Raw
1 #define USE_THE_REPOSITORY_VARIABLE
2 #define DISABLE_SIGN_COMPARE_WARNINGS
3
4 #include "git-compat-util.h"
5 #include "environment.h"
6 #include "gettext.h"
7 #include "range-diff.h"
8 #include "object-name.h"
9 #include "string-list.h"
10 #include "run-command.h"
11 #include "strvec.h"
12 #include "hashmap.h"
13 #include "xdiff-interface.h"
14 #include "linear-assignment.h"
15 #include "diffcore.h"
16 #include "commit.h"
17 #include "pager.h"
18 #include "pretty.h"
19 #include "repository.h"
20 #include "userdiff.h"
21 #include "apply.h"
22 #include "revision.h"
23
24 struct patch_util {
25 /* For the search for an exact match */
26 struct hashmap_entry e;
27 const char *diff, *patch;
28
29 int i, shown;
30 int diffsize;
31 size_t diff_offset;
32 /* the index of the matching item in the other branch, or -1 */
33 int matching;
34 struct object_id oid;
35 };
36
37 /*
38 * Reads the patches into a string list, with the `util` field being populated
39 * as struct object_id (will need to be free()d).
40 */
41 static int read_patches(const char *range, struct string_list *list,
42 const struct strvec *log_arg,
43 unsigned int include_merges)
44 {
45 struct child_process cp = CHILD_PROCESS_INIT;
46 struct strbuf buf = STRBUF_INIT, contents = STRBUF_INIT;
47 struct patch_util *util = NULL;
48 int in_header = 1;
49 char *line, *current_filename = NULL;
50 ssize_t len;
51 size_t size;
52 int ret = -1;
53
54 strvec_pushl(&cp.args, "log", "--no-color", "-p",
55 /*
56 * The patches being compared must be the builtin
57 * diff's: an external diff command or diff process
58 * could change either side of the comparison.
59 */
60 "--no-ext-diff",
61 "--reverse", "--date-order", "--decorate=no",
62 "--no-prefix", "--submodule=short",
63 /*
64 * Choose indicators that are not used anywhere
65 * else in diffs, but still look reasonable
66 * (e.g. will not be confusing when debugging)
67 */
68 "--output-indicator-new=>",
69 "--output-indicator-old=<",
70 "--output-indicator-context=#",
71 "--no-abbrev-commit",
72 "--pretty=medium",
73 "--show-notes-by-default",
74 NULL);
75 if (!include_merges)
76 strvec_push(&cp.args, "--no-merges");
77 strvec_push(&cp.args, range);
78 if (log_arg)
79 strvec_pushv(&cp.args, log_arg->v);
80 cp.out = -1;
81 cp.no_stdin = 1;
82 cp.git_cmd = 1;
83
84 if (start_command(&cp))
85 return error_errno(_("could not start `log`"));
86 if (strbuf_read(&contents, cp.out, 0) < 0) {
87 error_errno(_("could not read `log` output"));
88 finish_command(&cp);
89 goto cleanup;
90 }
91 if (finish_command(&cp))
92 goto cleanup;
93
94 line = contents.buf;
95 size = contents.len;
96 for (; size > 0; size -= len, line += len) {
97 char *p;
98 char *eol;
99
100 eol = memchr(line, '\n', size);
101 if (eol) {
102 *eol = '\0';
103 len = eol + 1 - line;
104 } else {
105 len = size;
106 }
107
108 if (skip_prefix(line, "commit ", &p)) {
109 char *q;
110 if (util) {
111 string_list_append(list, buf.buf)->util = util;
112 strbuf_reset(&buf);
113 }
