1 /* notmuch - Not much of an email program, (just index and search)
2 *
3 * Copyright © 2009 Carl Worth
4 *
5 * This program is free software: you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation, either version 3 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program. If not, see https://www.gnu.org/licenses/ .
17 *
18 * Author: Carl Worth <cworth@cworth.org>
19 */
20
21 #include "notmuch-client.h"
22 #include "tag-util.h"
23
24 #include <unistd.h>
25
26 typedef struct _filename_node {
27 char *filename;
28 time_t mtime;
29 struct _filename_node *next;
30 } _filename_node_t;
31
32 typedef struct _filename_list {
33 unsigned count;
34 _filename_node_t *head;
35 _filename_node_t **tail;
36 } _filename_list_t;
37
38 enum verbosity {
39 VERBOSITY_QUIET,
40 VERBOSITY_NORMAL,
41 VERBOSITY_VERBOSE,
42 };
43
44 typedef struct {
45 const char *db_path;
46 const char *mail_root;
47
48 int output_is_a_tty;
49 enum verbosity verbosity;
50 bool debug;
51 bool full_scan;
52 notmuch_config_values_t *new_tags;
53 const char **ignore_verbatim;
54 size_t ignore_verbatim_length;
55 regex_t *ignore_regex;
56 size_t ignore_regex_length;
57
58 int total_files;
59 int processed_files;
60 int added_messages, removed_messages, renamed_messages;
61 int vanished_files;
62 struct timeval tv_start;
63
64 _filename_list_t *removed_files;
65 _filename_list_t *removed_directories;
66 _filename_list_t *directory_mtimes;
67
68 notmuch_bool_t synchronize_flags;
69 } add_files_state_t;
70
71 static volatile sig_atomic_t do_print_progress = 0;
72
73 static void
handle_sigalrm(unused (int signal))74 handle_sigalrm (unused (int signal))
75 {
76 do_print_progress = 1;
77 }
78
79 static volatile sig_atomic_t interrupted;
80
81 static void
handle_sigint(unused (int sig))82 handle_sigint (unused (int sig))
83 {
84 static const char msg[] = "Stopping... \n";
85
86 /* This write is "opportunistic", so it's okay to ignore the
87 * result. It is not required for correctness, and if it does
88 * fail or produce a short write, we want to get out of the signal
89 * handler as quickly as possible, not retry it. */
90 IGNORE_RESULT (write (2, msg, sizeof (msg) - 1));
91 interrupted = 1;
92 }
93
94 static _filename_list_t *
_filename_list_create(const void * ctx)95 _filename_list_create (const void *ctx)
96 {
97 _filename_list_t *list;
98
99 list = talloc (ctx, _filename_list_t);
100 if (list == NULL)
101 return NULL;
102
103 list->head = NULL;
104 list->tail = &list->head;
105 list->count = 0;
106
107 return list;
108 }
109
110 static _filename_node_t *
_filename_list_add(_filename_list_t * list,const char * filename)111 _filename_list_add (_filename_list_t *list,
112 const char *filename)
113 {
114 _filename_node_t *node = talloc (list, _filename_node_t);
115
116 list->count++;
117
118 node->filename = talloc_strdup (list, filename);
119 node->next = NULL;
120
121 *(list->tail) = node;
122 list->tail = &node->next;
123
124 return node;
125 }
126
127 static void
generic_print_progress(const char * action,const char * object,struct timeval tv_start,unsigned processed,unsigned total)128 generic_print_progress (const char *action, const char *object,
129 struct timeval tv_start, unsigned processed, unsigned total)
130 {
131 struct timeval tv_now;
132 double elapsed_overall, rate_overall;
133
134 gettimeofday (&tv_now, NULL);
135
136 elapsed_overall = notmuch_time_elapsed (tv_start, tv_now);
137 rate_overall = processed / elapsed_overall;
138
139 printf ("%s %u ", action, processed);
140
141 if (total) {
142 printf ("of %u %s", total, object);
143 if (processed > 0 && elapsed_overall > 0.5) {
144 double time_remaining = ((total - processed) / rate_overall);
145 printf (" (");
146 notmuch_time_print_formatted_seconds (time_remaining);
147 printf (" remaining)");
148 }
149 } else {
150 printf ("%s", object);
151 if (elapsed_overall > 0.5)
152 printf (" (%d %s/sec.)", (int) rate_overall, object);
153 }
154 printf (".\033[K\r");
155
156 fflush (stdout);
157 }
158
159 static int
dirent_sort_inode(const struct dirent ** a,const struct dirent ** b)160 dirent_sort_inode (const struct dirent **a, const struct dirent **b)
161 {
162 return ((*a)->d_ino < (*b)->d_ino) ? -1 : 1;
163 }
164
165 static int
dirent_sort_strcmp_name(const struct dirent ** a,const struct dirent ** b)166 dirent_sort_strcmp_name (const struct dirent **a, const struct dirent **b)
167 {
168 return strcmp ((*a)->d_name, (*b)->d_name);
169 }
170
171 /* Return the type of a directory entry relative to path as a stat(2)
172 * mode. Like stat, this follows symlinks. Returns -1 and sets errno
173 * if the file's type cannot be determined (which includes dangling
174 * symlinks).
175 */
176 static int
dirent_type(const char * path,const struct dirent * entry)177 dirent_type (const char *path, const struct dirent *entry)
178 {
179 struct stat statbuf;
180 char *abspath;
181 int err, saved_errno;
182
183 #if HAVE_D_TYPE
184 /* Mapping from d_type to stat mode_t. We omit DT_LNK so that
185 * we'll fall through to stat and get the real file type. */
186 static const mode_t modes[] = {
187 [DT_BLK] = S_IFBLK,
188 [DT_CHR] = S_IFCHR,
189 [DT_DIR] = S_IFDIR,
190 [DT_FIFO] = S_IFIFO,
191 [DT_REG] = S_IFREG,
192 [DT_SOCK] = S_IFSOCK
193 };
194 if (entry->d_type < ARRAY_SIZE (modes) && modes[entry->d_type])
195 return modes[entry->d_type];
196 #endif
197
198 abspath = talloc_asprintf (NULL, "%s/%s", path, entry->d_name);
199 if (! abspath) {
200 errno = ENOMEM;
201 return -1;
202 }
203 err = stat (abspath, &statbuf);
204 saved_errno = errno;
205 talloc_free (abspath);
206 if (err < 0) {
207 errno = saved_errno;
208 return -1;
209 }
210 return statbuf.st_mode & S_IFMT;
211 }
212
213 /* Test if the directory looks like a Maildir directory.
