2 * wiggle - apply rejected patches
4 * Copyright (C) 2005 Neil Brown <neilb@cse.unsw.edu.au>
5 * Copyright (C) 2010-2013 Neil Brown <neilb@suse.de>
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program.
22 * Email: <neilb@suse.de>
29 * Second attempt at merging....
31 * We want to create a mergelist which identifies 'orig' and 'after'
32 * sections (from a and c) and conflicts (which are ranges of a,b,c which
34 * It is also helpful to differentiate 'orig' sections that aren't
35 * matched in 'b' with orig sections that are.
36 * To help with highlighting, it will be useful to know where
37 * the conflicts match the csl lists.
39 * This can all be achieved with a list of (a,b,c,c1,c1) 5-tuples.
40 * If two consecutive differ in more than one of a,b,c, it is a
42 * If only 'a' differ, it is un-matched original.
43 * If only 'b' differ, it is matched, unchanged original
44 * If only 'c' differ, it is 1
47 static inline int min(int a, int b)
52 static int check_alreadyapplied(struct file af, struct file cf,
58 for (i = 0; i < m->al; i++) {
59 if (af.list[m->a+i].len != cf.list[m->c+i].len)
61 if (strncmp(af.list[m->a+i].start,
62 cf.list[m->c+i].start,
63 af.list[m->a+i].len) != 0)
67 printf("already applied %d,%d,%d - %d,%d,%d\n",
68 m->a, m->b, m->c, m->al, m->bl, m->cl);
69 printf(" %.10s - %.10s\n", af.list[m->a].start,
72 m->type = AlreadyApplied;
76 /* A 'cut-point' is a location in the merger where it is reasonable
77 * the change the mode of display - between displaying the merger
78 * and displaying the separate streams.
79 * A 'conflict' can only be displayed as separate stream so when
80 * one is found, we need to find a preceding and trailing cut-point
81 * and enlarge the conflict to that range.
82 * A suitable location is one where all three streams are at a line-end.
84 static int is_cutpoint(struct merge m,
85 struct file af, struct file bf, struct file cf)
87 return ((m.a == 0 || ends_line(af.list[m.a-1])) &&
88 (m.b == 0 || ends_line(bf.list[m.b-1])) &&
89 (m.c == 0 || ends_line(cf.list[m.c-1])));
92 int isolate_conflicts(struct file af, struct file bf, struct file cf,
93 struct csl *csl1, struct csl *csl2, int words,
94 struct merge *m, int show_wiggles,
97 /* A Conflict indicates that something is definitely wrong
98 * and so we need to be a bit suspicious of nearby apparent matches.
99 * To display a conflict effectively we expands its effect to
100 * include any Extraneous, Unmatched, Changed or AlreadyApplied text.
101 * Also, unless 'words', we need to include any partial lines
102 * in the Unchanged text that forms the border of a conflict.
104 * A Changed text may also border a conflict, but it can
105 * only border one conflict (where as an Unchanged can border
106 * a preceding and a following conflict).
107 * The 'new' section of a Changed text appears in the
108 * conflict as does any part of the original before
111 * A hunk header (Extraneous) is never considered part of a
112 * conflict. It thereby can serve as a separator between
115 * Extended conflicts are marked by setting ->in_conflict in
116 * the "struct merge". This is '1' for an Unchanged, Changed,
117 * or (Extraneous) hunk header which borders the conflict,
118 * '2' for a merger which is truly in conflict, and '3' for
119 * a merger which is causing a 'wiggle'.
120 * When in_conflict == 1, the 'lo' and 'hi' fields indicate
121 * how much of the 'a' file is included in the conflict, the rest
122 * being part of the clean result.
123 * Elements in af from m->a to m->a+m->lo are in the preceding
124 * conflict, from m->a+m->lo to m->a+m->hi are clean, and
125 * m->a+m->hi to m->a+m->al are in the following conflict.
127 * We need to ensure there is adequate context for the conflict.
128 * So ensure there are at least 3 newlines in Extraneous or
129 * Unchanged on both sides of a Conflict - but don't go so far
130 * as including a hunk header.
131 * If there are 3, and they are all in 'Unchanged' sections, then
132 * that much context is not really needed - reduce it a bit.
134 * If a wiggle is adjacent to a conflict then:
135 * - if show_wiggles is set, we just merge them
136 * - if it is not set, we still want to count the wiggle.
139 int cnt = 0, wiggles = 0;
145 for (i = 0; m[i].type != End; i++)
146 m[i].in_conflict = 0;
148 for (i = 0; m[i].type != End; i++) {
149 /* The '3' here is a count of newlines. Once we find
150 * that many newlines of the particular type, we have escaped.