114 CALLOC_ARRAY(util, 1);
115 if (include_merges && (q = strstr(p, " (from ")))
116 *q = '\0';
117 if (repo_get_oid(the_repository, p, &util->oid)) {
118 error(_("could not parse commit '%s'"), p);
119 FREE_AND_NULL(util);
120 string_list_clear(list, 1);
121 goto cleanup;
122 }
123 util->matching = -1;
124 in_header = 1;
125 continue;
126 }
127
128 if (!util) {
129 error(_("could not parse first line of `log` output: "
130 "did not start with 'commit ': '%s'"),
131 line);
132 string_list_clear(list, 1);
133 goto cleanup;
134 }
135
136 if (starts_with(line, "diff --git")) {
137 struct patch patch = { 0 };
138 struct strbuf root = STRBUF_INIT;
139 int linenr = 0;
140 int orig_len;
141
142 in_header = 0;
143 strbuf_addch(&buf, '\n');
144 if (!util->diff_offset)
145 util->diff_offset = buf.len;
146 if (eol)
147 *eol = '\n';
148 orig_len = len;
149 len = parse_git_diff_header(&root, NULL, &linenr, 0, line,
150 len, size, &patch);
151 if (len < 0) {
152 error(_("could not parse git header '%.*s'"),
153 orig_len, line);
154 FREE_AND_NULL(util);
155 string_list_clear(list, 1);
156 goto cleanup;
157 }
158 strbuf_addstr(&buf, " ## ");
159 if (patch.is_new > 0)
160 strbuf_addf(&buf, "%s (new)", patch.new_name);
161 else if (patch.is_delete > 0)
162 strbuf_addf(&buf, "%s (deleted)", patch.old_name);
163 else if (patch.is_rename)
164 strbuf_addf(&buf, "%s => %s", patch.old_name, patch.new_name);
165 else
166 strbuf_addstr(&buf, patch.new_name);
167
168 free(current_filename);
169 if (patch.is_delete > 0)
170 current_filename = xstrdup(patch.old_name);
171 else
172 current_filename = xstrdup(patch.new_name);
173
174 if (patch.new_mode && patch.old_mode &&
175 patch.old_mode != patch.new_mode)
176 strbuf_addf(&buf, " (mode change %06o => %06o)",
177 patch.old_mode, patch.new_mode);
178
179 strbuf_addstr(&buf, " ##");
180 release_patch(&patch);
181 } else if (in_header) {
182 if (starts_with(line, "Author: ")) {
183 strbuf_addstr(&buf, " ## Metadata ##\n");
184 strbuf_addstr(&buf, line);
185 strbuf_addstr(&buf, "\n\n");
186 strbuf_addstr(&buf, " ## Commit message ##\n");
187 } else if (starts_with(line, "Notes") &&
188 line[strlen(line) - 1] == ':') {
189 strbuf_addstr(&buf, "\n\n");
190 /* strip the trailing colon */
191 strbuf_addf(&buf, " ## %.*s ##\n",
192 (int)(strlen(line) - 1), line);
193 } else if (starts_with(line, " ")) {
194 p = line + len - 2;
195 while (isspace(*p) && p >= line)
196 p--;
197 strbuf_add(&buf, line, p - line + 1);
198 strbuf_addch(&buf, '\n');
199 }
200 continue;
201 } else if (skip_prefix(line, "@@ ", &p)) {
202 p = strstr(p, "@@");
203 strbuf_addstr(&buf, "@@");
204 if (current_filename && p[2])
205 strbuf_addf(&buf, " %s:", current_filename);
206 if (p)
207 strbuf_addstr(&buf, p + 2);
208 } else if (!line[0])
209 /*
210 * A completely blank (not ' \n', which is context)
211 * line is not valid in a diff. We skip it
212 * silently, because this neatly handles the blank
213 * separator line between commits in git-log
214 * output.