214 *
215 * Search through the array of directory entries to see if we can find all
216 * three subdirectories typical for Maildir, that is "new", "cur", and "tmp".
217 *
218 * Return 1 if the directory looks like a Maildir and 0 otherwise.
219 */
220 static int
_entries_resemble_maildir(const char * path,struct dirent ** entries,int count)221 _entries_resemble_maildir (const char *path, struct dirent **entries, int count)
222 {
223 int i, found = 0;
224
225 for (i = 0; i < count; i++) {
226 if (dirent_type (path, entries[i]) != S_IFDIR)
227 continue;
228
229 if (strcmp (entries[i]->d_name, "new") == 0 ||
230 strcmp (entries[i]->d_name, "cur") == 0 ||
231 strcmp (entries[i]->d_name, "tmp") == 0) {
232 found++;
233 if (found == 3)
234 return 1;
235 }
236 }
237
238 return 0;
239 }
240
241 static bool
_special_directory(const char * entry)242 _special_directory (const char *entry)
243 {
244 return strcmp (entry, ".") == 0 || strcmp (entry, "..") == 0;
245 }
246
247 static bool
_setup_ignore(notmuch_database_t * notmuch,add_files_state_t * state)248 _setup_ignore (notmuch_database_t *notmuch, add_files_state_t *state)
249 {
250 notmuch_config_values_t *ignore_list;
251 int nregex = 0, nverbatim = 0;
252 const char **verbatim = NULL;
253 regex_t *regex = NULL;
254
255 for (ignore_list = notmuch_config_get_values (notmuch, NOTMUCH_CONFIG_NEW_IGNORE);
256 notmuch_config_values_valid (ignore_list);
257 notmuch_config_values_move_to_next (ignore_list)) {
258 const char *s = notmuch_config_values_get (ignore_list);
259 size_t len = strlen (s);
260
261 if (len == 0) {
262 fprintf (stderr, "Error: Empty string in new.ignore list\n");
263 return false;
264 }
265
266 if (s[0] == '/') {
267 regex_t *preg;
268 char *r;
269 int rerr;
270
271 if (len < 3 || s[len - 1] != '/') {
272 fprintf (stderr, "Error: Malformed pattern '%s' in new.ignore\n",
273 s);
274 return false;
275 }
276
277 r = talloc_strndup (notmuch, s + 1, len - 2);
278 regex = talloc_realloc (notmuch, regex, regex_t, nregex + 1);
279 preg = ®ex[nregex];
280
281 rerr = regcomp (preg, r, REG_EXTENDED | REG_NOSUB);
282 if (rerr) {
283 size_t error_size = regerror (rerr, preg, NULL, 0);
284 char *error = talloc_size (r, error_size);
285
286 regerror (rerr, preg, error, error_size);
287
288 fprintf (stderr, "Error: Invalid regex '%s' in new.ignore: %s\n",
289 r, error);
290 return false;
291 }
292 nregex++;
293
294 talloc_free (r);
295 } else {
296 verbatim = talloc_realloc (notmuch, verbatim, const char *,
297 nverbatim + 1);
298 verbatim[nverbatim++] = s;
299 }
300 }
301
302 state->ignore_regex = regex;
303 state->ignore_regex_length = nregex;
304 state->ignore_verbatim = verbatim;
305 state->ignore_verbatim_length = nverbatim;
306
307 return true;
308 }
309
310 static char *
_get_relative_path(const char * mail_root,const char * dirpath,const char * entry)311 _get_relative_path (const char *mail_root, const char *dirpath, const char *entry)
312 {
313 size_t mail_root_len = strlen (mail_root);
314
315 /* paranoia? */
316 if (strncmp (dirpath, mail_root, mail_root_len) != 0) {
317 fprintf (stderr, "Warning: '%s' is not a subdirectory of '%s'\n",
318 dirpath, mail_root);
319 return NULL;
320 }
321
322 dirpath += mail_root_len;
323 while (*dirpath == '/')
324 dirpath++;
325
326 if (*dirpath)
327 return talloc_asprintf (NULL, "%s/%s", dirpath, entry);
328 else
329 return talloc_strdup (NULL, entry);
330 }
331
332 /* Test if the file/directory is to be ignored.
