builtin/repack.c: avoid dir traversal in `collect_pack_filenames()`
[git.git] / builtin / rm.c
blob463eeabceaac3a47839733aee96ec14ff2863cfc
1 /*
2 * "git rm" builtin command
4 * Copyright (C) Linus Torvalds 2006
5 */
6 #define USE_THE_INDEX_VARIABLE
7 #include "builtin.h"
8 #include "alloc.h"
9 #include "advice.h"
10 #include "config.h"
11 #include "lockfile.h"
12 #include "dir.h"
13 #include "cache-tree.h"
14 #include "gettext.h"
15 #include "hash.h"
16 #include "tree-walk.h"
17 #include "object-name.h"
18 #include "parse-options.h"
19 #include "read-cache.h"
20 #include "repository.h"
21 #include "string-list.h"
22 #include "setup.h"
23 #include "sparse-index.h"
24 #include "submodule.h"
25 #include "pathspec.h"
27 static const char * const builtin_rm_usage[] = {
28 N_("git rm [-f | --force] [-n] [-r] [--cached] [--ignore-unmatch]\n"
29 " [--quiet] [--pathspec-from-file=<file> [--pathspec-file-nul]]\n"
30 " [--] [<pathspec>...]"),
31 NULL
34 static struct {
35 int nr, alloc;
36 struct {
37 const char *name;
38 char is_submodule;
39 } *entry;
40 } list;
42 static int get_ours_cache_pos(const char *path, int pos)
44 int i = -pos - 1;
46 while ((i < the_index.cache_nr) && !strcmp(the_index.cache[i]->name, path)) {
47 if (ce_stage(the_index.cache[i]) == 2)
48 return i;
49 i++;
51 return -1;
54 static void print_error_files(struct string_list *files_list,
55 const char *main_msg,
56 const char *hints_msg,
57 int *errs)
59 if (files_list->nr) {
60 int i;
61 struct strbuf err_msg = STRBUF_INIT;
63 strbuf_addstr(&err_msg, main_msg);
64 for (i = 0; i < files_list->nr; i++)
65 strbuf_addf(&err_msg,
66 "\n %s",
67 files_list->items[i].string);
68 if (advice_enabled(ADVICE_RM_HINTS))
69 strbuf_addstr(&err_msg, hints_msg);
70 *errs = error("%s", err_msg.buf);
71 strbuf_release(&err_msg);
75 static void submodules_absorb_gitdir_if_needed(void)
77 int i;
78 for (i = 0; i < list.nr; i++) {
79 const char *name = list.entry[i].name;
80 int pos;
81 const struct cache_entry *ce;
83 pos = index_name_pos(&the_index, name, strlen(name));
84 if (pos < 0) {
85 pos = get_ours_cache_pos(name, pos);
86 if (pos < 0)
87 continue;
89 ce = the_index.cache[pos];
91 if (!S_ISGITLINK(ce->ce_mode) ||
92 !file_exists(ce->name) ||
93 is_empty_dir(name))
94 continue;
96 if (!submodule_uses_gitfile(name))
97 absorb_git_dir_into_superproject(name, NULL);
101 static int check_local_mod(struct object_id *head, int index_only)
104 * Items in list are already sorted in the cache order,
105 * so we could do this a lot more efficiently by using
106 * tree_desc based traversal if we wanted to, but I am
107 * lazy, and who cares if removal of files is a tad
108 * slower than the theoretical maximum speed?
110 int i, no_head;
111 int errs = 0;
112 struct string_list files_staged = STRING_LIST_INIT_NODUP;
113 struct string_list files_cached = STRING_LIST_INIT_NODUP;
114 struct string_list files_local = STRING_LIST_INIT_NODUP;
116 no_head = is_null_oid(head);
117 for (i = 0; i < list.nr; i++) {
118 struct stat st;
119 int pos;
120 const struct cache_entry *ce;
121 const char *name = list.entry[i].name;
122 struct object_id oid;
123 unsigned short mode;
124 int local_changes = 0;
125 int staged_changes = 0;
127 pos = index_name_pos(&the_index, name, strlen(name));
128 if (pos < 0) {
130 * Skip unmerged entries except for populated submodules
131 * that could lose history when removed.
133 pos = get_ours_cache_pos(name, pos);
134 if (pos < 0)
135 continue;
137 if (!S_ISGITLINK(the_index.cache[pos]->ce_mode) ||
138 is_empty_dir(name))
139 continue;
141 ce = the_index.cache[pos];
143 if (lstat(ce->name, &st) < 0) {
144 if (!is_missing_file_error(errno))
145 warning_errno(_("failed to stat '%s'"), ce->name);
146 /* It already vanished from the working tree */
147 continue;
149 else if (S_ISDIR(st.st_mode)) {
150 /* if a file was removed and it is now a
151 * directory, that is the same as ENOENT as
152 * far as git is concerned; we do not track
153 * directories unless they are submodules.
155 if (!S_ISGITLINK(ce->ce_mode))
156 continue;
160 * "rm" of a path that has changes need to be treated
161 * carefully not to allow losing local changes
162 * accidentally. A local change could be (1) file in
163 * work tree is different since the index; and/or (2)
164 * the user staged a content that is different from
165 * the current commit in the index.
