Show More
@@ -1,2488 +1,2494 b'' | |||
|
1 | 1 | /* |
|
2 | 2 | parsers.c - efficient content parsing |
|
3 | 3 | |
|
4 | 4 | Copyright 2008 Matt Mackall <mpm@selenic.com> and others |
|
5 | 5 | |
|
6 | 6 | This software may be used and distributed according to the terms of |
|
7 | 7 | the GNU General Public License, incorporated herein by reference. |
|
8 | 8 | */ |
|
9 | 9 | |
|
10 | 10 | #include <Python.h> |
|
11 | 11 | #include <assert.h> |
|
12 | 12 | #include <ctype.h> |
|
13 | 13 | #include <stddef.h> |
|
14 | 14 | #include <string.h> |
|
15 | 15 | |
|
16 | 16 | #include "bitmanipulation.h" |
|
17 | 17 | #include "charencode.h" |
|
18 | 18 | #include "util.h" |
|
19 | 19 | |
|
20 | 20 | #ifdef IS_PY3K |
|
21 | 21 | /* The mapping of Python types is meant to be temporary to get Python |
|
22 | 22 | * 3 to compile. We should remove this once Python 3 support is fully |
|
23 | 23 | * supported and proper types are used in the extensions themselves. */ |
|
24 | 24 | #define PyInt_Check PyLong_Check |
|
25 | 25 | #define PyInt_FromLong PyLong_FromLong |
|
26 | 26 | #define PyInt_FromSsize_t PyLong_FromSsize_t |
|
27 | 27 | #define PyInt_AS_LONG PyLong_AS_LONG |
|
28 | 28 | #define PyInt_AsLong PyLong_AsLong |
|
29 | 29 | #endif |
|
30 | 30 | |
|
31 | 31 | typedef struct indexObjectStruct indexObject; |
|
32 | 32 | |
|
33 | 33 | typedef struct { |
|
34 | 34 | int children[16]; |
|
35 | 35 | } nodetreenode; |
|
36 | 36 | |
|
37 | 37 | /* |
|
38 | 38 | * A base-16 trie for fast node->rev mapping. |
|
39 | 39 | * |
|
40 | 40 | * Positive value is index of the next node in the trie |
|
41 | 41 | * Negative value is a leaf: -(rev + 2) |
|
42 | 42 | * Zero is empty |
|
43 | 43 | */ |
|
44 | 44 | typedef struct { |
|
45 | 45 | indexObject *index; |
|
46 | 46 | nodetreenode *nodes; |
|
47 | 47 | unsigned length; /* # nodes in use */ |
|
48 | 48 | unsigned capacity; /* # nodes allocated */ |
|
49 | 49 | int depth; /* maximum depth of tree */ |
|
50 | 50 | int splits; /* # splits performed */ |
|
51 | 51 | } nodetree; |
|
52 | 52 | |
|
53 | 53 | typedef struct { |
|
54 | 54 | PyObject_HEAD |
|
55 | 55 | nodetree nt; |
|
56 | 56 | } nodetreeObject; |
|
57 | 57 | |
|
58 | 58 | /* |
|
59 | 59 | * This class has two behaviors. |
|
60 | 60 | * |
|
61 | 61 | * When used in a list-like way (with integer keys), we decode an |
|
62 | 62 | * entry in a RevlogNG index file on demand. Our last entry is a |
|
63 | 63 | * sentinel, always a nullid. We have limited support for |
|
64 | 64 | * integer-keyed insert and delete, only at elements right before the |
|
65 | 65 | * sentinel. |
|
66 | 66 | * |
|
67 | 67 | * With string keys, we lazily perform a reverse mapping from node to |
|
68 | 68 | * rev, using a base-16 trie. |
|
69 | 69 | */ |
|
70 | 70 | struct indexObjectStruct { |
|
71 | 71 | PyObject_HEAD |
|
72 | 72 | /* Type-specific fields go here. */ |
|
73 | 73 | PyObject *data; /* raw bytes of index */ |
|
74 | 74 | Py_buffer buf; /* buffer of data */ |
|
75 | 75 | PyObject **cache; /* cached tuples */ |
|
76 | 76 | const char **offsets; /* populated on demand */ |
|
77 | 77 | Py_ssize_t raw_length; /* original number of elements */ |
|
78 | 78 | Py_ssize_t length; /* current number of elements */ |
|
79 | 79 | PyObject *added; /* populated on demand */ |
|
80 | 80 | PyObject *headrevs; /* cache, invalidated on changes */ |
|
81 | 81 | PyObject *filteredrevs;/* filtered revs set */ |
|
82 | 82 | nodetree nt; /* base-16 trie */ |
|
83 | 83 | int ntinitialized; /* 0 or 1 */ |
|
84 | 84 | int ntrev; /* last rev scanned */ |
|
85 | 85 | int ntlookups; /* # lookups */ |
|
86 | 86 | int ntmisses; /* # lookups that miss the cache */ |
|
87 | 87 | int inlined; |
|
88 | 88 | }; |
|
89 | 89 | |
|
90 | 90 | static Py_ssize_t index_length(const indexObject *self) |
|
91 | 91 | { |
|
92 | 92 | if (self->added == NULL) |
|
93 | 93 | return self->length; |
|
94 | 94 | return self->length + PyList_GET_SIZE(self->added); |
|
95 | 95 | } |
|
96 | 96 | |
|
97 | 97 | static PyObject *nullentry = NULL; |
|
98 | 98 | static const char nullid[20] = {0}; |
|
99 | 99 | |
|
100 | 100 | static Py_ssize_t inline_scan(indexObject *self, const char **offsets); |
|
101 | 101 | |
|
102 | 102 | #if LONG_MAX == 0x7fffffffL |
|
103 | 103 | static const char *const tuple_format = PY23("Kiiiiiis#", "Kiiiiiiy#"); |
|
104 | 104 | #else |
|
105 | 105 | static const char *const tuple_format = PY23("kiiiiiis#", "kiiiiiiy#"); |
|
