]>
Commit | Line | Data |
---|---|---|
a2ad79ce CC |
1 | #include "cache.h" |
2 | #include "commit.h" | |
3 | #include "diff.h" | |
4 | #include "revision.h" | |
5 | #include "bisect.h" | |
6 | ||
95188648 CC |
7 | static unsigned char (*skipped_sha1)[20]; |
8 | static int skipped_sha1_nr; | |
9 | ||
a2ad79ce CC |
10 | /* bits #0-15 in revision.h */ |
11 | ||
12 | #define COUNTED (1u<<16) | |
13 | ||
14 | /* | |
15 | * This is a truly stupid algorithm, but it's only | |
16 | * used for bisection, and we just don't care enough. | |
17 | * | |
18 | * We care just barely enough to avoid recursing for | |
19 | * non-merge entries. | |
20 | */ | |
21 | static int count_distance(struct commit_list *entry) | |
22 | { | |
23 | int nr = 0; | |
24 | ||
25 | while (entry) { | |
26 | struct commit *commit = entry->item; | |
27 | struct commit_list *p; | |
28 | ||
29 | if (commit->object.flags & (UNINTERESTING | COUNTED)) | |
30 | break; | |
31 | if (!(commit->object.flags & TREESAME)) | |
32 | nr++; | |
33 | commit->object.flags |= COUNTED; | |
34 | p = commit->parents; | |
35 | entry = p; | |
36 | if (p) { | |
37 | p = p->next; | |
38 | while (p) { | |
39 | nr += count_distance(p); | |
40 | p = p->next; | |
41 | } | |
42 | } | |
43 | } | |
44 | ||
45 | return nr; | |
46 | } | |
47 | ||
48 | static void clear_distance(struct commit_list *list) | |
49 | { | |
50 | while (list) { | |
51 | struct commit *commit = list->item; | |
52 | commit->object.flags &= ~COUNTED; | |
53 | list = list->next; | |
54 | } | |
55 | } | |
56 | ||
57 | #define DEBUG_BISECT 0 | |
58 | ||
59 | static inline int weight(struct commit_list *elem) | |
60 | { | |
61 | return *((int*)(elem->item->util)); | |
62 | } | |
63 | ||
64 | static inline void weight_set(struct commit_list *elem, int weight) | |
65 | { | |
66 | *((int*)(elem->item->util)) = weight; | |
67 | } | |
68 | ||
69 | static int count_interesting_parents(struct commit *commit) | |
70 | { | |
71 | struct commit_list *p; | |
72 | int count; | |
73 | ||
74 | for (count = 0, p = commit->parents; p; p = p->next) { | |
75 | if (p->item->object.flags & UNINTERESTING) | |
76 | continue; | |
77 | count++; | |
78 | } | |
79 | return count; | |
80 | } | |
81 | ||
82 | static inline int halfway(struct commit_list *p, int nr) | |
83 | { | |
84 | /* | |
85 | * Don't short-cut something we are not going to return! | |
86 | */ | |
87 | if (p->item->object.flags & TREESAME) | |
88 | return 0; | |
89 | if (DEBUG_BISECT) | |
90 | return 0; | |
91 | /* | |
92 | * 2 and 3 are halfway of 5. | |
93 | * 3 is halfway of 6 but 2 and 4 are not. | |
94 | */ | |
95 | switch (2 * weight(p) - nr) { | |
96 | case -1: case 0: case 1: | |
97 | return 1; | |
98 | default: | |
99 | return 0; | |
100 | } | |
101 | } | |
102 | ||
103 | #if !DEBUG_BISECT | |
104 | #define show_list(a,b,c,d) do { ; } while (0) | |
105 | #else | |
106 | static void show_list(const char *debug, int counted, int nr, | |
107 | struct commit_list *list) | |
108 | { | |
109 | struct commit_list *p; | |
110 | ||
111 | fprintf(stderr, "%s (%d/%d)\n", debug, counted, nr); | |
112 | ||
113 | for (p = list; p; p = p->next) { | |
114 | struct commit_list *pp; | |
115 | struct commit *commit = p->item; | |
116 | unsigned flags = commit->object.flags; | |
117 | enum object_type type; | |
118 | unsigned long size; | |
119 | char *buf = read_sha1_file(commit->object.sha1, &type, &size); | |
