles each loop:
+	 * /a/b/c/.git, /a/b/c, /a/b/.git, /a/b, /a/.git, /a
+	 * With GIT_REPOSITORY_OPEN_BARE or GIT_REPOSITORY_OPEN_NO_DOTGIT, we
+	 * assume we started with /a/b/c.git and don't append .git the first
+	 * time through.
+	 * min_iterations indicates the number of iterations left before going
+	 * further counts as a search. */
+	if (flags & (GIT_REPOSITORY_OPEN_BARE | GIT_REPOSITORY_OPEN_NO_DOTGIT)) {
+		in_dot_git = true;
+		min_iterations = 1;
+	} else {
+		in_dot_git = false;
+		min_iterations = 2;
+	}
 
-	if (!try_with_dot_git &&
-		(error = git_buf_joinpath(&path, path.ptr, DOT_GIT)) < 0)
-		return error;
+	while (!error && (min_iterations || !(path.ptr[ceiling_offset] == 0 ||
+					      (flags & GIT_REPOSITORY_OPEN_NO_SEARCH)))) {
+		if (!(flags & GIT_REPOSITORY_OPEN_NO_DOTGIT)) {
+			if (!in_dot_git)
+				if ((error = git_buf_joinpath(&path, path.ptr, DOT_GIT)) < 0)
+					break;
+			in_dot_git = !in_dot_git;
+		}
 
-	while (!error && !git_buf_len(repo_path)) {
 		if (p_stat(path.ptr, &st) == 0) {
 			/* check that we have not crossed device boundaries */
 			if (initial_device == 0)
@@ -414,17 +432,10 @@ static int find_repo(
 			break;
 		}
 
-		if (try_with_dot_git) {
-			/* if we tried original dir with and without .git AND either hit
-			 * directory ceiling or NO_SEARCH was requested, then be done.
-			 */
-			if (path.ptr[ceiling_offset] == '\0' ||
-				(flags & GIT_REPOSITORY_OPEN_NO_SEARCH) != 0)
-				break;
-			/* otherwise look first for .git item */
-			error = git_buf_joinpath(&path, path.ptr, DOT_GIT);
-		}
-		try_with_dot_git = !try_with_dot_git;
+		/* Once we've checked the directory (and .git if applicable),
+		 * find the ceiling for a search. */
+		if (min_iterations && (--min_iterations == 0))
+			ceiling_offset = find_ceiling_dir_offset(path.ptr, ceiling_dirs);
 	}
 
 	if (!error && parent_path && !(flags & GIT_REPOSITORY_OPEN_BARE)) {
@@ -480,6 +491,172 @@ int git_repository_open_bare(
 	return 0;
 }
 