167 * In such a case, you would need to --force the
168 * removal. However, "rm --cached" (remove only from
169 * the index) is safe if the index matches the file in
170 * the work tree or the HEAD commit, as it means that
171 * the content being removed is available elsewhere.
175 * Is the index different from the file in the work tree?
176 * If it's a submodule, is its work tree modified?
178 if (ie_match_stat(&the_index, ce, &st, 0) ||
179 (S_ISGITLINK(ce->ce_mode) &&
180 bad_to_remove_submodule(ce->name,
181 SUBMODULE_REMOVAL_DIE_ON_ERROR |
182 SUBMODULE_REMOVAL_IGNORE_IGNORED_UNTRACKED)))
183 local_changes = 1;
186 * Is the index different from the HEAD commit? By
187 * definition, before the very initial commit,
188 * anything staged in the index is treated by the same
189 * way as changed from the HEAD.
191 if (no_head
192 || get_tree_entry(the_repository, head, name, &oid, &mode)
193 || ce->ce_mode != create_ce_mode(mode)
194 || !oideq(&ce->oid, &oid))
195 staged_changes = 1;
198 * If the index does not match the file in the work
199 * tree and if it does not match the HEAD commit
200 * either, (1) "git rm" without --cached definitely
201 * will lose information; (2) "git rm --cached" will
202 * lose information unless it is about removing an
203 * "intent to add" entry.
205 if (local_changes && staged_changes) {
206 if (!index_only || !ce_intent_to_add(ce))
207 string_list_append(&files_staged, name);
209 else if (!index_only) {
210 if (staged_changes)
211 string_list_append(&files_cached, name);
212 if (local_changes)
213 string_list_append(&files_local, name);
216 print_error_files(&files_staged,
217 Q_("the following file has staged content different "
218 "from both the\nfile and the HEAD:",
219 "the following files have staged content different"
220 " from both the\nfile and the HEAD:",
221 files_staged.nr),
222 _("\n(use -f to force removal)"),
223 &errs);
224 string_list_clear(&files_staged, 0);
225 print_error_files(&files_cached,
226 Q_("the following file has changes "
227 "staged in the index:",
228 "the following files have changes "
229 "staged in the index:", files_cached.nr),
230 _("\n(use --cached to keep the file,"
231 " or -f to force removal)"),
232 &errs);
233 string_list_clear(&files_cached, 0);
235 print_error_files(&files_local,
236 Q_("the following file has local modifications:",
237 "the following files have local modifications:",
238 files_local.nr),
239 _("\n(use --cached to keep the file,"
240 " or -f to force removal)"),
241 &errs);
242 string_list_clear(&files_local, 0);
244 return errs;
247 static int show_only = 0, force = 0, index_only = 0, recursive = 0, quiet = 0;
248 static int ignore_unmatch = 0, pathspec_file_nul;
249 static int include_sparse;
250 static char *pathspec_from_file;
252 static struct option builtin_rm_options[] = {
253 OPT__DRY_RUN(&show_only, N_("dry run")),
254 OPT__QUIET(&quiet, N_("do not list removed files")),
255 OPT_BOOL( 0 , "cached", &index_only, N_("only remove from the index")),
256 OPT__FORCE(&force, N_("override the up-to-date check"), PARSE_OPT_NOCOMPLETE),
257 OPT_BOOL('r', NULL, &recursive, N_("allow recursive removal")),
258 OPT_BOOL( 0 , "ignore-unmatch", &ignore_unmatch,
259 N_("exit with a zero status even if nothing matched")),
260 OPT_BOOL(0, "sparse", &include_sparse, N_("allow updating entries outside of the sparse-checkout cone")),
261 OPT_PATHSPEC_FROM_FILE(&pathspec_from_file),
262 OPT_PATHSPEC_FILE_NUL(&pathspec_file_nul),
263 OPT_END(),
266 int cmd_rm(int argc, const char **argv, const char *prefix)
268 struct lock_file lock_file = LOCK_INIT;
269 int i, ret = 0;
270 struct pathspec pathspec;
271 char *seen;
273 git_config(git_default_config, NULL);
275 argc = parse_options(argc, argv, prefix, builtin_rm_options,
276 builtin_rm_usage, 0);
278 parse_pathspec(&pathspec, 0,
279 PATHSPEC_PREFER_CWD,
280 prefix, argv);
282 if (pathspec_from_file) {
283 if (pathspec.nr)
284 die(_("'%s' and pathspec arguments cannot be used together"), "--pathspec-from-file");
286 parse_pathspec_file(&pathspec, 0,
287 PATHSPEC_PREFER_CWD,
288 prefix, pathspec_from_file, pathspec_file_nul);
289 } else if (pathspec_file_nul) {
290 die(_("the option '%s' requires '%s'"), "--pathspec-file-nul", "--pathspec-from-file");
293 if (!pathspec.nr)