152 if (m[i].type == Changed)
154 if (m[i].type == Unmatched)
156 if (m[i].type == Extraneous && bf.list[m[i].b].start[0])
157 /* hunk headers don't imply wiggles, other
158 * extraneous text does.
162 if (m[i].type != Unchanged && changed && (unmatched || extraneous)) {
169 if ((m[i].type == Conflict) ||
170 (show_wiggles && in_wiggle)) {
171 /* We have a conflict or wiggle here.
172 * First search backwards for an Unchanged marking
173 * things as in_conflict. Then find the
174 * cut-point in the Unchanged. If there isn't one,
177 * Then search forward doing the same thing.
180 m[i].in_conflict = m[i].type == Conflict ? 2 : 3;
185 if (m[j].type == Extraneous &&
186 bf.list[m[j].b].start[0] == '\0')
187 /* hunk header - not conflict any more */
189 if (m[j].in_conflict > 1)
190 /* Merge the conflicts */
192 if (!m[j].in_conflict) {
193 m[j].in_conflict = 1;
196 /* Following must set m[j].hi, or set
199 if (m[j].type == Extraneous) {
200 for (k = m[j].bl; k > 0; k--)
201 if (ends_line(bf.list[m[j].b+k-1]))
205 if (m[j].type != Unchanged &&
206 m[j].type != Changed) {
207 if (m[j].type == Conflict)
208 m[j].in_conflict = 2;
210 m[j].in_conflict = m[i].in_conflict;
213 /* If we find enough newlines in this section,
214 * then we only really need 1, but would rather
215 * it wasn't the first one. 'firstk' allows us
216 * to track which newline we actually use
223 /* need to find the last line-break, which
224 * might be after the last newline, if there
225 * is one, or might be at the start
227 for (k = m[j].al; k > 0; k--) {
228 if (m[j].a + k >= af.elcnt)
229 /* FIXME impossible!*/
231 if (ends_line(af.list[m[j].a+k-1])) {
232 if (firstk > m[j].al)
245 else if (is_cutpoint(m[j], af,bf,cf))
248 /* no start-of-line found... */
251 (m[j].type == Changed)) {
252 /* this can only work if start is
253 * also a line break */
254 if (is_cutpoint(m[j], af,bf,cf))
261 m[j].in_conflict = m[i].in_conflict;
264 /* now the forward search */
266 for (j = i+1; m[j].type != End; j++) {
267 if (m[j].type == Extraneous &&
268 bf.list[m[j].b].start[0] == '\0')
269 /* hunk header - not conflict any more */
271 if (m[j].type == Extraneous) {
272 for (k = 0; k < m[j].bl; k++)
273 if (ends_line(bf.list[m[j].b+k]))
276 if (m[j].type != Unchanged &&
277 m[j].type != Changed) {
278 if (m[j].type == Conflict)
279 m[j].in_conflict = 2;
281 m[j].in_conflict = m[i].in_conflict;
284 m[j].in_conflict = 1;
290 /* need to find a line-break, which might be at
291 * the very beginning, or might be after the
292 * first newline - if there is one
294 if (is_cutpoint(m[j], af,bf,cf))
297 /* If we find enough newlines in this section,
298 * then we only really need 1, but would rather
299 * it wasn't the first one. 'firstk' allows us
300 * to track which newline we actually use
303 for (k = 0 ; k < m[j].al ; k++)
304 if (ends_line(af.list[m[j].a+k])) {
315 /* Hit end of file, pretend we found 3 newlines. */
319 m[j+1].type == Unmatched) {
320 /* If this Unmatched exceeds 3 lines, just stop here */
323 for (p = 0; p < m[j+1].al ; p++)
324 if (ends_line(af.list[m[j+1].a+p])) {
335 /* no start-of-line found */
338 if (m[j].lo <= m[j].al+1 &&
339 (m[j].type == Changed)) {
340 /* this can only work if the end is a line break */
341 if (is_cutpoint(m[j+1], af,bf,cf))
346 if (m[j].lo < m[j].al+1)
348 m[j].in_conflict = m[i].in_conflict;
350 if (m[j-1].in_conflict == 1)
353 /* A hunk header bordered the conflict */
356 /* If any of the merges are Changed or Conflict,
357 * then this really is a Conflict or Wiggle.
358 * If not they are just Unchanged, Unmatched,
359 * Extraneous or AlreadyApplied, and so don't
361 * Note that the first/last merges (in_conflict==1)
362 * can be Changed and so much be check separately.