215 */
216 continue;
217 else if (line[0] == '>') {
218 strbuf_addch(&buf, '+');
219 strbuf_addstr(&buf, line + 1);
220 } else if (line[0] == '<') {
221 strbuf_addch(&buf, '-');
222 strbuf_addstr(&buf, line + 1);
223 } else if (line[0] == '#') {
224 strbuf_addch(&buf, ' ');
225 strbuf_addstr(&buf, line + 1);
226 } else {
227 strbuf_addch(&buf, ' ');
228 strbuf_addstr(&buf, line);
229 }
230
231 strbuf_addch(&buf, '\n');
232 util->diffsize++;
233 }
234
235 ret = 0;
236 cleanup:
237 strbuf_release(&contents);
238
239 if (util)
240 string_list_append(list, buf.buf)->util = util;
241 strbuf_release(&buf);
242 free(current_filename);
243
244 return ret;
245 }
246
247 static int patch_util_cmp(const void *cmp_data UNUSED,
248 const struct hashmap_entry *ha,
249 const struct hashmap_entry *hb,
250 const void *keydata)
251 {
252 const struct patch_util
253 *a = container_of(ha, const struct patch_util, e),
254 *b = container_of(hb, const struct patch_util, e);
255 return strcmp(a->diff, keydata ? keydata : b->diff);
256 }
257
258 static void find_exact_matches(struct string_list *a, struct string_list *b)
259 {
260 struct hashmap map = HASHMAP_INIT(patch_util_cmp, NULL);
261 int i;
262
263 /* First, add the patches of a to a hash map */
264 for (i = 0; i < a->nr; i++) {
265 struct patch_util *util = a->items[i].util;
266
267 util->i = i;
268 util->patch = a->items[i].string;
269 util->diff = util->patch + util->diff_offset;
270 hashmap_entry_init(&util->e, strhash(util->diff));
271 hashmap_add(&map, &util->e);
272 }
273
274 /* Now try to find exact matches in b */
275 for (i = 0; i < b->nr; i++) {
276 struct patch_util *util = b->items[i].util, *other;
277
278 util->i = i;
279 util->patch = b->items[i].string;
280 util->diff = util->patch + util->diff_offset;
281 hashmap_entry_init(&util->e, strhash(util->diff));
282 other = hashmap_remove_entry(&map, util, e, NULL);
283 if (other) {
284 if (other->matching >= 0)
285 BUG("already assigned!");
286
287 other->matching = i;
288 util->matching = other->i;
289 }
290 }
291
292 hashmap_clear(&map);
293 }
294
295 static int diffsize_consume(void *data,
296 char *line UNUSED,
297 unsigned long len UNUSED)
298 {
299 (*(int *)data)++;
300 return 0;
301 }
302
303 static void diffsize_hunk(void *data,
304 long ob UNUSED, long on UNUSED,
305 long nb UNUSED, long nn UNUSED,
306 const char *func UNUSED, long funclen UNUSED)
307 {
308 diffsize_consume(data, NULL, 0);
309 }
310
311 static int diffsize(const char *a, const char *b)
312 {
313 xpparam_t pp = { 0 };
314 xdemitconf_t cfg = { 0 };
315 mmfile_t mf1, mf2;
316 int count = 0;
317
318 mf1.ptr = (char *)a;
319 mf1.size = strlen(a);
320 mf2.ptr = (char *)b;
321 mf2.size = strlen(b);
322
323 cfg.ctxlen = 3;
324 if (!xdi_diff_outf(&mf1, &mf2,
325 diffsize_hunk, diffsize_consume, &count,
326 &pp, &cfg))
327 return count;
328
329 error(_("failed to generate diff"));
330 return COST_MAX;
331 }
332
333 static void get_correspondences(struct string_list *a, struct string_list *b,
334 int creation_factor, size_t max_memory)
335 {
336 int n = a->nr + b->nr;
337 int *cost, c, *a2b, *b2a;
338 int i, j;
339 size_t cost_size = st_mult(n, n);
340 size_t cost_bytes = st_mult(sizeof(int), cost_size);
341 if (cost_bytes >= max_memory) {
342 struct strbuf cost_str = STRBUF_INIT;
343 struct strbuf max_str = STRBUF_INIT;
344 strbuf_humanise_bytes(&cost_str, cost_bytes);
345 strbuf_humanise_bytes(&max_str, max_memory);
346 die(_("range-diff: unable to compute the range-diff, since it "
347 "exceeds the maximum memory for the cost matrix: %s "
348 "(%"PRIuMAX" bytes) needed, limited to %s (%"PRIuMAX" bytes)"),
349 cost_str.buf, (uintmax_t)cost_bytes, max_str.buf, (uintmax_t)max_memory);
350 }
351 ALLOC_ARRAY(cost, cost_size);
352 ALLOC_ARRAY(a2b, n);
353 ALLOC_ARRAY(b2a, n);
354
355 for (i = 0; i < a->nr; i++) {
356 struct patch_util *a_util = a->items[i].util;
357
358 for (j = 0; j < b->nr; j++) {
359 struct patch_util *b_util = b->items[j].util;
360
361 if (a_util->matching == j)
362 c = 0;
363 else if (a_util->matching < 0 && b_util->matching < 0)
364 c = diffsize(a_util->diff, b_util->diff);
365 else
366 c = COST_MAX;
367 cost[i + n * j] = c;
368 }
369
370 c = a_util->matching < 0 ?
