]>
Commit | Line | Data |
---|---|---|
5e3f94df | 1 | #include "git-compat-util.h" |
41771fa4 | 2 | #include "hex.h" |
d4ff2072 | 3 | #include "match-trees.h" |
68faf689 JH |
4 | #include "tree.h" |
5 | #include "tree-walk.h" | |
cbd53a21 | 6 | #include "object-store.h" |
68faf689 | 7 | |
667f71c3 | 8 | static int score_missing(unsigned mode) |
68faf689 JH |
9 | { |
10 | int score; | |
11 | ||
12 | if (S_ISDIR(mode)) | |
13 | score = -1000; | |
14 | else if (S_ISLNK(mode)) | |
15 | score = -500; | |
16 | else | |
17 | score = -50; | |
18 | return score; | |
19 | } | |
20 | ||
667f71c3 | 21 | static int score_differs(unsigned mode1, unsigned mode2) |
68faf689 JH |
22 | { |
23 | int score; | |
24 | ||
25 | if (S_ISDIR(mode1) != S_ISDIR(mode2)) | |
26 | score = -100; | |
27 | else if (S_ISLNK(mode1) != S_ISLNK(mode2)) | |
28 | score = -50; | |
29 | else | |
30 | score = -5; | |
31 | return score; | |
32 | } | |
33 | ||
667f71c3 | 34 | static int score_matches(unsigned mode1, unsigned mode2) |
68faf689 JH |
35 | { |
36 | int score; | |
37 | ||
38 | /* Heh, we found SHA-1 collisions between different kind of objects */ | |
39 | if (S_ISDIR(mode1) != S_ISDIR(mode2)) | |
40 | score = -100; | |
41 | else if (S_ISLNK(mode1) != S_ISLNK(mode2)) | |
42 | score = -50; | |
43 | ||
44 | else if (S_ISDIR(mode1)) | |
45 | score = 1000; | |
46 | else if (S_ISLNK(mode1)) | |
47 | score = 500; | |
48 | else | |
49 | score = 250; | |
50 | return score; | |
51 | } | |
52 | ||
10a3fb00 | 53 | static void *fill_tree_desc_strict(struct tree_desc *desc, |
b6aec868 | 54 | const struct object_id *hash) |
10a3fb00 RS |
55 | { |
56 | void *buffer; | |
57 | enum object_type type; | |
58 | unsigned long size; | |
59 | ||
bc726bd0 | 60 | buffer = repo_read_object_file(the_repository, hash, &type, &size); |
10a3fb00 | 61 | if (!buffer) |
b6aec868 | 62 | die("unable to read tree (%s)", oid_to_hex(hash)); |
10a3fb00 | 63 | if (type != OBJ_TREE) |
b6aec868 | 64 | die("%s is not a tree", oid_to_hex(hash)); |
10a3fb00 RS |
65 | init_tree_desc(desc, buffer, size); |
66 | return buffer; | |
67 | } | |
68 | ||
d8febde3 RS |
69 | static int base_name_entries_compare(const struct name_entry *a, |
70 | const struct name_entry *b) | |
71 | { | |
72 | return base_name_compare(a->path, tree_entry_len(a), a->mode, | |
73 | b->path, tree_entry_len(b), b->mode); | |
74 | } | |
75 | ||
68faf689 JH |
76 | /* |
77 | * Inspect two trees, and give a score that tells how similar they are. | |
78 | */ | |
b6aec868 | 79 | static int score_trees(const struct object_id *hash1, const struct object_id *hash2) |
68faf689 JH |
80 | { |
81 | struct tree_desc one; | |
82 | struct tree_desc two; | |
10a3fb00 RS |
83 | void *one_buf = fill_tree_desc_strict(&one, hash1); |
84 | void *two_buf = fill_tree_desc_strict(&two, hash2); | |
68faf689 | 85 | int score = 0; |
68faf689 | 86 | |
d8febde3 | 87 | for (;;) { |
68faf689 JH |
88 | int cmp; |
89 | ||
2ec41507 JK |
90 | if (one.size && two.size) |
91 | cmp = base_name_entries_compare(&one.entry, &two.entry); | |
92 | else if (one.size) | |
68faf689 | 93 | /* two lacks this entry */ |
d8febde3 | 94 | cmp = -1; |
2ec41507 | 95 | else if (two.size) |
d8febde3 RS |
96 | /* two has more entries */ |
97 | cmp = 1; | |
98 | else | |
99 | break; | |
100 | ||