364 if (m[j].type == Changed)
366 for (j = i-1; j >= 0 && m[j].in_conflict > 1; j--)
367 if (m[j].type == Changed || m[j].type == Conflict)
369 if (j >= 0 && m[j].type == Changed)
371 /* False alarm, no real conflict/wiggle here as
372 * nothing changed. */
375 if (m[j].in_conflict == 1) {
378 m[j].in_conflict = 0;
382 m[j++].in_conflict = 0;
384 if (m[i].type == End)
387 for (k = 1; k < m[i].al; k++) {
388 if (m[i].a + k >= af.elcnt)
389 /* FIXME this should be impossible, but
393 if (words || ends_line(af.list[m[i].a+k])) {
405 /* Now count the conflicts and wiggles */
406 for (i = 0; m[i].type != End; i++) {
407 int true_conflict = 0;
408 if (!m[i].in_conflict)
411 for (j = i; m[j].type != End && m[j].in_conflict; j++) {
412 if (m[j].in_conflict == 2)
415 m[j].in_conflict == 1) {
417 if (!m[j+1].in_conflict)
433 struct ci make_merger(struct file af, struct file bf, struct file cf,
434 struct csl *csl1, struct csl *csl2, int words,
435 int ignore_already, int show_wiggles)
437 /* find the wiggles and conflicts between csl1 and csl2
442 int header_checked = -1;
443 int header_found = 0;
445 rv.conflicts = rv.wiggles = rv.ignored = 0;
447 for (i = 0; csl1[i].len; i++)
450 for (i = 0; csl2[i].len; i++)
453 /* maybe a bit of slack at each end */
456 rv.merger = xmalloc(sizeof(struct merge)*l);
458 a = b = c = c1 = c2 = 0;
462 match1 = (a >= csl1[c1].a && b >= csl1[c1].b); /* c1 doesn't match */
463 match2 = (b >= csl2[c2].a && c >= csl2[c2].b);
465 if (header_checked != c2) {
466 /* Check if there is a hunk header in this range */
469 for (j = b; j < csl2[c2].a + csl2[c2].len; j++)
470 if (bf.list[j].start[0] == '\0') {
479 rv.merger[i].c1 = c1;
480 rv.merger[i].c2 = c2;
481 rv.merger[i].in_conflict = 0;
483 if (!match1 && match2) {
484 /* This is either Unmatched or Extraneous - probably both.
485 * If the match2 has a hunk-header Extraneous, it must
486 * align with an end-of-line in 'a', so adjust endpoint
488 int newa = csl1[c1].a;
489 if (header_found >= 0) {
491 !ends_line(af.list[newa-1]))
494 if (a == newa && b == csl1[c1].b)
497 /* some unmatched text */
498 rv.merger[i].type = Unmatched;
499 rv.merger[i].al = newa - a;
504 assert(b < csl1[c1].b);
505 /* some Extraneous text */
506 /* length is min of unmatched on left
507 * and matched on right.
508 * However a hunk-header must be an
509 * Extraneous section by itself, so if this
510 * start with one, the length is 1, and if
511 * there is one in the middle, only take the
512 * text up to there for now.
514 rv.merger[i].type = Extraneous;
518 csl2[c2].len - (b-csl2[c2].a));
519 if (header_found == b) {
522 } else if (header_found > b && header_found < newb) {
528 rv.merger[i].bl = newb - b;
530 } else if (match1 && !match2) {
532 * if 'c' is currently at a suitable cut-point, then
533 * we can look for a triple-cut-point for start.
534 * Also, if csl2[c2].b isn't in a conflict, and is
535 * a suitable cut-point, then we could make a
536 * triple-cut-point for end of a conflict.
539 rv.merger[i].type = Changed;
540 rv.merger[i].bl = min(csl1[c1].b+csl1[c1].len, csl2[c2].a) - b;
541 rv.merger[i].al = rv.merger[i].bl;
542 rv.merger[i].cl = csl2[c2].b - c;
543 } else if (match1 && match2) {
544 /* Some unchanged text
546 rv.merger[i].type = Unchanged;
548 min(csl1[c1].len - (b-csl1[c1].b),
549 csl2[c2].len - (b-csl2[c2].a));
550 rv.merger[i].al = rv.merger[i].cl =
553 /* must be a conflict.