333 */
334 static bool
_entry_in_ignore_list(add_files_state_t * state,const char * dirpath,const char * entry)335 _entry_in_ignore_list (add_files_state_t *state, const char *dirpath,
336 const char *entry)
337 {
338 bool ret = false;
339 size_t i;
340 char *path;
341
342 for (i = 0; i < state->ignore_verbatim_length; i++) {
343 if (strcmp (entry, state->ignore_verbatim[i]) == 0)
344 return true;
345 }
346
347 if (state->ignore_regex_length == 0)
348 return false;
349
350 path = _get_relative_path (state->mail_root, dirpath, entry);
351 if (! path)
352 return false;
353
354 for (i = 0; i < state->ignore_regex_length; i++) {
355 if (regexec (&state->ignore_regex[i], path, 0, NULL, 0) == 0) {
356 ret = true;
357 break;
358 }
359 }
360
361 talloc_free (path);
362
363 return ret;
364 }
365
366 /* Add a single file to the database. */
367 static notmuch_status_t
add_file(notmuch_database_t * notmuch,const char * filename,add_files_state_t * state)368 add_file (notmuch_database_t *notmuch, const char *filename,
369 add_files_state_t *state)
370 {
371 notmuch_message_t *message = NULL;
372 const char *tag;
373 notmuch_status_t status;
374
375 status = notmuch_database_begin_atomic (notmuch);
376 if (status)
377 goto DONE;
378
379 status = notmuch_database_index_file (notmuch, filename, indexing_cli_choices.opts, &message);
380 switch (status) {
381 /* Success. */
382 case NOTMUCH_STATUS_SUCCESS:
383 state->added_messages++;
384 notmuch_message_freeze (message);
385 if (state->synchronize_flags)
386 notmuch_message_maildir_flags_to_tags (message);
387
388 for (notmuch_config_values_start (state->new_tags);
389 notmuch_config_values_valid (state->new_tags);
390 notmuch_config_values_move_to_next (state->new_tags)) {
391 notmuch_bool_t is_set;
392
393 tag = notmuch_config_values_get (state->new_tags);
394 /* Currently all errors from has_maildir_flag are fatal */
395 if ((status = notmuch_message_has_maildir_flag_st (message, 'S', &is_set)))
396 goto DONE;
397 if (strcmp ("unread", tag) != 0 || ! is_set) {
398 notmuch_message_add_tag (message, tag);
399 }
400 }
401
402 notmuch_message_thaw (message);
403 break;
404 /* Non-fatal issues (go on to next file). */
405 case NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID:
406 if (state->synchronize_flags) {
407 status = notmuch_message_maildir_flags_to_tags (message);
408 if (print_status_message ("add_file", message, status))
409 goto DONE;
410 }
411 break;
412 case NOTMUCH_STATUS_FILE_NOT_EMAIL:
413 fprintf (stderr, "Note: Ignoring non-mail file: %s\n", filename);
414 break;
415 case NOTMUCH_STATUS_FILE_ERROR:
416 /* Someone renamed/removed the file between scandir and now. */
417 state->vanished_files++;
418 fprintf (stderr, "Unexpected error with file %s\n", filename);
419 (void) print_status_database ("add_file", notmuch, status);
420 break;
421 /* Fatal issues. Don't process anymore. */
422 case NOTMUCH_STATUS_READ_ONLY_DATABASE:
423 case NOTMUCH_STATUS_XAPIAN_EXCEPTION:
424 case NOTMUCH_STATUS_OUT_OF_MEMORY:
425 (void) print_status_database ("add_file", notmuch, status);
426 goto DONE;
427 default:
428 INTERNAL_ERROR ("add_message returned unexpected value: %d", status);
429 goto DONE;
430 }
431
432 status = notmuch_database_end_atomic (notmuch);
433
434 DONE:
435 if (message)
436 notmuch_message_destroy (message);
437
438 return status;
439 }
440
441 /* Examine 'path' recursively as follows:
442 *
443 * o Ask the filesystem for the mtime of 'path' (fs_mtime)
444 * o Ask the database for its timestamp of 'path' (db_mtime)
445 *
446 * o Ask the filesystem for files and directories within 'path'
447 * (via scandir and stored in fs_entries)
448 *
449 * o Pass 1: For each directory in fs_entries, recursively call into
450 * this same function.
451 *
452 * o Compare fs_mtime to db_mtime. If they are equivalent, terminate
453 * the algorithm at this point, (this directory has not been
454 * updated in the filesystem since the last database scan of PASS
455 * 2).
456 *
457 * o Ask the database for files and directories within 'path'
458 * (db_files and db_subdirs)
459 *
460 * o Pass 2: Walk fs_entries simultaneously with db_files and
461 * db_subdirs. Look for one of three interesting cases:
462 *
463 * 1. Regular file in fs_entries and not in db_files
464 * This is a new file to add_message into the database.
465 *
466 * 2. Filename in db_files not in fs_entries.
467 * This is a file that has been removed from the mail store.
468 *
469 * 3. Directory in db_subdirs not in fs_entries
470 * This is a directory that has been removed from the mail store.
471 *
472 * Note that the addition of a directory is not interesting here,
473 * since that will have been taken care of in pass 1. Also, we
474 * don't immediately act on file/directory removal since we must
475 * ensure that in the case of a rename that the new filename is
476 * added before the old filename is removed, (so that no
477 * information is lost from the database).
478 *
479 * o Tell the database to update its time of 'path' to 'fs_mtime'
480 * if fs_mtime isn't the current wall-clock time.
481 */
482 static notmuch_status_t
add_files(notmuch_database_t * notmuch,const char * path,add_files_state_t * state)483 add_files (notmuch_database_t *notmuch,
484 const char *path,
485 add_files_state_t *state)
486 {
487 struct dirent *entry = NULL;
488 char *next = NULL;
489 time_t fs_mtime, db_mtime;
490 notmuch_status_t status, ret = NOTMUCH_STATUS_SUCCESS;
491 struct dirent **fs_entries = NULL;
492 int i, num_fs_entries = 0, entry_type;
493 notmuch_directory_t *directory;
494 notmuch_filenames_t *db_files = NULL;
495 notmuch_filenames_t *db_subdirs = NULL;
496 time_t stat_time;
497 struct stat st;
498 bool is_maildir;
499
500 if (stat (path, &st)) {
501 fprintf (stderr, "Error reading directory %s: %s\n",
502 path, strerror (errno));
503 return NOTMUCH_STATUS_FILE_ERROR;
504 }
505 stat_time = time (NULL);
506
507 if (! S_ISDIR (st.st_mode)) {
508 fprintf (stderr, "Error: %s is not a directory.\n", path);
509 return NOTMUCH_STATUS_FILE_ERROR;
510 }
511
512 fs_mtime = st.st_mtime;
513
514 status = notmuch_database_get_directory (notmuch, path, &directory);
515 if (status) {
516 ret = status;
517 goto DONE;
518 }
519 db_mtime = directory ? notmuch_directory_get_mtime (directory) : 0;
520
521 /* If the directory is unchanged from our last scan and has no
522 * sub-directories, then return without scanning it at all. In
523 * some situations, skipping the scan can substantially reduce the
524 * cost of notmuch new, especially since the huge numbers of files
525 * in Maildirs make scans expensive, but all files live in leaf
526 * directories.