371 a_util->diffsize * creation_factor / 100 : COST_MAX;
372 for (j = b->nr; j < n; j++)
373 cost[i + n * j] = c;
374 }
375
376 for (j = 0; j < b->nr; j++) {
377 struct patch_util *util = b->items[j].util;
378
379 c = util->matching < 0 ?
380 util->diffsize * creation_factor / 100 : COST_MAX;
381 for (i = a->nr; i < n; i++)
382 cost[i + n * j] = c;
383 }
384
385 for (i = a->nr; i < n; i++)
386 for (j = b->nr; j < n; j++)
387 cost[i + n * j] = 0;
388
389 compute_assignment(n, n, cost, a2b, b2a);
390
391 for (i = 0; i < a->nr; i++)
392 if (a2b[i] >= 0 && a2b[i] < b->nr) {
393 struct patch_util *a_util = a->items[i].util;
394 struct patch_util *b_util = b->items[a2b[i]].util;
395
396 a_util->matching = a2b[i];
397 b_util->matching = i;
398 }
399
400 free(cost);
401 free(a2b);
402 free(b2a);
403 }
404
405 static void output_pair_header(struct diff_options *diffopt,
406 int patch_no_width,
407 struct strbuf *buf,
408 struct strbuf *dashes,
409 struct patch_util *a_util,
410 struct patch_util *b_util)
411 {
412 struct object_id *oid = a_util ? &a_util->oid : &b_util->oid;
413 struct commit *commit;
414 char status;
415 const char *color_reset = diff_get_color_opt(diffopt, DIFF_RESET);
416 const char *color_old = diff_get_color_opt(diffopt, DIFF_FILE_OLD);
417 const char *color_new = diff_get_color_opt(diffopt, DIFF_FILE_NEW);
418 const char *color_commit = diff_get_color_opt(diffopt, DIFF_COMMIT);
419 const char *color;
420 int abbrev = diffopt->abbrev;
421
422 if (abbrev < 0)
423 abbrev = DEFAULT_ABBREV;
424
425 if (!dashes->len)
426 strbuf_addchars(dashes, '-',
427 strlen(repo_find_unique_abbrev(the_repository, oid, abbrev)));
428
429 if (!b_util) {
430 color = color_old;
431 status = '<';
432 } else if (!a_util) {
433 color = color_new;
434 status = '>';
435 } else if (strcmp(a_util->patch, b_util->patch)) {
436 color = color_commit;
437 status = '!';
438 } else {
439 color = color_commit;
440 status = '=';
441 }
442
443 strbuf_reset(buf);
444 strbuf_addstr(buf, status == '!' ? color_old : color);
445 if (!a_util)
446 strbuf_addf(buf, "%*s: %s ", patch_no_width, "-", dashes->buf);
447 else
448 strbuf_addf(buf, "%*d: %s ", patch_no_width, a_util->i + 1,
449 repo_find_unique_abbrev(the_repository, &a_util->oid, abbrev));
450
451 if (status == '!')