106 | 106 | #endif |
|
107 | 107 | |
|
108 | 108 | /* A RevlogNG v1 index entry is 64 bytes long. */ |
|
109 | 109 | static const long v1_hdrsize = 64; |
|
110 | 110 | |
|
111 | 111 | static void raise_revlog_error(void) |
|
112 | 112 | { |
|
113 | 113 | PyObject *mod = NULL, *dict = NULL, *errclass = NULL; |
|
114 | 114 | |
|
115 | 115 | mod = PyImport_ImportModule("mercurial.error"); |
|
116 | 116 | if (mod == NULL) { |
|
117 | 117 | goto cleanup; |
|
118 | 118 | } |
|
119 | 119 | |
|
120 | 120 | dict = PyModule_GetDict(mod); |
|
121 | 121 | if (dict == NULL) { |
|
122 | 122 | goto cleanup; |
|
123 | 123 | } |
|
124 | 124 | Py_INCREF(dict); |
|
125 | 125 | |
|
126 | 126 | errclass = PyDict_GetItemString(dict, "RevlogError"); |
|
127 | 127 | if (errclass == NULL) { |
|
128 | 128 | PyErr_SetString(PyExc_SystemError, |
|
129 | 129 | "could not find RevlogError"); |
|
130 | 130 | goto cleanup; |
|
131 | 131 | } |
|
132 | 132 | |
|
133 | 133 | /* value of exception is ignored by callers */ |
|
134 | 134 | PyErr_SetString(errclass, "RevlogError"); |
|
135 | 135 | |
|
136 | 136 | cleanup: |
|
137 | 137 | Py_XDECREF(dict); |
|
138 | 138 | Py_XDECREF(mod); |
|
139 | 139 | } |
|
140 | 140 | |
|
141 | 141 | /* |
|
142 | 142 | * Return a pointer to the beginning of a RevlogNG record. |
|
143 | 143 | */ |
|
144 | 144 | static const char *index_deref(indexObject *self, Py_ssize_t pos) |
|
145 | 145 | { |
|
146 | 146 | if (self->inlined && pos > 0) { |
|
147 | 147 | if (self->offsets == NULL) { |
|
148 | 148 | self->offsets = PyMem_Malloc(self->raw_length * |
|
149 | 149 | sizeof(*self->offsets)); |
|
150 | 150 | if (self->offsets == NULL) |
|
151 | 151 | return (const char *)PyErr_NoMemory(); |
|
152 | 152 | inline_scan(self, self->offsets); |
|
153 | 153 | } |
|
154 | 154 | return self->offsets[pos]; |
|
155 | 155 | } |
|
156 | 156 | |
|
157 | 157 | return (const char *)(self->buf.buf) + pos * v1_hdrsize; |
|
158 | 158 | } |
|
159 | 159 | |
|
160 | /* | |
|
161 | * Get parents of the given rev. | |
|
162 | * | |
|
163 | * The specified rev must be valid and must not be nullrev. A returned | |
|
164 | * parent revision may be nullrev, but is guaranteed to be in valid range. | |
|
165 | */ | |
|
160 | 166 | static inline int index_get_parents(indexObject *self, Py_ssize_t rev, |
|
161 | 167 | int *ps, int maxrev) |
|
162 | 168 | { |
|
163 | 169 | if (rev >= self->length) { |
|
164 | 170 | PyObject *tuple = PyList_GET_ITEM(self->added, rev - self->length); |
|
165 | 171 | ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5)); |
|
166 | 172 | ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6)); |
|
167 | 173 | } else { |
|
168 | 174 | const char *data = index_deref(self, rev); |
|
169 | 175 | ps[0] = getbe32(data + 24); |
|
170 | 176 | ps[1] = getbe32(data + 28); |
|
171 | 177 | } |
|
172 | 178 | /* If index file is corrupted, ps[] may point to invalid revisions. So |
|
173 | 179 | * there is a risk of buffer overflow to trust them unconditionally. */ |
|
174 | if (ps[0] > maxrev || ps[1] > maxrev) { | |
|
180 | if (ps[0] < -1 || ps[0] > maxrev || ps[1] < -1 || ps[1] > maxrev) { | |
|
175 | 181 | PyErr_SetString(PyExc_ValueError, "parent out of range"); |
|
176 | 182 | return -1; |
|
177 | 183 | } |
|
178 | 184 | return 0; |
|
179 | 185 | } |
|
180 | 186 | |
|
181 | 187 | |
|
182 | 188 | /* |
|
183 | 189 | * RevlogNG format (all in big endian, data may be inlined): |
|
184 | 190 | * 6 bytes: offset |
|
185 | 191 | * 2 bytes: flags |
|
186 | 192 | * 4 bytes: compressed length |
|
187 | 193 | * 4 bytes: uncompressed length |
|
188 | 194 | * 4 bytes: base revision |
|
189 | 195 | * 4 bytes: link revision |
|
190 | 196 | * 4 bytes: parent 1 revision |
|
191 | 197 | * 4 bytes: parent 2 revision |
|
192 | 198 | * 32 bytes: nodeid (only 20 bytes used) |
|
193 | 199 | */ |
|
194 | 200 | static PyObject *index_get(indexObject *self, Py_ssize_t pos) |
|
195 | 201 | { |
|
196 | 202 | uint64_t offset_flags; |
|
197 | 203 | int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2; |
|
198 | 204 | const char *c_node_id; |
|
199 | 205 | const char *data; |
|
200 | 206 | Py_ssize_t length = index_length(self); |
|
201 | 207 | PyObject *entry; |
|
202 | 208 | |
|
203 | 209 | if (pos == -1) { |
|
204 | 210 | Py_INCREF(nullentry); |
|
205 | 211 | return nullentry; |
|
206 | 212 | } |
|
207 | 213 | |
|
208 | 214 | if (pos < 0 || pos >= length) { |
|
209 | 215 | PyErr_SetString(PyExc_IndexError, "revlog index out of range"); |