2ec41507 | 101 | if (cmp < 0) { |
68faf689 | 102 | /* path1 does not appear in two */ |
667f71c3 | 103 | score += score_missing(one.entry.mode); |
2ec41507 JK |
104 | update_tree_entry(&one); |
105 | } else if (cmp > 0) { | |
68faf689 | 106 | /* path2 does not appear in one */ |
667f71c3 | 107 | score += score_missing(two.entry.mode); |
2ec41507 JK |
108 | update_tree_entry(&two); |
109 | } else { | |
110 | /* path appears in both */ | |
ea82b2a0 | 111 | if (!oideq(&one.entry.oid, &two.entry.oid)) { |
2ec41507 JK |
112 | /* they are different */ |
113 | score += score_differs(one.entry.mode, | |
667f71c3 | 114 | two.entry.mode); |
2ec41507 JK |
115 | } else { |
116 | /* same subtree or blob */ | |
117 | score += score_matches(one.entry.mode, | |
667f71c3 | 118 | two.entry.mode); |
2ec41507 JK |
119 | } |
120 | update_tree_entry(&one); | |
121 | update_tree_entry(&two); | |
122 | } | |
68faf689 JH |
123 | } |
124 | free(one_buf); | |
125 | free(two_buf); | |
126 | return score; | |
127 | } | |
128 | ||
129 | /* | |
130 | * Match one itself and its subtrees with two and pick the best match. | |
131 | */ | |
b6aec868 | 132 | static void match_trees(const struct object_id *hash1, |
133 | const struct object_id *hash2, | |
68faf689 JH |
134 | int *best_score, |
135 | char **best_match, | |
538dfe73 | 136 | const char *base, |
68faf689 JH |
137 | int recurse_limit) |
138 | { | |
139 | struct tree_desc one; | |
10a3fb00 | 140 | void *one_buf = fill_tree_desc_strict(&one, hash1); |
68faf689 JH |
141 | |
142 | while (one.size) { | |
143 | const char *path; | |
ce6663a9 | 144 | const struct object_id *elem; |
5ec1e728 | 145 | unsigned short mode; |
68faf689 JH |
146 | int score; |
147 | ||
148 | elem = tree_entry_extract(&one, &path, &mode); | |
149 | if (!S_ISDIR(mode)) | |
150 | goto next; | |
b6aec868 | 151 | score = score_trees(elem, hash2); |
68faf689 | 152 | if (*best_score < score) { |
68faf689 | 153 | free(*best_match); |
28310186 | 154 | *best_match = xstrfmt("%s%s", base, path); |
68faf689 JH |
155 | *best_score = score; |
156 | } | |
157 | if (recurse_limit) { | |
28310186 | 158 | char *newbase = xstrfmt("%s%s/", base, path); |
b6aec868 | 159 | match_trees(elem, hash2, best_score, best_match, |
68faf689 JH |
160 | newbase, recurse_limit - 1); |
161 | free(newbase); | |
162 | } | |
163 | ||
164 | next: | |
165 | update_tree_entry(&one); | |
166 | } | |
167 | free(one_buf); | |
168 | } | |
169 | ||
170 | /* | |
3b34934d PO |
171 | * A tree "oid1" has a subdirectory at "prefix". Come up with a tree object by |
172 | * replacing it with another tree "oid2". | |
68faf689 | 173 | */ |
3b34934d PO |
174 | static int splice_tree(const struct object_id *oid1, const char *prefix, |
175 | const struct object_id *oid2, struct object_id *result) | |
68faf689 JH |
176 | { |
177 | char *subpath; | |
178 | int toplen; | |
179 | char *buf; | |
180 | unsigned long sz; | |
181 | struct tree_desc desc; | |
f55ac431 | 182 | unsigned char *rewrite_here; |
3b34934d PO |
183 | const struct object_id *rewrite_with; |
184 | struct object_id subtree; | |
68faf689 JH |
185 | enum object_type type; |
186 | int status; | |
187 | ||
2c5495f7 RM |
188 | subpath = strchrnul(prefix, '/'); |
189 | toplen = subpath - prefix; | |
190 | if (*subpath) | |
68faf689 | 191 | subpath++; |