554 * Move a and c to next match, and b to closest of the two
556 rv.merger[i].type = Conflict;
557 rv.merger[i].al = csl1[c1].a - a;
558 rv.merger[i].cl = csl2[c2].b - c;
559 rv.merger[i].bl = min(csl1[c1].b, csl2[c2].a) - b;
560 if (ignore_already &&
561 check_alreadyapplied(af, cf, &rv.merger[i]))
563 else if (rv.merger[i].bl == 0 &&
565 /* As the 'before' stream is empty, this
566 * could look like Unmatched in the
567 * original, and an insertion in the
568 * diff. Reporting it like that is
569 * probably more useful that as a full
571 * Leave the type for the insertion as
572 * Conflict (not Changed) as there is some
573 * real uncertainty here, but allow the
574 * original to become Unmatched.
578 rv.merger[i].oldtype = rv.merger[i].type;
579 a += rv.merger[i].al;
580 b += rv.merger[i].bl;
581 c += rv.merger[i].cl;
584 while (csl1[c1].a + csl1[c1].len <= a && csl1[c1].len)
586 assert(csl1[c1].b + csl1[c1].len >= b);
587 while (csl2[c2].b + csl2[c2].len <= c && csl2[c2].len)
589 assert(csl2[c2].a + csl2[c2].len >= b);
590 if (csl1[c1].len == 0 && csl2[c2].len == 0 &&
591 a == csl1[c1].a && b == csl1[c1].b &&
592 b == csl2[c2].a && c == csl2[c2].b)
595 rv.merger[i].type = End;
596 rv.merger[i].oldtype = End;
600 rv.merger[i].c1 = c1;
601 rv.merger[i].c2 = c2;
602 rv.merger[i].in_conflict = 0;
605 /* Now revert any AlreadyApplied that aren't bounded by
606 * Unchanged or Changed.
608 for (i = 0; rv.merger[i].type != End; i++) {
609 if (rv.merger[i].type != AlreadyApplied)
611 if (i > 0 && rv.merger[i-1].type != Unchanged &&
612 rv.merger[i-1].type != Changed)
613 rv.merger[i].type = Conflict;
614 if (rv.merger[i+1].type != Unchanged &&
615 rv.merger[i+1].type != Changed &&
616 rv.merger[i+1].type != End)
617 rv.merger[i].type = Conflict;
619 rv.conflicts = isolate_conflicts(af, bf, cf, csl1, csl2, words,
620 rv.merger, show_wiggles, &rv.wiggles);
624 static int printrange(FILE *out, struct file *f, int start, int len,
628 while (len > 0 && start < f->elcnt) {
629 struct elmnt e = f->list[start];
631 if (e.start[e.plen-1] == '\n' &&
641 int print_merge(FILE *out, struct file *a, struct file *b, struct file *c,
642 int words, struct merge *merger,
643 struct merge *mpos, int streampos, int offsetpos)
648 int offset = INT_MAX;
650 for (m = merger; m->type != End ; m++) {
653 printf("[%s: %d-%d,%d-%d,%d-%d%s(%d,%d)]\n",
654 m->type==Unmatched ? "Unmatched" :
655 m->type==Unchanged ? "Unchanged" :
656 m->type==Extraneous ? "Extraneous" :
657 m->type==Changed ? "Changed" :
658 m->type==AlreadyApplied ? "AlreadyApplied" :
659 m->type==Conflict ? "Conflict":"unknown",
663 m->in_conflict ? " in_conflict" : "",
666 while (m->in_conflict) {
667 /* need to print from 'hi' to 'lo' of next
668 * Unchanged which is < it's hi
670 int found_conflict = 0;
672 if (m->in_conflict == 1)
678 if (m->in_conflict == 1 && m->type == Unchanged)
679 lineno += printrange(out, a, m->a+m->lo, m->hi - m->lo, offset - m->lo);
684 for (cm = m; cm->in_conflict; cm++) {
685 printf("{%s: %d-%d,%d-%d,%d-%d%s(%d,%d)}\n",
686 cm->type==Unmatched?"Unmatched":
687 cm->type==Unchanged?"Unchanged":
688 cm->type==Extraneous?"Extraneous":
689 cm->type==Changed?"Changed":
690 cm->type==AlreadyApplied?"AlreadyApplied":
691 cm->type==Conflict?"Conflict":"unknown",
692 cm->a, cm->a+cm->al-1,
693 cm->b, cm->b+cm->bl-1,
694 cm->c, cm->c+cm->cl-1,
695 cm->in_conflict ? " in_conflict" : "",
697 if (cm->in_conflict == 1 && cm != m)
701 if (m->in_conflict == 1 &&
702 m[1].in_conflict == 1) {
703 /* Nothing between two conflicts */
708 fputs(words ? "<<<---" : "<<<<<<< found\n", out);
711 for (cm = m; cm->in_conflict; cm++) {