527 *
528 * To check for sub-directories, we borrow a trick from find,
529 * kpathsea, and many other UNIX tools: since a directory's link
530 * count is the number of sub-directories (specifically, their
531 * '..' entries) plus 2 (the link from the parent and the link for
532 * '.'). This check is safe even on weird file systems, since
533 * file systems that can't compute this will return 0 or 1. This
534 * is safe even on *really* weird file systems like HFS+ that
535 * mistakenly return the total number of directory entries, since
536 * that only inflates the count beyond 2.
537 */
538 if (directory && (! state->full_scan) && fs_mtime == db_mtime && st.st_nlink == 2) {
539 /* There's one catch: pass 1 below considers symlinks to
540 * directories to be directories, but these don't increase the
541 * file system link count. So, only bail early if the
542 * database agrees that there are no sub-directories. */
543 db_subdirs = notmuch_directory_get_child_directories (directory);
544 if (! notmuch_filenames_valid (db_subdirs))
545 goto DONE;
546 notmuch_filenames_destroy (db_subdirs);
547 db_subdirs = NULL;
548 }
549
550 /* If the database knows about this directory, then we sort based
551 * on strcmp to match the database sorting. Otherwise, we can do
552 * inode-based sorting for faster filesystem operation. */
553 num_fs_entries = scandir (path, &fs_entries, 0,
554 directory ?
555 dirent_sort_strcmp_name : dirent_sort_inode);
556
557 if (num_fs_entries == -1) {
558 fprintf (stderr, "Error opening directory %s: %s\n",
559 path, strerror (errno));
560 /* We consider this a fatal error because, if a user moved a
561 * message from another directory that we were able to scan
562 * into this directory, skipping this directory will cause
563 * that message to be lost. */
564 ret = NOTMUCH_STATUS_FILE_ERROR;
565 goto DONE;
566 }
567
568 /* Pass 1: Recurse into all sub-directories. */
569 is_maildir = _entries_resemble_maildir (path, fs_entries, num_fs_entries);
570
571 for (i = 0; i < num_fs_entries && ! interrupted; i++) {
572 entry = fs_entries[i];
573
574 /* Ignore special directories to avoid infinite recursion. */
575 if (_special_directory (entry->d_name))
576 continue;
577
578 /* Ignore any files/directories the user has configured to
579 * ignore. We do this before dirent_type both for performance
580 * and because we don't care if dirent_type fails on entries
581 * that are explicitly ignored.
582 */
583 if (_entry_in_ignore_list (state, path, entry->d_name)) {
584 if (state->debug)
585 printf ("(D) add_files, pass 1: explicitly ignoring %s/%s\n",
586 path, entry->d_name);
587 continue;
588 }
589
590 /* We only want to descend into directories (and symlinks to
591 * directories). */
592 entry_type = dirent_type (path, entry);
593 if (entry_type == -1) {
594 /* Be pessimistic, e.g. so we don't lose lots of mail just
595 * because a user broke a symlink. */
596 fprintf (stderr, "Error reading file %s/%s: %s\n",
597 path, entry->d_name, strerror (errno));
598 return NOTMUCH_STATUS_FILE_ERROR;
599 } else if (entry_type != S_IFDIR) {
600 continue;
601 }
602
603 /* Ignore the .notmuch directory and any "tmp" directory
604 * that appears within a maildir.
605 */
606 if ((is_maildir && strcmp (entry->d_name, "tmp") == 0) ||
607 strcmp (entry->d_name, ".notmuch") == 0)
608 continue;
609
610 next = talloc_asprintf (notmuch, "%s/%s", path, entry->d_name);
611 status = add_files (notmuch, next, state);
612 if (status) {
613 ret = status;
614 goto DONE;
615 }
616 talloc_free (next);
617 next = NULL;
618 }
619
620 /* If the directory's modification time in the filesystem is the
621 * same as what we recorded in the database the last time we
622 * scanned it, then we can skip the second pass entirely.
623 *
624 * We test for strict equality here to avoid a bug that can happen
625 * if the system clock jumps backward, (preventing new mail from
626 * being discovered until the clock catches up and the directory
627 * is modified again).
628 */
629 if (directory && (! state->full_scan) && fs_mtime == db_mtime)
630 goto DONE;
631
632 /* If the database has never seen this directory before, we can
633 * simply leave db_files and db_subdirs NULL. */
634 if (directory) {
635 db_files = notmuch_directory_get_child_files (directory);
636 db_subdirs = notmuch_directory_get_child_directories (directory);
637 }
638
639 /* Pass 2: Scan for new files, removed files, and removed directories. */
640 for (i = 0; i < num_fs_entries && ! interrupted; i++) {
641 entry = fs_entries[i];
642
643 /* Ignore special directories early. */
644 if (_special_directory (entry->d_name))
645 continue;
646
647 /* Ignore files & directories user has configured to be ignored */
648 if (_entry_in_ignore_list (state, path, entry->d_name)) {
649 if (state->debug)
650 printf ("(D) add_files, pass 2: explicitly ignoring %s/%s\n",
651 path, entry->d_name);
652 continue;
653 }
654
655 /* Check if we've walked past any names in db_files or
656 * db_subdirs. If so, these have been deleted. */
657 while (notmuch_filenames_valid (db_files) &&
658 strcmp (notmuch_filenames_get (db_files), entry->d_name) < 0) {
659 char *absolute = talloc_asprintf (state->removed_files,
660 "%s/%s", path,
661 notmuch_filenames_get (db_files));
662
663 if (state->debug)
664 printf ("(D) add_files, pass 2: queuing passed file %s for deletion from database\n",
665 absolute);
666
667 _filename_list_add (state->removed_files, absolute);
668
669 notmuch_filenames_move_to_next (db_files);
670 }
671
672 while (notmuch_filenames_valid (db_subdirs) &&
673 strcmp (notmuch_filenames_get (db_subdirs), entry->d_name) <= 0) {
674 const char *filename = notmuch_filenames_get (db_subdirs);
675
676 if (strcmp (filename, entry->d_name) < 0) {