|
210 | 216 | return NULL; |
|
211 | 217 | } |
|
212 | 218 | |
|
213 | 219 | if (pos >= self->length) { |
|
214 | 220 | PyObject *obj; |
|
215 | 221 | obj = PyList_GET_ITEM(self->added, pos - self->length); |
|
216 | 222 | Py_INCREF(obj); |
|
217 | 223 | return obj; |
|
218 | 224 | } |
|
219 | 225 | |
|
220 | 226 | if (self->cache) { |
|
221 | 227 | if (self->cache[pos]) { |
|
222 | 228 | Py_INCREF(self->cache[pos]); |
|
223 | 229 | return self->cache[pos]; |
|
224 | 230 | } |
|
225 | 231 | } else { |
|
226 | 232 | self->cache = calloc(self->raw_length, sizeof(PyObject *)); |
|
227 | 233 | if (self->cache == NULL) |
|
228 | 234 | return PyErr_NoMemory(); |
|
229 | 235 | } |
|
230 | 236 | |
|
231 | 237 | data = index_deref(self, pos); |
|
232 | 238 | if (data == NULL) |
|
233 | 239 | return NULL; |
|
234 | 240 | |
|
235 | 241 | offset_flags = getbe32(data + 4); |
|
236 | 242 | if (pos == 0) /* mask out version number for the first entry */ |
|
237 | 243 | offset_flags &= 0xFFFF; |
|
238 | 244 | else { |
|
239 | 245 | uint32_t offset_high = getbe32(data); |
|
240 | 246 | offset_flags |= ((uint64_t)offset_high) << 32; |
|
241 | 247 | } |
|
242 | 248 | |
|
243 | 249 | comp_len = getbe32(data + 8); |
|
244 | 250 | uncomp_len = getbe32(data + 12); |
|
245 | 251 | base_rev = getbe32(data + 16); |
|
246 | 252 | link_rev = getbe32(data + 20); |
|
247 | 253 | parent_1 = getbe32(data + 24); |
|
248 | 254 | parent_2 = getbe32(data + 28); |
|
249 | 255 | c_node_id = data + 32; |
|
250 | 256 | |
|
251 | 257 | entry = Py_BuildValue(tuple_format, offset_flags, comp_len, |
|
252 | 258 | uncomp_len, base_rev, link_rev, |
|
253 | 259 | parent_1, parent_2, c_node_id, 20); |
|
254 | 260 | |
|
255 | 261 | if (entry) { |
|
256 | 262 | PyObject_GC_UnTrack(entry); |
|
257 | 263 | Py_INCREF(entry); |
|
258 | 264 | } |
|
259 | 265 | |
|
260 | 266 | self->cache[pos] = entry; |
|
261 | 267 | |
|
262 | 268 | return entry; |
|
263 | 269 | } |
|
264 | 270 | |
|
265 | 271 | /* |
|
266 | 272 | * Return the 20-byte SHA of the node corresponding to the given rev. |
|
267 | 273 | */ |
|
268 | 274 | static const char *index_node(indexObject *self, Py_ssize_t pos) |
|
269 | 275 | { |
|
270 | 276 | Py_ssize_t length = index_length(self); |
|
271 | 277 | const char *data; |
|
272 | 278 | |
|
273 | 279 | if (pos == -1) |
|
274 | 280 | return nullid; |
|
275 | 281 | |
|
276 | 282 | if (pos >= length) |
|
277 | 283 | return NULL; |
|
278 | 284 | |
|
279 | 285 | if (pos >= self->length) { |
|
280 | 286 | PyObject *tuple, *str; |
|
281 | 287 | tuple = PyList_GET_ITEM(self->added, pos - self->length); |
|
282 | 288 | str = PyTuple_GetItem(tuple, 7); |
|
283 | 289 | return str ? PyBytes_AS_STRING(str) : NULL; |
|
284 | 290 | } |
|
285 | 291 | |
|
286 | 292 | data = index_deref(self, pos); |
|
287 | 293 | return data ? data + 32 : NULL; |
|
288 | 294 | } |
|
289 | 295 | |
|
290 | 296 | /* |
|
291 | 297 | * Return the 20-byte SHA of the node corresponding to the given rev. The |
|
292 | 298 | * rev is assumed to be existing. If not, an exception is set. |
|
293 | 299 | */ |
|
294 | 300 | static const char *index_node_existing(indexObject *self, Py_ssize_t pos) |
|
295 | 301 | { |
|
296 | 302 | const char *node = index_node(self, pos); |
|
297 | 303 | if (node == NULL) { |
|
298 | 304 | PyErr_Format(PyExc_IndexError, "could not access rev %d", |
|
299 | 305 | (int)pos); |
|
300 | 306 | } |
|
301 | 307 | return node; |
|
302 | 308 | } |
|
303 | 309 | |
|
304 | 310 | static int nt_insert(nodetree *self, const char *node, int rev); |
|
305 | 311 | |
|
306 | 312 | static int node_check(PyObject *obj, char **node) |
|
307 | 313 | { |
|
308 | 314 | Py_ssize_t nodelen; |
|
309 | 315 | if (PyBytes_AsStringAndSize(obj, node, &nodelen) == -1) |
|
310 | 316 | return -1; |
|
311 | 317 | if (nodelen == 20) |
|
312 | 318 | return 0; |
|
313 | 319 | PyErr_SetString(PyExc_ValueError, "20-byte hash required"); |
|
314 | 320 | return -1; |
|
315 | 321 | } |
|
316 | 322 | |
|
317 | 323 | static PyObject *index_append(indexObject *self, PyObject *obj) |
|
318 | 324 | { |
|
319 | 325 | char *node; |
|
320 | 326 | Py_ssize_t len; |
|
321 | 327 | |
|
322 | 328 | if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) { |
|
323 | 329 | PyErr_SetString(PyExc_TypeError, "8-tuple required"); |
|
324 | 330 | return NULL; |
|
325 | 331 | } |
|
326 | 332 | |
|
327 | 333 | if (node_check(PyTuple_GET_ITEM(obj, 7), &node) == -1) |
|
328 | 334 | return NULL; |