452 strbuf_addf(buf, "%s%s", color_reset, color);
453 strbuf_addch(buf, status);
454 if (status == '!')
455 strbuf_addf(buf, "%s%s", color_reset, color_new);
456
457 if (!b_util)
458 strbuf_addf(buf, " %*s: %s", patch_no_width, "-", dashes->buf);
459 else
460 strbuf_addf(buf, " %*d: %s", patch_no_width, b_util->i + 1,
461 repo_find_unique_abbrev(the_repository, &b_util->oid, abbrev));
462
463 commit = lookup_commit_reference(the_repository, oid);
464 if (commit) {
465 if (status == '!')
466 strbuf_addf(buf, "%s%s", color_reset, color);
467
468 strbuf_addch(buf, ' ');
469 pp_commit_easy(CMIT_FMT_ONELINE, commit, buf);
470 }
471 strbuf_addf(buf, "%s\n", color_reset);
472
473 fwrite(buf->buf, buf->len, 1, diffopt->file);
474 }
475
476 static struct userdiff_driver section_headers = {
477 .funcname = {
478 .pattern = "^ ## (.*) ##$\n^.?@@ (.*)$",
479 .cflags = REG_EXTENDED,
480 },
481 };
482
483 static struct diff_filespec *get_filespec(const char *name, const char *p)
484 {
485 struct diff_filespec *spec = alloc_filespec(name);
486
487 fill_filespec(spec, null_oid(the_hash_algo), 0, 0100644);
488 spec->data = (char *)p;
489 spec->size = strlen(p);
490 spec->should_munmap = 0;
491 spec->is_stdin = 1;
492 spec->driver = &section_headers;
493
494 return spec;
495 }
496
497 static void patch_diff(const char *a, const char *b,
498 struct diff_options *diffopt)
499 {
500 diff_queue(&diff_queued_diff,
501 get_filespec("a", a), get_filespec("b", b));
502
503 diffcore_std(diffopt);
504 diff_flush(diffopt);
505 }
506
507 static const char *output_prefix_cb(struct diff_options *opt UNUSED, void *data)
508 {
509 return data;
510 }
511
512 static void output(struct string_list *a, struct string_list *b,
513 struct range_diff_options *range_diff_opts)
514 {
515 struct strbuf buf = STRBUF_INIT, dashes = STRBUF_INIT;
516 int patch_no_width = decimal_width(1 + (a->nr > b->nr ? a->nr : b->nr));
517 int i = 0, j = 0;
518 struct diff_options opts;
519 struct strbuf indent = STRBUF_INIT;
520
521 if (range_diff_opts->diffopt)
522 memcpy(&opts, range_diff_opts->diffopt, sizeof(opts));
523 else
524 repo_diff_setup(the_repository, &opts);
525
526 opts.no_free = 1;
527 if (!opts.output_format)
528 opts.output_format = DIFF_FORMAT_PATCH;
529 opts.flags.suppress_diff_headers = 1;
530 opts.flags.dual_color_diffed_diffs =
531 range_diff_opts->dual_color;
532 opts.flags.suppress_hunk_header_line_count = 1;
533 opts.output_prefix = output_prefix_cb;
534 strbuf_addstr(&indent, " ");
535 opts.output_prefix_data = indent.buf;
536 diff_setup_done(&opts);
537
538 /*
539 * We assume the user is really more interested in the second argument
540 * ("newer" version). To that end, we print the output in the order of
541 * the RHS (the `b` parameter). To put the LHS (the `a` parameter)
542 * commits that are no longer in the RHS into a good place, we place
543 * them once we have shown all of their predecessors in the LHS.