677 char *absolute = talloc_asprintf (state->removed_directories,
678 "%s/%s", path, filename);
679 if (state->debug)
680 printf (
681 "(D) add_files, pass 2: queuing passed directory %s for deletion from database\n",
682 absolute);
683
684 _filename_list_add (state->removed_directories, absolute);
685 }
686
687 notmuch_filenames_move_to_next (db_subdirs);
688 }
689
690 /* Only add regular files (and symlinks to regular files). */
691 entry_type = dirent_type (path, entry);
692 if (entry_type == -1) {
693 fprintf (stderr, "Error reading file %s/%s: %s\n",
694 path, entry->d_name, strerror (errno));
695 return NOTMUCH_STATUS_FILE_ERROR;
696 } else if (entry_type != S_IFREG) {
697 continue;
698 }
699
700 /* Don't add a file that we've added before. */
701 if (notmuch_filenames_valid (db_files) &&
702 strcmp (notmuch_filenames_get (db_files), entry->d_name) == 0) {
703 notmuch_filenames_move_to_next (db_files);
704 continue;
705 }
706
707 /* We're now looking at a regular file that doesn't yet exist
708 * in the database, so add it. */
709 next = talloc_asprintf (notmuch, "%s/%s", path, entry->d_name);
710
711 state->processed_files++;
712
713 if (state->verbosity >= VERBOSITY_VERBOSE) {
714 if (state->output_is_a_tty)
715 printf ("\r\033[K");
716
717 printf ("%i/%i: %s", state->processed_files, state->total_files,
718 next);
719
720 putchar ((state->output_is_a_tty) ? '\r' : '\n');
721 fflush (stdout);
722 }
723
724 status = add_file (notmuch, next, state);
725 if (status) {
726 ret = status;
727 goto DONE;
728 }
729
730 if (do_print_progress) {
731 do_print_progress = 0;
732 generic_print_progress ("Processed", "files", state->tv_start,
733 state->processed_files, state->total_files);
734 }
735
736 talloc_free (next);
737 next = NULL;
738 }
739
740 if (interrupted)
741 goto DONE;
742
743 /* Now that we've walked the whole filesystem list, anything left
744 * over in the database lists has been deleted. */
745 while (notmuch_filenames_valid (db_files)) {
746 char *absolute = talloc_asprintf (state->removed_files,
747 "%s/%s", path,
748 notmuch_filenames_get (db_files));
749 if (state->debug)
750 printf ("(D) add_files, pass 3: queuing leftover file %s for deletion from database\n",
751 absolute);
752
753 _filename_list_add (state->removed_files, absolute);
754
755 notmuch_filenames_move_to_next (db_files);
756 }
757
758 while (notmuch_filenames_valid (db_subdirs)) {
759 char *absolute = talloc_asprintf (state->removed_directories,
760 "%s/%s", path,
761 notmuch_filenames_get (db_subdirs));
762
763 if (state->debug)
764 printf (
765 "(D) add_files, pass 3: queuing leftover directory %s for deletion from database\n",
766 absolute);
767
768 _filename_list_add (state->removed_directories, absolute);
769
770 notmuch_filenames_move_to_next (db_subdirs);
771 }
772
773 /* If the directory's mtime is the same as the wall-clock time
774 * when we stat'ed the directory, we skip updating the mtime in
775 * the database because a message could be delivered later in this
776 * same second. This may lead to unnecessary re-scans, but it
777 * avoids overlooking messages. */
778 if (fs_mtime != stat_time)
779 _filename_list_add (state->directory_mtimes, path)->mtime = fs_mtime;
780
781 DONE:
782 if (next)
783 talloc_free (next);
784 if (fs_entries) {
785 for (i = 0; i < num_fs_entries; i++)
786 free (fs_entries[i]);
787
788 free (fs_entries);
789 }
790 if (db_subdirs)
791 notmuch_filenames_destroy (db_subdirs);
792 if (db_files)
793 notmuch_filenames_destroy (db_files);
794 if (directory)
795 notmuch_directory_destroy (directory);
796
797 return ret;
798 }
799
800 static void
setup_progress_printing_timer(void)801 setup_progress_printing_timer (void)
802 {
803 struct sigaction action;
804 struct itimerval timerval;
805
806 /* Set up our handler for SIGALRM */
807 memset (&action, 0, sizeof (struct sigaction));
808 action.sa_handler = handle_sigalrm;
809 sigemptyset (&action.sa_mask);
810 action.sa_flags = SA_RESTART;
811 sigaction (SIGALRM, &action, NULL);
812
813 /* Then start a timer to send SIGALRM once per second. */
814 timerval.it_interval.tv_sec = 1;
815 timerval.it_interval.tv_usec = 0;
816 timerval.it_value.tv_sec = 1;
817 timerval.it_value.tv_usec = 0;
818 setitimer (ITIMER_REAL, &timerval, NULL);
819 }
820
821 static void
stop_progress_printing_timer(void)822 stop_progress_printing_timer (void)
823 {
824 struct sigaction action;
825 struct itimerval timerval;
826
827 /* Now stop the timer. */
828 timerval.it_interval.tv_sec = 0;
829 timerval.it_interval.tv_usec = 0;
830 timerval.it_value.tv_sec = 0;
831 timerval.it_value.tv_usec = 0;
832 setitimer (ITIMER_REAL, &timerval, NULL);
833
834 /* And disable the signal handler. */
835 action.sa_handler = SIG_IGN;
836 sigaction (SIGALRM, &action, NULL);
837 }
838
839
840 /* XXX: This should be merged with the add_files function since it
841 * shares a lot of logic with it. */
842 /* Recursively count all regular files in path and all sub-directories
843 * of path. The result is added to *count (which should be
844 * initialized to zero by the top-level caller before calling
845 * count_files). */
846 static void
count_files(const char * path,int * count,add_files_state_t * state)847 count_files (const char *path, int *count, add_files_state_t *state)
848 {
849 struct dirent *entry = NULL;
850 char *next;
851 struct dirent **fs_entries = NULL;
852 int num_fs_entries = scandir (path, &fs_entries, 0, dirent_sort_inode);
853 int entry_type, i;
854
855 if (num_fs_entries == -1) {
856 fprintf (stderr, "Warning: failed to open directory %s: %s\n",
857 path, strerror (errno));
858 goto DONE;
859 }
860
861 for (i = 0; i < num_fs_entries && ! interrupted; i++) {
862 entry = fs_entries[i];
863
864 /* Ignore special directories to avoid infinite recursion.