|
329 | 335 | |
|
330 | 336 | len = index_length(self); |
|
331 | 337 | |
|
332 | 338 | if (self->added == NULL) { |
|
333 | 339 | self->added = PyList_New(0); |
|
334 | 340 | if (self->added == NULL) |
|
335 | 341 | return NULL; |
|
336 | 342 | } |
|
337 | 343 | |
|
338 | 344 | if (PyList_Append(self->added, obj) == -1) |
|
339 | 345 | return NULL; |
|
340 | 346 | |
|
341 | 347 | if (self->ntinitialized) |
|
342 | 348 | nt_insert(&self->nt, node, (int)len); |
|
343 | 349 | |
|
344 | 350 | Py_CLEAR(self->headrevs); |
|
345 | 351 | Py_RETURN_NONE; |
|
346 | 352 | } |
|
347 | 353 | |
|
348 | 354 | static PyObject *index_stats(indexObject *self) |
|
349 | 355 | { |
|
350 | 356 | PyObject *obj = PyDict_New(); |
|
351 | 357 | PyObject *t = NULL; |
|
352 | 358 | |
|
353 | 359 | if (obj == NULL) |
|
354 | 360 | return NULL; |
|
355 | 361 | |
|
356 | 362 | #define istat(__n, __d) \ |
|
357 | 363 | do { \ |
|
358 | 364 | t = PyInt_FromSsize_t(self->__n); \ |
|
359 | 365 | if (!t) \ |
|
360 | 366 | goto bail; \ |
|
361 | 367 | if (PyDict_SetItemString(obj, __d, t) == -1) \ |
|
362 | 368 | goto bail; \ |
|
363 | 369 | Py_DECREF(t); \ |
|
364 | 370 | } while (0) |
|
365 | 371 | |
|
366 | 372 | if (self->added) { |
|
367 | 373 | Py_ssize_t len = PyList_GET_SIZE(self->added); |
|
368 | 374 | t = PyInt_FromSsize_t(len); |
|
369 | 375 | if (!t) |
|
370 | 376 | goto bail; |
|
371 | 377 | if (PyDict_SetItemString(obj, "index entries added", t) == -1) |
|
372 | 378 | goto bail; |
|
373 | 379 | Py_DECREF(t); |
|
374 | 380 | } |
|
375 | 381 | |
|
376 | 382 | if (self->raw_length != self->length) |
|
377 | 383 | istat(raw_length, "revs on disk"); |
|
378 | 384 | istat(length, "revs in memory"); |
|
379 | 385 | istat(ntlookups, "node trie lookups"); |
|
380 | 386 | istat(ntmisses, "node trie misses"); |
|
381 | 387 | istat(ntrev, "node trie last rev scanned"); |
|
382 | 388 | if (self->ntinitialized) { |
|
383 | 389 | istat(nt.capacity, "node trie capacity"); |
|
384 | 390 | istat(nt.depth, "node trie depth"); |
|
385 | 391 | istat(nt.length, "node trie count"); |
|
386 | 392 | istat(nt.splits, "node trie splits"); |
|
387 | 393 | } |
|
388 | 394 | |
|
389 | 395 | #undef istat |
|
390 | 396 | |
|
391 | 397 | return obj; |
|
392 | 398 | |
|
393 | 399 | bail: |
|
394 | 400 | Py_XDECREF(obj); |
|
395 | 401 | Py_XDECREF(t); |
|
396 | 402 | return NULL; |
|
397 | 403 | } |
|
398 | 404 | |
|
399 | 405 | /* |
|
400 | 406 | * When we cache a list, we want to be sure the caller can't mutate |
|
401 | 407 | * the cached copy. |
|
402 | 408 | */ |
|
403 | 409 | static PyObject *list_copy(PyObject *list) |
|
404 | 410 | { |
|
405 | 411 | Py_ssize_t len = PyList_GET_SIZE(list); |
|
406 | 412 | PyObject *newlist = PyList_New(len); |
|
407 | 413 | Py_ssize_t i; |
|
408 | 414 | |
|
409 | 415 | if (newlist == NULL) |
|
410 | 416 | return NULL; |
|
411 | 417 | |
|
412 | 418 | for (i = 0; i < len; i++) { |
|
413 | 419 | PyObject *obj = PyList_GET_ITEM(list, i); |
|
414 | 420 | Py_INCREF(obj); |
|
415 | 421 | PyList_SET_ITEM(newlist, i, obj); |
|
416 | 422 | } |
|
417 | 423 | |
|
418 | 424 | return newlist; |
|
419 | 425 | } |
|
420 | 426 | |
|
421 | 427 | static int check_filter(PyObject *filter, Py_ssize_t arg) |
|
422 | 428 | { |
|
423 | 429 | if (filter) { |
|
424 | 430 | PyObject *arglist, *result; |
|
425 | 431 | int isfiltered; |
|
426 | 432 | |
|
427 | 433 | arglist = Py_BuildValue("(n)", arg); |
|
428 | 434 | if (!arglist) { |
|
429 | 435 | return -1; |
|
430 | 436 | } |
|
431 | 437 | |
|
432 | 438 | result = PyEval_CallObject(filter, arglist); |
|
433 | 439 | Py_DECREF(arglist); |
|
434 | 440 | if (!result) { |
|
435 | 441 | return -1; |
|
436 | 442 | } |
|
437 | 443 | |
|
438 | 444 | /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error, |
|
439 | 445 | * same as this function, so we can just return it directly.*/ |
|
440 | 446 | isfiltered = PyObject_IsTrue(result); |
|
441 | 447 | Py_DECREF(result); |
|
442 | 448 | return isfiltered; |
|
443 | 449 | } else { |
|
444 | 450 | return 0; |
|
445 | 451 | } |
|
446 | 452 | } |
|
447 | 453 | |
|
448 | 454 | static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list, |
|
449 | 455 | Py_ssize_t marker, char *phases) |
|
450 | 456 | { |
|
451 | 457 | PyObject *iter = NULL; |
|
452 | 458 | PyObject *iter_item = NULL; |
|
453 | 459 | Py_ssize_t min_idx = index_length(self) + 2; |
|
454 | 460 | long iter_item_long; |
|
455 | 461 | |
|