120 | char *ep, *sp; | |
121 | ||
122 | fprintf(stderr, "%c%c%c ", | |
123 | (flags & TREESAME) ? ' ' : 'T', | |
124 | (flags & UNINTERESTING) ? 'U' : ' ', | |
125 | (flags & COUNTED) ? 'C' : ' '); | |
126 | if (commit->util) | |
127 | fprintf(stderr, "%3d", weight(p)); | |
128 | else | |
129 | fprintf(stderr, "---"); | |
130 | fprintf(stderr, " %.*s", 8, sha1_to_hex(commit->object.sha1)); | |
131 | for (pp = commit->parents; pp; pp = pp->next) | |
132 | fprintf(stderr, " %.*s", 8, | |
133 | sha1_to_hex(pp->item->object.sha1)); | |
134 | ||
135 | sp = strstr(buf, "\n\n"); | |
136 | if (sp) { | |
137 | sp += 2; | |
138 | for (ep = sp; *ep && *ep != '\n'; ep++) | |
139 | ; | |
140 | fprintf(stderr, " %.*s", (int)(ep - sp), sp); | |
141 | } | |
142 | fprintf(stderr, "\n"); | |
143 | } | |
144 | } | |
145 | #endif /* DEBUG_BISECT */ | |
146 | ||
147 | static struct commit_list *best_bisection(struct commit_list *list, int nr) | |
148 | { | |
149 | struct commit_list *p, *best; | |
150 | int best_distance = -1; | |
151 | ||
152 | best = list; | |
153 | for (p = list; p; p = p->next) { | |
154 | int distance; | |
155 | unsigned flags = p->item->object.flags; | |
156 | ||
157 | if (flags & TREESAME) | |
158 | continue; | |
159 | distance = weight(p); | |
160 | if (nr - distance < distance) | |
161 | distance = nr - distance; | |
162 | if (distance > best_distance) { | |
163 | best = p; | |
164 | best_distance = distance; | |
165 | } | |
166 | } | |
167 | ||
168 | return best; | |
169 | } | |
170 | ||
171 | struct commit_dist { | |
172 | struct commit *commit; | |
173 | int distance; | |
174 | }; | |
175 | ||
176 | static int compare_commit_dist(const void *a_, const void *b_) | |
177 | { | |
178 | struct commit_dist *a, *b; | |
179 | ||
180 | a = (struct commit_dist *)a_; | |
181 | b = (struct commit_dist *)b_; | |
182 | if (a->distance != b->distance) | |
183 | return b->distance - a->distance; /* desc sort */ | |
184 | return hashcmp(a->commit->object.sha1, b->commit->object.sha1); | |
185 | } | |
186 | ||
187 | static struct commit_list *best_bisection_sorted(struct commit_list *list, int nr) | |
188 | { | |
189 | struct commit_list *p; | |
190 | struct commit_dist *array = xcalloc(nr, sizeof(*array)); | |
191 | int cnt, i; | |
192 | ||
193 | for (p = list, cnt = 0; p; p = p->next) { | |
194 | int distance; | |
195 | unsigned flags = p->item->object.flags; | |
196 | ||
197 | if (flags & TREESAME) | |
198 | continue; | |
199 | distance = weight(p); | |
200 | if (nr - distance < distance) | |
201 | distance = nr - distance; | |
202 | array[cnt].commit = p->item; | |
203 | array[cnt].distance = distance; | |
204 | cnt++; | |
205 | } | |
206 | qsort(array, cnt, sizeof(*array), compare_commit_dist); | |
207 | for (p = list, i = 0; i < cnt; i++) { | |
208 | struct name_decoration *r = xmalloc(sizeof(*r) + 100); | |
209 | struct object *obj = &(array[i].commit->object); | |
210 | ||
211 | sprintf(r->name, "dist=%d", array[i].distance); | |
212 | r->next = add_decoration(&name_decoration, obj, r); | |
213 | p->item = array[i].commit; | |
214 | p = p->next; | |
215 | } | |
216 | if (p) | |
217 | p->next = NULL; | |
218 | free(array); | |
219 | return list; | |
220 | } | |
221 | ||
222 | /* | |
223 | * zero or positive weight is the number of interesting commits it can | |
224 | * reach, including itself. Especially, weight = 0 means it does not | |