712 if (cm == mpos && streampos == 0)
714 if (cm->type == Conflict)
716 if (cm->in_conflict == 1 && cm != m) {
717 lineno += printrange(out, a, cm->a, cm->lo, offset);
720 lineno += printrange(out, a, cm->a+st1, cm->al-st1, offset-st1);
722 if (cm == mpos && streampos == 0)
725 if (cm == mpos && streampos == 0)
727 fputs(words ? "|||" : "||||||| expected\n", out);
731 for (cm = m; cm->in_conflict; cm++) {
732 if (cm == mpos && streampos == 1)
734 if (cm->in_conflict == 1 && cm != m) {
735 lineno += printrange(out, a, cm->a, cm->lo, offset);
738 lineno += printrange(out, b, cm->b+st1, cm->bl-st1, offset-st1);
740 if (cm == mpos && streampos == 1)
743 if (cm == mpos && streampos == 1)
745 fputs(words ? "===" : "=======\n", out);
749 for (cm = m; cm->in_conflict; cm++) {
750 if (cm == mpos && streampos == 2)
752 if (cm->in_conflict == 1 && cm != m) {
753 if (cm->type == Unchanged)
754 lineno += printrange(out, a, cm->a, cm->lo, offset);
756 lineno += printrange(out, c, cm->c, cm->cl, offset);
759 if (cm->type == Changed)
760 st1 = 0; /* All of result of change must be printed */
761 lineno += printrange(out, c, cm->c+st1, cm->cl-st1, offset-st1);
763 if (cm == mpos && streampos == 2)
766 if (cm == mpos && streampos == 2)
768 if (!found_conflict) {
769 /* This section was wiggled in successfully,
770 * but full conflict display was requested.
771 * So now print out the wiggled result as well.
773 fputs(words ? "&&&" : "&&&&&&& resolution\n", out);
777 for (cm = m; cm->in_conflict; cm++) {
779 if (cm->in_conflict == 1 && cm != m)
785 lineno += printrange(out, a, cm->a+st1,
786 last ? cm->lo : cm->al-st1, offset-st1);
791 lineno += printrange(out, c, cm->c,
792 last ? cm->lo : cm->cl, offset);
803 fputs(words ? "--->>>" : ">>>>>>> replacement\n", out);
807 if (m->in_conflict == 1 && m[1].in_conflict == 0) {
808 /* End of a conflict, no conflict follows */
811 if (m->type == Unchanged)
812 lineno += printrange(out, a, m->a+m->lo, m->hi-m->lo, offset-m->lo);
819 /* there is always some non-conflict after a conflict,
820 * unless we hit the end
826 printf("<<%s: %d-%d,%d-%d,%d-%d%s(%d,%d)>>\n",
827 m->type==Unmatched?"Unmatched":
828 m->type==Unchanged?"Unchanged":
829 m->type==Extraneous?"Extraneous":
830 m->type==Changed?"Changed":
831 m->type==AlreadyApplied?"AlreadyApplied":
832 m->type==Conflict?"Conflict":"unknown",
836 m->in_conflict ? " in_conflict" : "",
846 lineno += printrange(out, a, m->a, m->al, offset);
851 lineno += printrange(out, c, m->c, m->cl, offset);
863 int save_merge(struct file a, struct file b, struct file c,
864 struct merge *merger, char *file, int backup)
866 char *replacename = xmalloc(strlen(file) + 20);
867 char *orignew = xmalloc(strlen(file) + 20);
872 strcpy(replacename, file);
873 strcat(replacename, "XXXXXX");
874 strcpy(orignew, file);
875 strcat(orignew, ".porig");
877 fd = mkstemp(replacename);
882 outfile = fdopen(fd, "w");
883 lineno = print_merge(outfile, &a, &b, &c, 0, merger,
886 if (backup && rename(file, orignew) != 0)
888 else if (rename(replacename, file) != 0)
894 return err < 0 ? err : lineno;
897 int save_tmp_merge(struct file a, struct file b, struct file c,
898 struct merge *merger, char **filep,
899 struct merge *mpos, int streampos, int offsetpos)
908 dir = getenv("TMPDIR");
912 asprintf(&fname, "%s/wiggle-tmp-XXXXXX", dir);
916 base = strrchr(dir, '/');
921 asprintf(&fname, "%.*stmp-XXXXXX-%s", (int)(base-dir), dir, base);
922 suffix = strlen(base)+1;
924 fd = mkstemps(fname, suffix);
931 outfile = fdopen(fd, "w");
932 lineno = print_merge(outfile, &a, &b, &c, 0, merger,
933 mpos, streampos, offsetpos);