456 | 462 | if (PyList_GET_SIZE(list) != 0) { |
|
457 | 463 | iter = PyObject_GetIter(list); |
|
458 | 464 | if (iter == NULL) |
|
459 | 465 | return -2; |
|
460 | 466 | while ((iter_item = PyIter_Next(iter))) { |
|
461 | 467 | iter_item_long = PyInt_AS_LONG(iter_item); |
|
462 | 468 | Py_DECREF(iter_item); |
|
463 | 469 | if (iter_item_long < min_idx) |
|
464 | 470 | min_idx = iter_item_long; |
|
465 | 471 | phases[iter_item_long] = (char)marker; |
|
466 | 472 | } |
|
467 | 473 | Py_DECREF(iter); |
|
468 | 474 | } |
|
469 | 475 | |
|
470 | 476 | return min_idx; |
|
471 | 477 | } |
|
472 | 478 | |
|
473 | 479 | static inline void set_phase_from_parents(char *phases, int parent_1, |
|
474 | 480 | int parent_2, Py_ssize_t i) |
|
475 | 481 | { |
|
476 | 482 | if (parent_1 >= 0 && phases[parent_1] > phases[i]) |
|
477 | 483 | phases[i] = phases[parent_1]; |
|
478 | 484 | if (parent_2 >= 0 && phases[parent_2] > phases[i]) |
|
479 | 485 | phases[i] = phases[parent_2]; |
|
480 | 486 | } |
|
481 | 487 | |
|
482 | 488 | static PyObject *reachableroots2(indexObject *self, PyObject *args) |
|
483 | 489 | { |
|
484 | 490 | |
|
485 | 491 | /* Input */ |
|
486 | 492 | long minroot; |
|
487 | 493 | PyObject *includepatharg = NULL; |
|
488 | 494 | int includepath = 0; |
|
489 | 495 | /* heads and roots are lists */ |
|
490 | 496 | PyObject *heads = NULL; |
|
491 | 497 | PyObject *roots = NULL; |
|
492 | 498 | PyObject *reachable = NULL; |
|
493 | 499 | |
|
494 | 500 | PyObject *val; |
|
495 | 501 | Py_ssize_t len = index_length(self); |
|
496 | 502 | long revnum; |
|
497 | 503 | Py_ssize_t k; |
|
498 | 504 | Py_ssize_t i; |
|
499 | 505 | Py_ssize_t l; |
|
500 | 506 | int r; |
|
501 | 507 | int parents[2]; |
|
502 | 508 | |
|
503 | 509 | /* Internal data structure: |
|
504 | 510 | * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit |
|
505 | 511 | * revstates: array of length len+1 (all revs + nullrev) */ |
|
506 | 512 | int *tovisit = NULL; |
|
507 | 513 | long lentovisit = 0; |
|
508 | 514 | enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 }; |
|
509 | 515 | char *revstates = NULL; |
|
510 | 516 | |
|
511 | 517 | /* Get arguments */ |
|
512 | 518 | if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads, |
|
513 | 519 | &PyList_Type, &roots, |
|
514 | 520 | &PyBool_Type, &includepatharg)) |
|
515 | 521 | goto bail; |
|
516 | 522 | |
|
517 | 523 | if (includepatharg == Py_True) |
|
518 | 524 | includepath = 1; |
|
519 | 525 | |
|
520 | 526 | /* Initialize return set */ |
|
521 | 527 | reachable = PyList_New(0); |
|
522 | 528 | if (reachable == NULL) |
|
523 | 529 | goto bail; |
|
524 | 530 | |
|
525 | 531 | /* Initialize internal datastructures */ |
|
526 | 532 | tovisit = (int *)malloc((len + 1) * sizeof(int)); |
|
527 | 533 | if (tovisit == NULL) { |
|
528 | 534 | PyErr_NoMemory(); |
|
529 | 535 | goto bail; |
|
530 | 536 | } |
|
531 | 537 | |
|
532 | 538 | revstates = (char *)calloc(len + 1, 1); |
|
533 | 539 | if (revstates == NULL) { |
|
534 | 540 | PyErr_NoMemory(); |
|
535 | 541 | goto bail; |
|
536 | 542 | } |
|
537 | 543 | |
|
538 | 544 | l = PyList_GET_SIZE(roots); |
|
539 | 545 | for (i = 0; i < l; i++) { |
|
540 | 546 | revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i)); |
|
541 | 547 | if (revnum == -1 && PyErr_Occurred()) |
|
542 | 548 | goto bail; |
|
543 | 549 | /* If root is out of range, e.g. wdir(), it must be unreachable |
|
544 | 550 | * from heads. So we can just ignore it. */ |
|
545 | 551 | if (revnum + 1 < 0 || revnum + 1 >= len + 1) |
|
546 | 552 | continue; |
|
547 | 553 | revstates[revnum + 1] |= RS_ROOT; |
|
548 | 554 | } |
|
549 | 555 | |
|
550 | 556 | /* Populate tovisit with all the heads */ |
|
551 | 557 | l = PyList_GET_SIZE(heads); |
|
552 | 558 | for (i = 0; i < l; i++) { |
|
553 | 559 | revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i)); |
|
554 | 560 | if (revnum == -1 && PyErr_Occurred()) |
|
555 | 561 | goto bail; |
|
556 | 562 | if (revnum + 1 < 0 || revnum + 1 >= len + 1) { |
|
557 | 563 | PyErr_SetString(PyExc_IndexError, "head out of range"); |
|
558 | 564 | goto bail; |
|
559 | 565 | } |
|
560 | 566 | if (!(revstates[revnum + 1] & RS_SEEN)) { |
|
561 | 567 | tovisit[lentovisit++] = (int)revnum; |
|
562 | 568 | revstates[revnum + 1] |= RS_SEEN; |
|
563 | 569 | } |
|
564 | 570 | } |
|
565 | 571 | |
|
566 | 572 | /* Visit the tovisit list and find the reachable roots */ |
|
567 | 573 | k = 0; |
|
568 | 574 | while (k < lentovisit) { |
|