865 * Also ignore the .notmuch directory.
866 */
867 if (_special_directory (entry->d_name) ||
868 strcmp (entry->d_name, ".notmuch") == 0)
869 continue;
870
871 /* Ignore any files/directories the user has configured to be
872 * ignored
873 */
874 if (_entry_in_ignore_list (state, path, entry->d_name)) {
875 if (state->debug)
876 printf ("(D) count_files: explicitly ignoring %s/%s\n",
877 path, entry->d_name);
878 continue;
879 }
880
881 if (asprintf (&next, "%s/%s", path, entry->d_name) == -1) {
882 next = NULL;
883 fprintf (stderr, "Error descending from %s to %s: Out of memory\n",
884 path, entry->d_name);
885 continue;
886 }
887
888 entry_type = dirent_type (path, entry);
889 if (entry_type == S_IFREG) {
890 *count = *count + 1;
891 if (*count % 1000 == 0 && state->verbosity >= VERBOSITY_NORMAL) {
892 printf ("Found %d files so far.\r", *count);
893 fflush (stdout);
894 }
895 } else if (entry_type == S_IFDIR) {
896 count_files (next, count, state);
897 }
898
899 free (next);
900 }
901
902 DONE:
903 if (fs_entries) {
904 for (i = 0; i < num_fs_entries; i++)
905 free (fs_entries[i]);
906
907 free (fs_entries);
908 }
909 }
910
911 static void
upgrade_print_progress(void * closure,double progress)912 upgrade_print_progress (void *closure,
913 double progress)
914 {
915 add_files_state_t *state = closure;
916
917 printf ("Upgrading database: %.2f%% complete", progress * 100.0);
918
919 if (progress > 0) {
920 struct timeval tv_now;
921 double elapsed, time_remaining;
922
923 gettimeofday (&tv_now, NULL);
924
925 elapsed = notmuch_time_elapsed (state->tv_start, tv_now);
926 time_remaining = (elapsed / progress) * (1.0 - progress);
927 printf (" (");
928 notmuch_time_print_formatted_seconds (time_remaining);
929 printf (" remaining)");
930 }
931
932 printf (". \r");
933
934 fflush (stdout);
935 }
936
937 /* Remove one message filename from the database. */
938 static notmuch_status_t
remove_filename(notmuch_database_t * notmuch,const char * path,add_files_state_t * add_files_state)939 remove_filename (notmuch_database_t *notmuch,
940 const char *path,
941 add_files_state_t *add_files_state)
942 {
943 notmuch_status_t status;
944 notmuch_message_t *message;
945
946 status = notmuch_database_begin_atomic (notmuch);
947 if (status)
948 return status;
949 status = notmuch_database_find_message_by_filename (notmuch, path, &message);
950 if (status || message == NULL)
951 goto DONE;
952
953 status = notmuch_database_remove_message (notmuch, path);
954 if (status == NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID) {
955 add_files_state->renamed_messages++;
956 if (add_files_state->synchronize_flags == true)
957 notmuch_message_maildir_flags_to_tags (message);
958 status = NOTMUCH_STATUS_SUCCESS;
959 } else if (status == NOTMUCH_STATUS_SUCCESS) {
960 add_files_state->removed_messages++;
961 }
962 notmuch_message_destroy (message);
963
964 DONE:
965 notmuch_database_end_atomic (notmuch);
966 return status;
967 }
968
969 /* Recursively remove all filenames from the database referring to
970 * 'path' (or to any of its children). */
971 static notmuch_status_t
_remove_directory(notmuch_database_t * notmuch,const char * path,add_files_state_t * add_files_state)972 _remove_directory (notmuch_database_t *notmuch,
973 const char *path,
974 add_files_state_t *add_files_state)
975 {
976 notmuch_status_t status;
977 notmuch_directory_t *directory;
978 notmuch_filenames_t *files, *subdirs;
979 char *absolute;
980
981 status = notmuch_database_get_directory (notmuch, path, &directory);
982 if (status || ! directory)
983 return status;
984
985 for (files = notmuch_directory_get_child_files (directory);
986 notmuch_filenames_valid (files);
987 notmuch_filenames_move_to_next (files)) {
988 absolute = talloc_asprintf (notmuch, "%s/%s", path,
989 notmuch_filenames_get (files));
990 status = remove_filename (notmuch, absolute, add_files_state);
991 talloc_free (absolute);
992 if (status)
993 goto DONE;
994 }
995
996 for (subdirs = notmuch_directory_get_child_directories (directory);
997 notmuch_filenames_valid (subdirs);
998 notmuch_filenames_move_to_next (subdirs)) {
999 absolute = talloc_asprintf (notmuch, "%s/%s", path,
1000 notmuch_filenames_get (subdirs));
1001 status = _remove_directory (notmuch, absolute, add_files_state);
1002 talloc_free (absolute);
1003 if (status)
1004 goto DONE;
1005 }
1006
1007 status = notmuch_directory_delete (directory);
1008
1009 DONE:
1010 if (status)
1011 notmuch_directory_destroy (directory);
1012 return status;
1013 }
1014
1015 static void
print_results(const add_files_state_t * state)1016 print_results (const add_files_state_t *state)
1017 {
1018 double elapsed;
1019 struct timeval tv_now;
1020
1021 gettimeofday (&tv_now, NULL);
1022 elapsed = notmuch_time_elapsed (state->tv_start, tv_now);
1023
1024 if (state->processed_files) {
1025 printf ("Processed %d %s in ", state->processed_files,
1026 state->processed_files == 1 ? "file" : "total files");