68faf689 | 192 | |
bc726bd0 | 193 | buf = repo_read_object_file(the_repository, oid1, &type, &sz); |
68faf689 | 194 | if (!buf) |
3b34934d | 195 | die("cannot read tree %s", oid_to_hex(oid1)); |
68faf689 JH |
196 | init_tree_desc(&desc, buf, sz); |
197 | ||
198 | rewrite_here = NULL; | |
199 | while (desc.size) { | |
200 | const char *name; | |
5ec1e728 | 201 | unsigned short mode; |
68faf689 | 202 | |
36775ab5 | 203 | tree_entry_extract(&desc, &name, &mode); |
68faf689 JH |
204 | if (strlen(name) == toplen && |
205 | !memcmp(name, prefix, toplen)) { | |
206 | if (!S_ISDIR(mode)) | |
3b34934d PO |
207 | die("entry %s in tree %s is not a tree", name, |
208 | oid_to_hex(oid1)); | |
f55ac431 | 209 | |
210 | /* | |
211 | * We cast here for two reasons: | |
212 | * | |
213 | * - to flip the "char *" (for the path) to "unsigned | |
214 | * char *" (for the hash stored after it) | |
215 | * | |
216 | * - to discard the "const"; this is OK because we | |
217 | * know it points into our non-const "buf" | |
218 | */ | |
219 | rewrite_here = (unsigned char *)(desc.entry.path + | |
220 | strlen(desc.entry.path) + | |
221 | 1); | |
68faf689 JH |
222 | break; |
223 | } | |
224 | update_tree_entry(&desc); | |
225 | } | |
226 | if (!rewrite_here) | |
3b34934d PO |
227 | die("entry %.*s not found in tree %s", toplen, prefix, |
228 | oid_to_hex(oid1)); | |
2c5495f7 | 229 | if (*subpath) { |
f55ac431 | 230 | struct object_id tree_oid; |
92e2cab9 | 231 | oidread(&tree_oid, rewrite_here); |
f55ac431 | 232 | status = splice_tree(&tree_oid, subpath, oid2, &subtree); |
68faf689 JH |
233 | if (status) |
234 | return status; | |
3b34934d PO |
235 | rewrite_with = &subtree; |
236 | } else { | |
237 | rewrite_with = oid2; | |
68faf689 | 238 | } |
f55ac431 | 239 | hashcpy(rewrite_here, rewrite_with->hash); |
c80d226a | 240 | status = write_object_file(buf, sz, OBJ_TREE, result); |
68faf689 JH |
241 | free(buf); |
242 | return status; | |
243 | } | |
244 | ||
245 | /* | |
246 | * We are trying to come up with a merge between one and two that | |
247 | * results in a tree shape similar to one. The tree two might | |
248 | * correspond to a subtree of one, in which case it needs to be | |
249 | * shifted down by prefixing otherwise empty directories. On the | |
250 | * other hand, it could cover tree one and we might need to pick a | |
251 | * subtree of it. | |
252 | */ | |
90d34051 NTND |
253 | void shift_tree(struct repository *r, |
254 | const struct object_id *hash1, | |
82db3d44 | 255 | const struct object_id *hash2, |
256 | struct object_id *shifted, | |
68faf689 JH |
257 | int depth_limit) |
258 | { | |
259 | char *add_prefix; | |
260 | char *del_prefix; | |
261 | int add_score, del_score; | |
262 | ||
85e51b78 JH |
263 | /* |
264 | * NEEDSWORK: this limits the recursion depth to hardcoded | |
265 | * value '2' to avoid excessive overhead. | |
266 | */ | |
267 | if (!depth_limit) | |
268 | depth_limit = 2; | |
269 | ||
b6aec868 | 270 | add_score = del_score = score_trees(hash1, hash2); |
68faf689 JH |
271 | add_prefix = xcalloc(1, 1); |
272 | del_prefix = xcalloc(1, 1); | |
273 | ||
274 | /* | |
275 | * See if one's subtree resembles two; if so we need to prefix | |
276 | * two with a few fake trees to match the prefix. | |
277 | */ | |