569 | 575 | /* Add the node to reachable if it is a root*/ |
|
570 | 576 | revnum = tovisit[k++]; |
|
571 | 577 | if (revstates[revnum + 1] & RS_ROOT) { |
|
572 | 578 | revstates[revnum + 1] |= RS_REACHABLE; |
|
573 | 579 | val = PyInt_FromLong(revnum); |
|
574 | 580 | if (val == NULL) |
|
575 | 581 | goto bail; |
|
576 | 582 | r = PyList_Append(reachable, val); |
|
577 | 583 | Py_DECREF(val); |
|
578 | 584 | if (r < 0) |
|
579 | 585 | goto bail; |
|
580 | 586 | if (includepath == 0) |
|
581 | 587 | continue; |
|
582 | 588 | } |
|
583 | 589 | |
|
584 | 590 | /* Add its parents to the list of nodes to visit */ |
|
585 | 591 | if (revnum == -1) |
|
586 | 592 | continue; |
|
587 | 593 | r = index_get_parents(self, revnum, parents, (int)len - 1); |
|
588 | 594 | if (r < 0) |
|
589 | 595 | goto bail; |
|
590 | 596 | for (i = 0; i < 2; i++) { |
|
591 | 597 | if (!(revstates[parents[i] + 1] & RS_SEEN) |
|
592 | 598 | && parents[i] >= minroot) { |
|
593 | 599 | tovisit[lentovisit++] = parents[i]; |
|
594 | 600 | revstates[parents[i] + 1] |= RS_SEEN; |
|
595 | 601 | } |
|
596 | 602 | } |
|
597 | 603 | } |
|
598 | 604 | |
|
599 | 605 | /* Find all the nodes in between the roots we found and the heads |
|
600 | 606 | * and add them to the reachable set */ |
|
601 | 607 | if (includepath == 1) { |
|
602 | 608 | long minidx = minroot; |
|
603 | 609 | if (minidx < 0) |
|
604 | 610 | minidx = 0; |
|
605 | 611 | for (i = minidx; i < len; i++) { |
|
606 | 612 | if (!(revstates[i + 1] & RS_SEEN)) |
|
607 | 613 | continue; |
|
608 | 614 | r = index_get_parents(self, i, parents, (int)len - 1); |
|
609 | 615 | /* Corrupted index file, error is set from |
|
610 | 616 | * index_get_parents */ |
|
611 | 617 | if (r < 0) |
|
612 | 618 | goto bail; |
|
613 | 619 | if (((revstates[parents[0] + 1] | |
|
614 | 620 | revstates[parents[1] + 1]) & RS_REACHABLE) |
|
615 | 621 | && !(revstates[i + 1] & RS_REACHABLE)) { |
|
616 | 622 | revstates[i + 1] |= RS_REACHABLE; |
|
617 | 623 | val = PyInt_FromSsize_t(i); |
|
618 | 624 | if (val == NULL) |
|
619 | 625 | goto bail; |
|
620 | 626 | r = PyList_Append(reachable, val); |
|
621 | 627 | Py_DECREF(val); |
|
622 | 628 | if (r < 0) |
|
623 | 629 | goto bail; |
|
624 | 630 | } |
|
625 | 631 | } |
|
626 | 632 | } |
|
627 | 633 | |
|
628 | 634 | free(revstates); |
|
629 | 635 | free(tovisit); |
|
630 | 636 | return reachable; |
|
631 | 637 | bail: |
|
632 | 638 | Py_XDECREF(reachable); |
|
633 | 639 | free(revstates); |
|
634 | 640 | free(tovisit); |
|
635 | 641 | return NULL; |
|
636 | 642 | } |
|
637 | 643 | |
|
638 | 644 | static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args) |
|
639 | 645 | { |
|
640 | 646 | PyObject *roots = Py_None; |
|
641 | 647 | PyObject *ret = NULL; |
|
642 | 648 | PyObject *phasessize = NULL; |
|
643 | 649 | PyObject *phaseroots = NULL; |
|
644 | 650 | PyObject *phaseset = NULL; |
|
645 | 651 | PyObject *phasessetlist = NULL; |
|
646 | 652 | PyObject *rev = NULL; |
|
647 | 653 | Py_ssize_t len = index_length(self); |
|
648 | 654 | Py_ssize_t numphase = 0; |
|
649 | 655 | Py_ssize_t minrevallphases = 0; |
|
650 | 656 | Py_ssize_t minrevphase = 0; |
|
651 | 657 | Py_ssize_t i = 0; |
|
652 | 658 | char *phases = NULL; |
|
653 | 659 | long phase; |
|
654 | 660 | |
|
655 | 661 | if (!PyArg_ParseTuple(args, "O", &roots)) |
|
656 | 662 | goto done; |
|
657 | 663 | if (roots == NULL || !PyList_Check(roots)) { |
|
658 | 664 | PyErr_SetString(PyExc_TypeError, "roots must be a list"); |
|
659 | 665 | goto done; |
|
660 | 666 | } |
|
661 | 667 | |
|
662 | 668 | phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */ |
|
663 | 669 | if (phases == NULL) { |
|
664 | 670 | PyErr_NoMemory(); |
|
665 | 671 | goto done; |
|
666 | 672 | } |
|
667 | 673 | /* Put the phase information of all the roots in phases */ |
|
668 | 674 | numphase = PyList_GET_SIZE(roots)+1; |
|
669 | 675 | minrevallphases = len + 1; |
|
670 | 676 | phasessetlist = PyList_New(numphase); |
|
671 | 677 | if (phasessetlist == NULL) |
|
672 | 678 | goto done; |
|
673 | 679 | |
|
674 | 680 | PyList_SET_ITEM(phasessetlist, 0, Py_None); |
|
675 | 681 | Py_INCREF(Py_None); |
|
676 | 682 | |
|
677 | 683 | for (i = 0; i < numphase-1; i++) { |
|
678 | 684 | phaseroots = PyList_GET_ITEM(roots, i); |
|
679 | 685 | phaseset = PySet_New(NULL); |
|
680 | 686 | if (phaseset == NULL) |
|
681 | 687 | goto release; |
|
682 | 688 | PyList_SET_ITEM(phasessetlist, i+1, phaseset); |
|