225 | * reach any tree-changing commits (e.g. just above uninteresting one | |
226 | * but traversal is with pathspec). | |
227 | * | |
228 | * weight = -1 means it has one parent and its distance is yet to | |
229 | * be computed. | |
230 | * | |
231 | * weight = -2 means it has more than one parent and its distance is | |
232 | * unknown. After running count_distance() first, they will get zero | |
233 | * or positive distance. | |
234 | */ | |
235 | static struct commit_list *do_find_bisection(struct commit_list *list, | |
236 | int nr, int *weights, | |
237 | int find_all) | |
238 | { | |
239 | int n, counted; | |
240 | struct commit_list *p; | |
241 | ||
242 | counted = 0; | |
243 | ||
244 | for (n = 0, p = list; p; p = p->next) { | |
245 | struct commit *commit = p->item; | |
246 | unsigned flags = commit->object.flags; | |
247 | ||
248 | p->item->util = &weights[n++]; | |
249 | switch (count_interesting_parents(commit)) { | |
250 | case 0: | |
251 | if (!(flags & TREESAME)) { | |
252 | weight_set(p, 1); | |
253 | counted++; | |
254 | show_list("bisection 2 count one", | |
255 | counted, nr, list); | |
256 | } | |
257 | /* | |
258 | * otherwise, it is known not to reach any | |
259 | * tree-changing commit and gets weight 0. | |
260 | */ | |
261 | break; | |
262 | case 1: | |
263 | weight_set(p, -1); | |
264 | break; | |
265 | default: | |
266 | weight_set(p, -2); | |
267 | break; | |
268 | } | |
269 | } | |
270 | ||
271 | show_list("bisection 2 initialize", counted, nr, list); | |
272 | ||
273 | /* | |
274 | * If you have only one parent in the resulting set | |
275 | * then you can reach one commit more than that parent | |
276 | * can reach. So we do not have to run the expensive | |
277 | * count_distance() for single strand of pearls. | |
278 | * | |
279 | * However, if you have more than one parents, you cannot | |
280 | * just add their distance and one for yourself, since | |
281 | * they usually reach the same ancestor and you would | |
282 | * end up counting them twice that way. | |
283 | * | |
284 | * So we will first count distance of merges the usual | |
285 | * way, and then fill the blanks using cheaper algorithm. | |
286 | */ | |
287 | for (p = list; p; p = p->next) { | |
288 | if (p->item->object.flags & UNINTERESTING) | |
289 | continue; | |
290 | if (weight(p) != -2) | |
291 | continue; | |
292 | weight_set(p, count_distance(p)); | |
293 | clear_distance(list); | |
294 | ||
295 | /* Does it happen to be at exactly half-way? */ | |
296 | if (!find_all && halfway(p, nr)) | |
297 | return p; | |
298 | counted++; | |
299 | } | |
300 | ||
301 | show_list("bisection 2 count_distance", counted, nr, list); | |
302 | ||
303 | while (counted < nr) { | |
304 | for (p = list; p; p = p->next) { | |
305 | struct commit_list *q; | |
306 | unsigned flags = p->item->object.flags; | |
307 | ||
308 | if (0 <= weight(p)) | |
309 | continue; | |
310 | for (q = p->item->parents; q; q = q->next) { | |
311 | if (q->item->object.flags & UNINTERESTING) | |
312 | continue; | |
313 | if (0 <= weight(q)) | |
314 | break; | |
315 | } | |
316 | if (!q) | |
317 | continue; | |
318 | ||
319 | /* | |
320 | * weight for p is unknown but q is known. | |
321 | * add one for p itself if p is to be counted, | |
322 | * otherwise inherit it from q directly. | |
323 | */ | |
324 | if (!(flags & TREESAME)) { | |
325 | weight_set(p, weight(q)+1); | |
326 | counted++; | |
327 | show_list("bisection 2 count one", | |