1027 notmuch_time_print_formatted_seconds (elapsed);
1028 if (elapsed > 1)
1029 printf (" (%d files/sec.)",
1030 (int) (state->processed_files / elapsed));
1031 printf (".%s\n", (state->output_is_a_tty) ? "\033[K" : "");
1032 }
1033
1034 if (state->added_messages)
1035 printf ("Added %d new %s to the database.", state->added_messages,
1036 state->added_messages == 1 ? "message" : "messages");
1037 else
1038 printf ("No new mail.");
1039
1040 if (state->removed_messages)
1041 printf (" Removed %d %s.", state->removed_messages,
1042 state->removed_messages == 1 ? "message" : "messages");
1043
1044 if (state->renamed_messages)
1045 printf (" Detected %d file %s.", state->renamed_messages,
1046 state->renamed_messages == 1 ? "rename" : "renames");
1047
1048 printf ("\n");
1049 }
1050
1051 static int
_maybe_upgrade(notmuch_database_t * notmuch,add_files_state_t * state)1052 _maybe_upgrade (notmuch_database_t *notmuch, add_files_state_t *state)
1053 {
1054 if (notmuch_database_needs_upgrade (notmuch)) {
1055 time_t now = time (NULL);
1056 struct tm *gm_time = gmtime (&now);
1057 int err;
1058 notmuch_status_t status;
1059 const char *backup_dir = notmuch_config_get (notmuch, NOTMUCH_CONFIG_BACKUP_DIR);
1060 const char *backup_name;
1061
1062 err = mkdir (backup_dir, 0755);
1063 if (err && errno != EEXIST) {
1064 fprintf (stderr, "Failed to create %s: %s\n", backup_dir, strerror (errno));
1065 return EXIT_FAILURE;
1066 }
1067
1068 /* since dump files are written atomically, the amount of
1069 * harm from overwriting one within a second seems
1070 * relatively small. */
1071 backup_name = talloc_asprintf (notmuch, "%s/dump-%04d%02d%02dT%02d%02d%02d.gz",
1072 backup_dir,
1073 gm_time->tm_year + 1900,
1074 gm_time->tm_mon + 1,
1075 gm_time->tm_mday,
1076 gm_time->tm_hour,
1077 gm_time->tm_min,
1078 gm_time->tm_sec);
1079
1080 if (state->verbosity >= VERBOSITY_NORMAL) {
1081 printf ("Welcome to a new version of notmuch! Your database will now be upgraded.\n");
1082 printf ("This process is safe to interrupt.\n");
1083 printf ("Backing up tags to %s...\n", backup_name);
1084 }
1085
1086 if (notmuch_database_dump (notmuch, backup_name, "",
1087 DUMP_FORMAT_BATCH_TAG, DUMP_INCLUDE_DEFAULT, true)) {
1088 fprintf (stderr, "Backup failed. Aborting upgrade.");
1089 return EXIT_FAILURE;
1090 }
1091
1092 gettimeofday (&state->tv_start, NULL);
1093 status = notmuch_database_upgrade (
1094 notmuch,
1095 state->verbosity >= VERBOSITY_NORMAL ? upgrade_print_progress : NULL,
1096 state);
1097 if (status) {
1098 printf ("Upgrade failed: %s\n",
1099 notmuch_status_to_string (status));
1100 notmuch_database_destroy (notmuch);
1101 return EXIT_FAILURE;
1102 }
1103 if (state->verbosity >= VERBOSITY_NORMAL)
1104 printf ("Your notmuch database has now been upgraded.\n");
1105 }
1106 return EXIT_SUCCESS;
1107 }
1108
1109 int
notmuch_new_command(notmuch_database_t * notmuch,int argc,char * argv[])1110 notmuch_new_command (notmuch_database_t *notmuch, int argc, char *argv[])
1111 {
1112 add_files_state_t add_files_state = {
1113 .verbosity = VERBOSITY_NORMAL,
1114 .debug = false,
1115 .full_scan = false,
1116 .output_is_a_tty = isatty (fileno (stdout)),
1117 };
1118 struct timeval tv_start;
1119 int ret = 0;
1120 const char *db_path, *mail_root;
1121 struct sigaction action;
1122 _filename_node_t *f;
1123 int opt_index;
1124 unsigned int i;
1125 bool timer_is_active = false;
1126 bool hooks = true;
1127 bool quiet = false, verbose = false;
1128 notmuch_status_t status;
1129
1130 notmuch_opt_desc_t options[] = {
1131 { .opt_bool = &quiet, .name = "quiet" },
1132 { .opt_bool = &verbose, .name = "verbose" },
1133 { .opt_bool = &add_files_state.debug, .name = "debug" },
1134 { .opt_bool = &add_files_state.full_scan, .name = "full-scan" },
1135 { .opt_bool = &hooks, .name = "hooks" },
1136 { .opt_inherit = notmuch_shared_indexing_options },
1137 { .opt_inherit = notmuch_shared_options },
1138 { }
1139 };
1140
1141 opt_index = parse_arguments (argc, argv, options, 1);
1142 if (opt_index < 0)
1143 return EXIT_FAILURE;
1144
1145 notmuch_process_shared_options (notmuch, argv[0]);
1146
1147 /* quiet trumps verbose */
1148 if (quiet)
1149 add_files_state.verbosity = VERBOSITY_QUIET;
1150 else if (verbose)
1151 add_files_state.verbosity = VERBOSITY_VERBOSE;
1152
1153 add_files_state.new_tags = notmuch_config_get_values (notmuch, NOTMUCH_CONFIG_NEW_TAGS);
1154
1155 if (print_status_database (
1156 "notmuch new",
1157 notmuch,
1158 notmuch_config_get_bool (notmuch, NOTMUCH_CONFIG_SYNC_MAILDIR_FLAGS,
1159 &add_files_state.synchronize_flags)))
1160 return EXIT_FAILURE;
1161
1162 db_path = notmuch_config_get (notmuch, NOTMUCH_CONFIG_DATABASE_PATH);
1163 add_files_state.db_path = db_path;
1164
1165 mail_root = notmuch_config_get (notmuch, NOTMUCH_CONFIG_MAIL_ROOT);
1166 add_files_state.mail_root = mail_root;
1167
1168 if (! _setup_ignore (notmuch, &add_files_state))
1169 return EXIT_FAILURE;
1170