+static int _git_repository_open_ext_from_env(
+	git_repository **out,
+	const char *start_path)
+{
+	git_repository *repo = NULL;
+	git_index *index = NULL;
+	git_odb *odb = NULL;
+	git_buf dir_buf = GIT_BUF_INIT;
+	git_buf ceiling_dirs_buf = GIT_BUF_INIT;
+	git_buf across_fs_buf = GIT_BUF_INIT;
+	git_buf index_file_buf = GIT_BUF_INIT;
+	git_buf namespace_buf = GIT_BUF_INIT;
+	git_buf object_dir_buf = GIT_BUF_INIT;
+	git_buf alts_buf = GIT_BUF_INIT;
+	git_buf work_tree_buf = GIT_BUF_INIT;
+	git_buf common_dir_buf = GIT_BUF_INIT;
+	const char *ceiling_dirs = NULL;
+	unsigned flags = 0;
+	int error;
+
+	if (!start_path) {
+		error = git__getenv(&dir_buf, "GIT_DIR");
+		if (error == GIT_ENOTFOUND) {
+			giterr_clear();
+			start_path = ".";
+		} else if (error < 0)
+			goto error;
+		else {
+			start_path = git_buf_cstr(&dir_buf);
+			flags |= GIT_REPOSITORY_OPEN_NO_SEARCH;
+			flags |= GIT_REPOSITORY_OPEN_NO_DOTGIT;
+		}
+	}
+
+	error = git__getenv(&ceiling_dirs_buf, "GIT_CEILING_DIRECTORIES");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else
+		ceiling_dirs = git_buf_cstr(&ceiling_dirs_buf);
+
+	error = git__getenv(&across_fs_buf, "GIT_DISCOVERY_ACROSS_FILESYSTEM");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else {
+		int across_fs = 0;
+		error = git_config_parse_bool(&across_fs, git_buf_cstr(&across_fs_buf));
+		if (error < 0)
+			goto error;
+		if (across_fs)
+			flags |= GIT_REPOSITORY_OPEN_CROSS_FS;
+	}
+
+	error = git__getenv(&index_file_buf, "GIT_INDEX_FILE");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else {
+		error = git_index_open(&index, git_buf_cstr(&index_file_buf));
+		if (error < 0)
+			goto error;
+	}
+
+	error = git__getenv(&namespace_buf, "GIT_NAMESPACE");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+
+	error = git__getenv(&object_dir_buf, "GIT_OBJECT_DIRECTORY");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else {
+		error = git_odb_open(&odb, git_buf_cstr(&object_dir_buf));
+		if (error < 0)
+			goto error;
+	}
+
+	error = git__getenv(&work_tree_buf, "GIT_WORK_TREE");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else {
+		giterr_set(GITERR_INVALID, "GIT_WORK_TREE unimplemented");
+		error = GIT_ERROR;
+		goto error;
+	}
+
+	error = git__getenv(&work_tree_buf, "GIT_COMMON_DIR");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+	else {
+		giterr_set(GITERR_INVALID, "GIT_COMMON_DIR unimplemented");
+		error = GIT_ERROR;
+		goto error;
+	}
+
+	error = git_repository_open_ext(&repo, start_path, flags, ceiling_dirs);
+	if (error < 0)
+		goto error;
+
+	if (odb)
+		git_repository_set_odb(repo, odb);
+
+	error = git__getenv(&alts_buf, "GIT_ALTERNATE_OBJECT_DIRECTORIES");
+	if (error == GIT_ENOTFOUND)
+		giterr_clear();
+	else if (error < 0)
+		goto error;
+        else {
+		const char *end;
+		char *alt, *sep;
+		if (!odb) {
+			error = git_repository_odb(&odb, repo);
+			if (error < 0)
+				goto error;
+		}
+
+		end = git_buf_cstr(&alts_buf) + git_buf_len(&alts_buf);
+		for (sep = alt = alts_buf.ptr; sep != end; alt = sep+1) {
+			for (sep = alt; *sep && *sep != GIT_PATH_LIST_SEPARATOR; sep++)
+				;
+			if (*sep)
+				*sep = '\0';
+			error = git_odb_add_disk_alternate(odb, alt);
+			if (error < 0)
+				goto error;
+		}
+	}
+
+	error = git_repository_set_namespace(repo, git_buf_cstr(&namespace_buf));
+	if (error < 0)
+		goto error;
+
+	git_repository_set_index(repo, index);
+
+	if (out) {
+		*out = repo;
+		goto success;
+	}
+error:
+	git_repository_free(repo);
+success:
+	git_odb_free(odb);
+	git_index_free(index);
+	git_buf_free(&common_dir_buf);
+	git_buf_free(&work_tree_buf);
+	git_buf_free(&alts_buf);
+	git_buf_free(&object_dir_buf);
+	git_buf_free(&namespace_buf);
+	git_buf_free(&index_file_buf);
+	git_buf_free(&across_fs_buf);
+	git_buf_free(&ceiling_dirs_buf);
+	git_buf_free(&dir_buf);
+	return error;
+}
+
 int git_repository_open_ext(
 	git_repository **repo_ptr,
 	const char *start_path,
@@ -492,6 +669,9 @@ int git_repository_open_ext(
 	git_repository *repo;
 	git_config *config = NULL;
 
+	if (flags & GIT_REPOSITORY_OPEN_FROM_ENV)
+		return _git_repository_open_ext_from_env(repo_ptr, start_path);
+
 	if (repo_ptr)
 		*repo_ptr = NULL;
 

tests/repo/discover.c

@@ -118,12 +118,22 @@ void test_repo_discover__0(void)
 	cl_git_fail(git_repository_discover(&found_path, ALTERNATE_MALFORMED_FOLDER3, 0, ceiling_dirs));
 	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, ALTERNATE_NOT_FOUND_FOLDER, 0, ceiling_dirs));
 
+	append_ceiling_dir(&ceiling_dirs_buf, SUB_REPOSITORY_FOLDER_SUB);
+	ceiling_dirs = git_buf_cstr(&ceiling_dirs_buf);
+
+	/* this must pass as ceiling_directories cannot prevent the current
+	 * working directory to be checked */
+	ensure_repository_discover(SUB_REPOSITORY_FOLDER, ceiling_dirs, &sub_repository_path);
+	ensure_repository_discover(SUB_REPOSITORY_FOLDER_SUB, ceiling_dirs, &sub_repository_path);
+	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER_SUB_SUB, 0, ceiling_dirs));
+	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER_SUB_SUB_SUB, 0, ceiling_dirs));
+
 	append_ceiling_dir(&ceiling_dirs_buf, SUB_REPOSITORY_FOLDER);
 	ceiling_dirs = git_buf_cstr(&ceiling_dirs_buf);
 
 	//this must pass as ceiling_directories cannot predent the current
 	//working directory to be checked
-	cl_git_pass(git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER, 0, ceiling_dirs));
+	ensure_repository_discover(SUB_REPOSITORY_FOLDER, ceiling_dirs, &sub_repository_path);
 	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER_SUB, 0, ceiling_dirs));
 	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER_SUB_SUB, 0, ceiling_dirs));
 	cl_assert_equal_i(GIT_ENOTFOUND, git_repository_discover(&found_path, SUB_REPOSITORY_FOLDER_SUB_SUB_SUB, 0, ceiling_dirs));