b6aec868 | 278 | match_trees(hash1, hash2, &add_score, &add_prefix, "", depth_limit); |
68faf689 JH |
279 | |
280 | /* | |
281 | * See if two's subtree resembles one; if so we need to | |
282 | * pick only subtree of two. | |
283 | */ | |
b6aec868 | 284 | match_trees(hash2, hash1, &del_score, &del_prefix, "", depth_limit); |
68faf689 JH |
285 | |
286 | /* Assume we do not have to do any shifting */ | |
82db3d44 | 287 | oidcpy(shifted, hash2); |
68faf689 JH |
288 | |
289 | if (add_score < del_score) { | |
290 | /* We need to pick a subtree of two */ | |
5ec1e728 | 291 | unsigned short mode; |
68faf689 JH |
292 | |
293 | if (!*del_prefix) | |
294 | return; | |
295 | ||
90d34051 | 296 | if (get_tree_entry(r, hash2, del_prefix, shifted, &mode)) |
68faf689 | 297 | die("cannot find path %s in tree %s", |
82db3d44 | 298 | del_prefix, oid_to_hex(hash2)); |
68faf689 JH |
299 | return; |
300 | } | |
301 | ||
302 | if (!*add_prefix) | |
303 | return; | |
304 | ||
3b34934d | 305 | splice_tree(hash1, add_prefix, hash2, shifted); |
68faf689 | 306 | } |
85e51b78 JH |
307 | |
308 | /* | |
309 | * The user says the trees will be shifted by this much. | |
310 | * Unfortunately we cannot fundamentally tell which one to | |
311 | * be prefixed, as recursive merge can work in either direction. | |
312 | */ | |
90d34051 NTND |
313 | void shift_tree_by(struct repository *r, |
314 | const struct object_id *hash1, | |
82db3d44 | 315 | const struct object_id *hash2, |
316 | struct object_id *shifted, | |
85e51b78 JH |
317 | const char *shift_prefix) |
318 | { | |
82db3d44 | 319 | struct object_id sub1, sub2; |
5ec1e728 | 320 | unsigned short mode1, mode2; |
85e51b78 JH |
321 | unsigned candidate = 0; |
322 | ||
323 | /* Can hash2 be a tree at shift_prefix in tree hash1? */ | |
90d34051 | 324 | if (!get_tree_entry(r, hash1, shift_prefix, &sub1, &mode1) && |
85e51b78 JH |
325 | S_ISDIR(mode1)) |
326 | candidate |= 1; | |
327 | ||
328 | /* Can hash1 be a tree at shift_prefix in tree hash2? */ | |
90d34051 | 329 | if (!get_tree_entry(r, hash2, shift_prefix, &sub2, &mode2) && |
85e51b78 JH |
330 | S_ISDIR(mode2)) |
331 | candidate |= 2; | |
332 | ||
333 | if (candidate == 3) { | |
334 | /* Both are plausible -- we need to evaluate the score */ | |
b6aec868 | 335 | int best_score = score_trees(hash1, hash2); |
85e51b78 JH |
336 | int score; |
337 | ||
338 | candidate = 0; | |
b6aec868 | 339 | score = score_trees(&sub1, hash2); |
85e51b78 JH |
340 | if (score > best_score) { |
341 | candidate = 1; | |
342 | best_score = score; | |
343 | } | |
b6aec868 | 344 | score = score_trees(&sub2, hash1); |
85e51b78 JH |
345 | if (score > best_score) |
346 | candidate = 2; | |
347 | } | |
348 | ||
349 | if (!candidate) { | |
350 | /* Neither is plausible -- do not shift */ | |
82db3d44 | 351 | oidcpy(shifted, hash2); |
85e51b78 JH |
352 | return; |
353 | } | |
354 | ||
355 | if (candidate == 1) | |
356 | /* | |
357 | * shift tree2 down by adding shift_prefix above it | |
358 | * to match tree1. | |
359 | */ | |
3b34934d | 360 | splice_tree(hash1, shift_prefix, hash2, shifted); |
85e51b78 JH |
361 | else |
362 | /* | |
363 | * shift tree2 up by removing shift_prefix from it | |
364 | * to match tree1. | |
365 | */ | |
82db3d44 | 366 | oidcpy(shifted, &sub2); |
85e51b78 | 367 | } |