1171 for (notmuch_config_values_start (add_files_state.new_tags);
1172 notmuch_config_values_valid (add_files_state.new_tags);
1173 notmuch_config_values_move_to_next (add_files_state.new_tags)) {
1174 const char *tag, *error_msg;
1175
1176 tag = notmuch_config_values_get (add_files_state.new_tags);
1177 error_msg = illegal_tag (tag, false);
1178 if (error_msg) {
1179 fprintf (stderr, "Error: tag '%s' in new.tags: %s\n", tag, error_msg);
1180 return EXIT_FAILURE;
1181 }
1182 }
1183
1184 if (hooks) {
1185 /* Drop write lock to run hook */
1186 status = notmuch_database_reopen (notmuch, NOTMUCH_DATABASE_MODE_READ_ONLY);
1187 if (print_status_database ("notmuch new", notmuch, status))
1188 return EXIT_FAILURE;
1189
1190 ret = notmuch_run_hook (notmuch, "pre-new");
1191 if (ret)
1192 return EXIT_FAILURE;
1193
1194 /* acquire write lock again */
1195 status = notmuch_database_reopen (notmuch, NOTMUCH_DATABASE_MODE_READ_WRITE);
1196 if (print_status_database ("notmuch new", notmuch, status))
1197 return EXIT_FAILURE;
1198 }
1199
1200 if (notmuch_database_get_revision (notmuch, NULL) == 0) {
1201 int count = 0;
1202 count_files (mail_root, &count, &add_files_state);
1203 if (interrupted)
1204 return EXIT_FAILURE;
1205
1206 if (add_files_state.verbosity >= VERBOSITY_NORMAL)
1207 printf ("Found %d total files (that's not much mail).\n", count);
1208
1209 add_files_state.total_files = count;
1210 } else {
1211 if (_maybe_upgrade (notmuch, &add_files_state))
1212 return EXIT_FAILURE;
1213
1214 add_files_state.total_files = 0;
1215 }
1216
1217 if (notmuch == NULL)
1218 return EXIT_FAILURE;
1219
1220 status = notmuch_process_shared_indexing_options (notmuch);
1221 if (status != NOTMUCH_STATUS_SUCCESS) {
1222 fprintf (stderr, "Error: Failed to process index options. (%s)\n",
1223 notmuch_status_to_string (status));
1224 return EXIT_FAILURE;
1225 }
1226
1227 /* Set up our handler for SIGINT. We do this after having
1228 * potentially done a database upgrade we this interrupt handler
1229 * won't support. */
1230 memset (&action, 0, sizeof (struct sigaction));
1231 action.sa_handler = handle_sigint;
1232 sigemptyset (&action.sa_mask);
1233 action.sa_flags = SA_RESTART;
1234 sigaction (SIGINT, &action, NULL);
1235
1236 gettimeofday (&add_files_state.tv_start, NULL);
1237
1238 add_files_state.removed_files = _filename_list_create (notmuch);
1239 add_files_state.removed_directories = _filename_list_create (notmuch);
1240 add_files_state.directory_mtimes = _filename_list_create (notmuch);
1241
1242 if (add_files_state.verbosity == VERBOSITY_NORMAL &&
1243 add_files_state.output_is_a_tty && ! debugger_is_active ()) {
1244 setup_progress_printing_timer ();
1245 timer_is_active = true;
1246 }
1247
1248 ret = add_files (notmuch, mail_root, &add_files_state);
1249 if (ret)
1250 goto DONE;
1251
1252 gettimeofday (&tv_start, NULL);
1253 for (f = add_files_state.removed_files->head; f && ! interrupted; f = f->next) {
1254 ret = remove_filename (notmuch, f->filename, &add_files_state);
1255 if (ret)
1256 goto DONE;
1257 if (do_print_progress) {
1258 do_print_progress = 0;
1259 generic_print_progress ("Cleaned up", "messages",
1260 tv_start, add_files_state.removed_messages +
1261 add_files_state.renamed_messages,
1262 add_files_state.removed_files->count);
1263 }
1264 }
1265
1266 gettimeofday (&tv_start, NULL);
1267 for (f = add_files_state.removed_directories->head, i = 0; f && ! interrupted; f = f->next, i++) {
1268 ret = _remove_directory (notmuch, f->filename, &add_files_state);
1269 if (ret)
1270 goto DONE;
1271 if (do_print_progress) {
1272 do_print_progress = 0;
1273 generic_print_progress ("Cleaned up", "directories",
1274 tv_start, i,
1275 add_files_state.removed_directories->count);
1276 }
1277 }
1278
1279 for (f = add_files_state.directory_mtimes->head; f && ! interrupted; f = f->next) {
1280 notmuch_directory_t *directory;
1281 status = notmuch_database_get_directory (notmuch, f->filename, &directory);
1282 if (status == NOTMUCH_STATUS_SUCCESS && directory) {
1283 notmuch_directory_set_mtime (directory, f->mtime);
1284 notmuch_directory_destroy (directory);
1285 }
1286 }
1287
1288 DONE:
1289 talloc_free (add_files_state.removed_files);
1290 talloc_free (add_files_state.removed_directories);
1291 talloc_free (add_files_state.directory_mtimes);
1292
1293 if (timer_is_active)
1294 stop_progress_printing_timer ();
1295
1296 if (add_files_state.verbosity >= VERBOSITY_NORMAL)
1297 print_results (&add_files_state);
1298
1299 if (ret)
1300 fprintf (stderr, "Note: A fatal error was encountered: %s\n",
1301 notmuch_status_to_string (ret));
1302
1303 notmuch_database_close (notmuch);
1304
1305 if (hooks && ! ret && ! interrupted)
1306 ret = notmuch_run_hook (notmuch, "post-new");
1307
1308 notmuch_database_destroy (notmuch);
1309
1310 if (ret || interrupted)
1311 return EXIT_FAILURE;
1312
1313 if (add_files_state.vanished_files)
1314 return NOTMUCH_EXIT_TEMPFAIL;
1315
1316 return EXIT_SUCCESS;
1317 }
1318