683 | 689 | if (!PyList_Check(phaseroots)) { |
|
684 | 690 | PyErr_SetString(PyExc_TypeError, |
|
685 | 691 | "roots item must be a list"); |
|
686 | 692 | goto release; |
|
687 | 693 | } |
|
688 | 694 | minrevphase = add_roots_get_min(self, phaseroots, i+1, phases); |
|
689 | 695 | if (minrevphase == -2) /* Error from add_roots_get_min */ |
|
690 | 696 | goto release; |
|
691 | 697 | minrevallphases = MIN(minrevallphases, minrevphase); |
|
692 | 698 | } |
|
693 | 699 | /* Propagate the phase information from the roots to the revs */ |
|
694 | 700 | if (minrevallphases != -1) { |
|
695 | 701 | int parents[2]; |
|
696 | 702 | for (i = minrevallphases; i < len; i++) { |
|
697 | 703 | if (index_get_parents(self, i, parents, |
|
698 | 704 | (int)len - 1) < 0) |
|
699 | 705 | goto release; |
|
700 | 706 | set_phase_from_parents(phases, parents[0], parents[1], i); |
|
701 | 707 | } |
|
702 | 708 | } |
|
703 | 709 | /* Transform phase list to a python list */ |
|
704 | 710 | phasessize = PyInt_FromSsize_t(len); |
|
705 | 711 | if (phasessize == NULL) |
|
706 | 712 | goto release; |
|
707 | 713 | for (i = 0; i < len; i++) { |
|
708 | 714 | phase = phases[i]; |
|
709 | 715 | /* We only store the sets of phase for non public phase, the public phase |
|
710 | 716 | * is computed as a difference */ |
|
711 | 717 | if (phase != 0) { |
|
712 | 718 | phaseset = PyList_GET_ITEM(phasessetlist, phase); |
|
713 | 719 | rev = PyInt_FromSsize_t(i); |
|
714 | 720 | if (rev == NULL) |
|
715 | 721 | goto release; |
|
716 | 722 | PySet_Add(phaseset, rev); |
|
717 | 723 | Py_XDECREF(rev); |
|
718 | 724 | } |
|
719 | 725 | } |
|
720 | 726 | ret = PyTuple_Pack(2, phasessize, phasessetlist); |
|
721 | 727 | |
|
722 | 728 | release: |
|
723 | 729 | Py_XDECREF(phasessize); |
|
724 | 730 | Py_XDECREF(phasessetlist); |
|
725 | 731 | done: |
|
726 | 732 | free(phases); |
|
727 | 733 | return ret; |
|
728 | 734 | } |
|
729 | 735 | |
|
730 | 736 | static PyObject *index_headrevs(indexObject *self, PyObject *args) |
|
731 | 737 | { |
|
732 | 738 | Py_ssize_t i, j, len; |
|
733 | 739 | char *nothead = NULL; |
|
734 | 740 | PyObject *heads = NULL; |
|
735 | 741 | PyObject *filter = NULL; |
|
736 | 742 | PyObject *filteredrevs = Py_None; |
|
737 | 743 | |
|
738 | 744 | if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) { |
|
739 | 745 | return NULL; |
|
740 | 746 | } |
|
741 | 747 | |
|
742 | 748 | if (self->headrevs && filteredrevs == self->filteredrevs) |
|
743 | 749 | return list_copy(self->headrevs); |
|
744 | 750 | |
|
745 | 751 | Py_DECREF(self->filteredrevs); |
|
746 | 752 | self->filteredrevs = filteredrevs; |
|
747 | 753 | Py_INCREF(filteredrevs); |
|
748 | 754 | |
|
749 | 755 | if (filteredrevs != Py_None) { |
|
750 | 756 | filter = PyObject_GetAttrString(filteredrevs, "__contains__"); |
|
751 | 757 | if (!filter) { |
|
752 | 758 | PyErr_SetString(PyExc_TypeError, |
|
753 | 759 | "filteredrevs has no attribute __contains__"); |
|
754 | 760 | goto bail; |
|
755 | 761 | } |
|
756 | 762 | } |
|
757 | 763 | |
|
758 | 764 | len = index_length(self); |
|
759 | 765 | heads = PyList_New(0); |
|
760 | 766 | if (heads == NULL) |
|
761 | 767 | goto bail; |
|
762 | 768 | if (len == 0) { |
|
763 | 769 | PyObject *nullid = PyInt_FromLong(-1); |
|
764 | 770 | if (nullid == NULL || PyList_Append(heads, nullid) == -1) { |
|
765 | 771 | Py_XDECREF(nullid); |
|
766 | 772 | goto bail; |
|
767 | 773 | } |
|
768 | 774 | goto done; |
|
769 | 775 | } |
|
770 | 776 | |
|
771 | 777 | nothead = calloc(len, 1); |
|
772 | 778 | if (nothead == NULL) { |
|
773 | 779 | PyErr_NoMemory(); |
|
774 | 780 | goto bail; |
|
775 | 781 | } |
|
776 | 782 | |
|
777 | 783 | for (i = len - 1; i >= 0; i--) { |
|
778 | 784 | int isfiltered; |
|
779 | 785 | int parents[2]; |
|
780 | 786 | |
|
781 | 787 | /* If nothead[i] == 1, it means we've seen an unfiltered child of this |
|
782 | 788 | * node already, and therefore this node is not filtered. So we can skip |
|
783 | 789 | * the expensive check_filter step. |
|
784 | 790 | */ |
|
785 | 791 | if (nothead[i] != 1) { |
|
786 | 792 | isfiltered = check_filter(filter, i); |
|
787 | 793 | if (isfiltered == -1) { |
|
788 | 794 | PyErr_SetString(PyExc_TypeError, |
|
789 | 795 | "unable to check filter"); |
|
790 | 796 | goto bail; |
|
791 | 797 | } |
|
792 | 798 | |
|
793 | 799 | if (isfiltered) { |
|
794 | 800 | nothead[i] = 1; |
|
795 | 801 | continue; |
|
796 | 802 | } |
|
797 | 803 | } |
|
798 | 804 | |
|