328 | counted, nr, list); | |
329 | } | |
330 | else | |
331 | weight_set(p, weight(q)); | |
332 | ||
333 | /* Does it happen to be at exactly half-way? */ | |
334 | if (!find_all && halfway(p, nr)) | |
335 | return p; | |
336 | } | |
337 | } | |
338 | ||
339 | show_list("bisection 2 counted all", counted, nr, list); | |
340 | ||
341 | if (!find_all) | |
342 | return best_bisection(list, nr); | |
343 | else | |
344 | return best_bisection_sorted(list, nr); | |
345 | } | |
346 | ||
347 | struct commit_list *find_bisection(struct commit_list *list, | |
348 | int *reaches, int *all, | |
349 | int find_all) | |
350 | { | |
351 | int nr, on_list; | |
352 | struct commit_list *p, *best, *next, *last; | |
353 | int *weights; | |
354 | ||
355 | show_list("bisection 2 entry", 0, 0, list); | |
356 | ||
357 | /* | |
358 | * Count the number of total and tree-changing items on the | |
359 | * list, while reversing the list. | |
360 | */ | |
361 | for (nr = on_list = 0, last = NULL, p = list; | |
362 | p; | |
363 | p = next) { | |
364 | unsigned flags = p->item->object.flags; | |
365 | ||
366 | next = p->next; | |
367 | if (flags & UNINTERESTING) | |
368 | continue; | |
369 | p->next = last; | |
370 | last = p; | |
371 | if (!(flags & TREESAME)) | |
372 | nr++; | |
373 | on_list++; | |
374 | } | |
375 | list = last; | |
376 | show_list("bisection 2 sorted", 0, nr, list); | |
377 | ||
378 | *all = nr; | |
379 | weights = xcalloc(on_list, sizeof(*weights)); | |
380 | ||
381 | /* Do the real work of finding bisection commit. */ | |
382 | best = do_find_bisection(list, nr, weights, find_all); | |
383 | if (best) { | |
384 | if (!find_all) | |
385 | best->next = NULL; | |
386 | *reaches = weight(best); | |
387 | } | |
388 | free(weights); | |
389 | return best; | |
390 | } | |
391 | ||
95188648 CC |
392 | static int skipcmp(const void *a, const void *b) |
393 | { | |
394 | return hashcmp(a, b); | |
395 | } | |
396 | ||
397 | static void prepare_skipped(void) | |
398 | { | |
399 | qsort(skipped_sha1, skipped_sha1_nr, sizeof(*skipped_sha1), skipcmp); | |
400 | } | |
401 | ||
402 | static int lookup_skipped(unsigned char *sha1) | |
403 | { | |
404 | int lo, hi; | |
405 | lo = 0; | |
406 | hi = skipped_sha1_nr; | |
407 | while (lo < hi) { | |
408 | int mi = (lo + hi) / 2; | |
409 | int cmp = hashcmp(sha1, skipped_sha1[mi]); | |
410 | if (!cmp) | |
411 | return mi; | |
412 | if (cmp < 0) | |
413 | hi = mi; | |
414 | else | |
415 | lo = mi + 1; | |
416 | } | |
417 | return -lo - 1; | |
418 | } | |
419 | ||
420 | struct commit_list *filter_skipped(struct commit_list *list, | |
421 | struct commit_list **tried, | |
422 | int show_all) | |
423 | { | |
424 | struct commit_list *filtered = NULL, **f = &filtered; | |
425 | ||
426 | *tried = NULL; | |
427 | ||
428 | if (!skipped_sha1_nr) | |
429 | return list; | |
430 | ||
431 | prepare_skipped(); | |
432 | ||
433 | while (list) { | |
434 | struct commit_list *next = list->next; | |
435 | list->next = NULL; | |
436 | if (0 <= lookup_skipped(list->item->object.sha1)) { | |
437 | /* Move current to tried list */ | |
438 | *tried = list; | |
439 | tried = &list->next; | |
440 | } else { | |
441 | if (!show_all) | |
442 | return list; | |
443 | /* Move current to filtered list */ | |
444 | *f = list; | |
445 | f = &list->next; | |
446 | } | |
447 | list = next; | |
448 | } | |
449 | ||
450 | return filtered; | |
451 | } |