544 */
545
546 while (i < a->nr || j < b->nr) {
547 struct patch_util *a_util, *b_util;
548 a_util = i < a->nr ? a->items[i].util : NULL;
549 b_util = j < b->nr ? b->items[j].util : NULL;
550
551 /* Skip all the already-shown commits from the LHS. */
552 while (i < a->nr && a_util->shown)
553 a_util = ++i < a->nr ? a->items[i].util : NULL;
554
555 /* Show unmatched LHS commit whose predecessors were shown. */
556 if (i < a->nr && a_util->matching < 0) {
557 if (!range_diff_opts->right_only)
558 output_pair_header(&opts, patch_no_width,
559 &buf, &dashes, a_util, NULL);
560 i++;
561 continue;
562 }
563
564 /* Show unmatched RHS commits. */
565 while (j < b->nr && b_util->matching < 0) {
566 if (!range_diff_opts->left_only)
567 output_pair_header(&opts, patch_no_width,
568 &buf, &dashes, NULL, b_util);
569 b_util = ++j < b->nr ? b->items[j].util : NULL;
570 }
571
572 /* Show matching LHS/RHS pair. */
573 if (j < b->nr) {
574 a_util = a->items[b_util->matching].util;
575 output_pair_header(&opts, patch_no_width,
576 &buf, &dashes, a_util, b_util);
577 if (!(opts.output_format & DIFF_FORMAT_NO_OUTPUT))
578 patch_diff(a->items[b_util->matching].string,
579 b->items[j].string, &opts);
580 a_util->shown = 1;
581 j++;
582 }
583 }
584 strbuf_release(&buf);
585 strbuf_release(&dashes);
586 strbuf_release(&indent);
587 opts.no_free = 0;
588 diff_free(&opts);
589 }
590
591 int show_range_diff(const char *range1, const char *range2,
592 struct range_diff_options *range_diff_opts)
593 {
594 int res = 0;
595
596 struct string_list branch1 = STRING_LIST_INIT_DUP;
597 struct string_list branch2 = STRING_LIST_INIT_DUP;
598 unsigned int include_merges = range_diff_opts->include_merges;
599
600 if (range_diff_opts->left_only && range_diff_opts->right_only)
601 res = error(_("options '%s' and '%s' cannot be used together"), "--left-only", "--right-only");
602
603 if (!res && read_patches(range1, &branch1, range_diff_opts->log_arg, include_merges))
604 res = error(_("could not parse log for '%s'"), range1);
605 if (!res && read_patches(range2, &branch2, range_diff_opts->log_arg, include_merges))
606 res = error(_("could not parse log for '%s'"), range2);
607
608 if (!res) {
609 find_exact_matches(&branch1, &branch2);
610 get_correspondences(&branch1, &branch2,
611 range_diff_opts->creation_factor,
612 range_diff_opts->max_memory);
613 output(&branch1, &branch2, range_diff_opts);
614 }
615
616 string_list_clear(&branch1, 1);
617 string_list_clear(&branch2, 1);
618
619 return res;
620 }
621
622 int is_range_diff_range(const char *arg)
623 {
624 char *copy = xstrdup(arg); /* setup_revisions() modifies it */
625 const char *argv[] = { "", copy, "--", NULL };
626 int i, positive = 0, negative = 0;
627 struct rev_info revs;
628
629 repo_init_revisions(the_repository, &revs, NULL);
630 if (setup_revisions(3, argv, &revs, NULL) == 1) {
631 for (i = 0; i < revs.pending.nr; i++)
632 if (revs.pending.objects[i].item->flags & UNINTERESTING)
633 negative++;
634 else
635 positive++;
636 for (i = 0; i < revs.pending.nr; i++) {
637 struct object *obj = revs.pending.objects[i].item;
638
639 if (obj->type == OBJ_COMMIT)
640 clear_commit_marks((struct commit *)obj,
641 ALL_REV_FLAGS);
642 }
643 }
644
645 free(copy);
646 release_revisions(&revs);
647 return negative > 0 && positive > 0;
648 }