294 die(_("No pathspec was given. Which files should I remove?"));
296 if (!index_only)
297 setup_work_tree();
299 prepare_repo_settings(the_repository);
300 the_repository->settings.command_requires_full_index = 0;
301 repo_hold_locked_index(the_repository, &lock_file, LOCK_DIE_ON_ERROR);
303 if (repo_read_index(the_repository) < 0)
304 die(_("index file corrupt"));
306 refresh_index(&the_index, REFRESH_QUIET|REFRESH_UNMERGED, &pathspec, NULL, NULL);
308 seen = xcalloc(pathspec.nr, 1);
310 if (pathspec_needs_expanded_index(&the_index, &pathspec))
311 ensure_full_index(&the_index);
313 for (i = 0; i < the_index.cache_nr; i++) {
314 const struct cache_entry *ce = the_index.cache[i];
316 if (!include_sparse &&
317 (ce_skip_worktree(ce) ||
318 !path_in_sparse_checkout(ce->name, &the_index)))
319 continue;
320 if (!ce_path_match(&the_index, ce, &pathspec, seen))
321 continue;
322 ALLOC_GROW(list.entry, list.nr + 1, list.alloc);
323 list.entry[list.nr].name = xstrdup(ce->name);
324 list.entry[list.nr].is_submodule = S_ISGITLINK(ce->ce_mode);
325 if (list.entry[list.nr++].is_submodule &&
326 !is_staging_gitmodules_ok(&the_index))
327 die(_("please stage your changes to .gitmodules or stash them to proceed"));
330 if (pathspec.nr) {
331 const char *original;
332 int seen_any = 0;
333 char *skip_worktree_seen = NULL;
334 struct string_list only_match_skip_worktree = STRING_LIST_INIT_NODUP;
336 for (i = 0; i < pathspec.nr; i++) {
337 original = pathspec.items[i].original;
338 if (seen[i])
339 seen_any = 1;
340 else if (ignore_unmatch)
341 continue;
342 else if (!include_sparse &&
343 matches_skip_worktree(&pathspec, i, &skip_worktree_seen))
344 string_list_append(&only_match_skip_worktree, original);
345 else
346 die(_("pathspec '%s' did not match any files"), original);
348 if (!recursive && seen[i] == MATCHED_RECURSIVELY)
349 die(_("not removing '%s' recursively without -r"),
350 *original ? original : ".");
353 if (only_match_skip_worktree.nr) {
354 advise_on_updating_sparse_paths(&only_match_skip_worktree);
355 ret = 1;
357 free(skip_worktree_seen);
358 string_list_clear(&only_match_skip_worktree, 0);
360 if (!seen_any)
361 exit(ret);
363 clear_pathspec(&pathspec);
364 free(seen);
366 if (!index_only)
367 submodules_absorb_gitdir_if_needed();
370 * If not forced, the file, the index and the HEAD (if exists)
371 * must match; but the file can already been removed, since
372 * this sequence is a natural "novice" way:
374 * rm F; git rm F
376 * Further, if HEAD commit exists, "diff-index --cached" must
377 * report no changes unless forced.
379 if (!force) {
380 struct object_id oid;
381 if (repo_get_oid(the_repository, "HEAD", &oid))
382 oidclr(&oid);
383 if (check_local_mod(&oid, index_only))
384 exit(1);
388 * First remove the names from the index: we won't commit
389 * the index unless all of them succeed.
391 for (i = 0; i < list.nr; i++) {
392 const char *path = list.entry[i].name;
393 if (!quiet)
394 printf("rm '%s'\n", path);
396 if (remove_file_from_index(&the_index, path))
397 die(_("git rm: unable to remove %s"), path);
400 if (show_only)
401 return 0;
404 * Then, unless we used "--cached", remove the filenames from
405 * the workspace. If we fail to remove the first one, we
406 * abort the "git rm" (but once we've successfully removed
407 * any file at all, we'll go ahead and commit to it all:
408 * by then we've already committed ourselves and can't fail
409 * in the middle)
411 if (!index_only) {
412 int removed = 0, gitmodules_modified = 0;
413 struct strbuf buf = STRBUF_INIT;
414 int flag = force ? REMOVE_DIR_PURGE_ORIGINAL_CWD : 0;
415 for (i = 0; i < list.nr; i++) {
416 const char *path = list.entry[i].name;
417 if (list.entry[i].is_submodule) {
418 strbuf_reset(&buf);
419 strbuf_addstr(&buf, path);
420 if (remove_dir_recursively(&buf, flag))
421 die(_("could not remove '%s'"), path);
423 removed = 1;
424 if (!remove_path_from_gitmodules(path))
425 gitmodules_modified = 1;
426 continue;
428 if (!remove_path(path)) {
429 removed = 1;
430 continue;
432 if (!removed)
433 die_errno("git rm: '%s'", path);
435 strbuf_release(&buf);
436 if (gitmodules_modified)
437 stage_updated_gitmodules(&the_index);
440 if (write_locked_index(&the_index, &lock_file,
441 COMMIT_LOCK | SKIP_IF_UNCHANGED))
442 die(_("Unable to write new index file"));
444 return ret;