799 | 805 | if (index_get_parents(self, i, parents, (int)len - 1) < 0) |
|
800 | 806 | goto bail; |
|
801 | 807 | for (j = 0; j < 2; j++) { |
|
802 | 808 | if (parents[j] >= 0) |
|
803 | 809 | nothead[parents[j]] = 1; |
|
804 | 810 | } |
|
805 | 811 | } |
|
806 | 812 | |
|
807 | 813 | for (i = 0; i < len; i++) { |
|
808 | 814 | PyObject *head; |
|
809 | 815 | |
|
810 | 816 | if (nothead[i]) |
|
811 | 817 | continue; |
|
812 | 818 | head = PyInt_FromSsize_t(i); |
|
813 | 819 | if (head == NULL || PyList_Append(heads, head) == -1) { |
|
814 | 820 | Py_XDECREF(head); |
|
815 | 821 | goto bail; |
|
816 | 822 | } |
|
817 | 823 | } |
|
818 | 824 | |
|
819 | 825 | done: |
|
820 | 826 | self->headrevs = heads; |
|
821 | 827 | Py_XDECREF(filter); |
|
822 | 828 | free(nothead); |
|
823 | 829 | return list_copy(self->headrevs); |
|
824 | 830 | bail: |
|
825 | 831 | Py_XDECREF(filter); |
|
826 | 832 | Py_XDECREF(heads); |
|
827 | 833 | free(nothead); |
|
828 | 834 | return NULL; |
|
829 | 835 | } |
|
830 | 836 | |
|
831 | 837 | /** |
|
832 | 838 | * Obtain the base revision index entry. |
|
833 | 839 | * |
|
834 | 840 | * Callers must ensure that rev >= 0 or illegal memory access may occur. |
|
835 | 841 | */ |
|
836 | 842 | static inline int index_baserev(indexObject *self, int rev) |
|
837 | 843 | { |
|
838 | 844 | const char *data; |
|
839 | 845 | |
|
840 | 846 | if (rev >= self->length) { |
|
841 | 847 | PyObject *tuple = PyList_GET_ITEM(self->added, rev - self->length); |
|
842 | 848 | return (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 3)); |
|
843 | 849 | } |
|
844 | 850 | else { |
|
845 | 851 | data = index_deref(self, rev); |
|
846 | 852 | if (data == NULL) { |
|
847 | 853 | return -2; |
|
848 | 854 | } |
|
849 | 855 | |
|
850 | 856 | return getbe32(data + 16); |
|
851 | 857 | } |
|
852 | 858 | } |
|
853 | 859 | |
|
854 | 860 | static PyObject *index_deltachain(indexObject *self, PyObject *args) |
|
855 | 861 | { |
|
856 | 862 | int rev, generaldelta; |
|
857 | 863 | PyObject *stoparg; |
|
858 | 864 | int stoprev, iterrev, baserev = -1; |
|
859 | 865 | int stopped; |
|
860 | 866 | PyObject *chain = NULL, *result = NULL; |
|
861 | 867 | const Py_ssize_t length = index_length(self); |
|
862 | 868 | |
|
863 | 869 | if (!PyArg_ParseTuple(args, "iOi", &rev, &stoparg, &generaldelta)) { |
|
864 | 870 | return NULL; |
|
865 | 871 | } |
|
866 | 872 | |
|
867 | 873 | if (PyInt_Check(stoparg)) { |
|
868 | 874 | stoprev = (int)PyInt_AsLong(stoparg); |
|
869 | 875 | if (stoprev == -1 && PyErr_Occurred()) { |
|
870 | 876 | return NULL; |
|
871 | 877 | } |
|
872 | 878 | } |
|
873 | 879 | else if (stoparg == Py_None) { |
|
874 | 880 | stoprev = -2; |
|
875 | 881 | } |
|
876 | 882 | else { |
|
877 | 883 | PyErr_SetString(PyExc_ValueError, |
|
878 | 884 | "stoprev must be integer or None"); |
|
879 | 885 | return NULL; |
|
880 | 886 | } |
|
881 | 887 | |
|
882 | 888 | if (rev < 0 || rev >= length) { |
|
883 | 889 | PyErr_SetString(PyExc_ValueError, "revlog index out of range"); |
|
884 | 890 | return NULL; |
|
885 | 891 | } |
|
886 | 892 | |
|
887 | 893 | chain = PyList_New(0); |
|
888 | 894 | if (chain == NULL) { |
|
889 | 895 | return NULL; |
|
890 | 896 | } |
|
891 | 897 | |
|
892 | 898 | baserev = index_baserev(self, rev); |
|
893 | 899 | |
|
894 | 900 | /* This should never happen. */ |
|
895 | 901 | if (baserev <= -2) { |
|
896 | 902 | /* Error should be set by index_deref() */ |
|
897 | 903 | assert(PyErr_Occurred()); |
|
898 | 904 | goto bail; |
|
899 | 905 | } |
|
900 | 906 | |
|
901 | 907 | iterrev = rev; |
|
902 | 908 | |
|
903 | 909 | while (iterrev != baserev && iterrev != stoprev) { |
|
904 | 910 | PyObject *value = PyInt_FromLong(iterrev); |
|
905 | 911 | if (value == NULL) { |
|
906 | 912 | goto bail; |
|
907 | 913 | } |
|
908 | 914 | if (PyList_Append(chain, value)) { |
|
909 | 915 | Py_DECREF(value); |
|
910 | 916 | goto bail; |
|
911 | 917 | } |
|
912 | 918 | Py_DECREF(value); |
|
913 | 919 | |
|
914 | 920 | if (generaldelta) { |
|
915 | 921 | iterrev = baserev; |
|
916 | 922 | } |
|
917 | 923 | else { |
|
918 | 924 | iterrev--; |
|
919 | 925 | } |
|
920 | 926 | |
|
921 | 927 | if (iterrev < 0) { |
|
922 | 928 | break; |
|
923 | 929 | } |
|
924 | 930 | |
|
925 | 931 | if (iterrev >= length) { |
|
926 | 932 | PyErr_SetString(PyExc_IndexError, "revision outside index"); |
|
927 | 933 | return NULL; |
|
928 | 934 | } |
|
929 | 935 | |
|
930 | 936 | baserev = index_baserev(self, iterrev); |
|
931 | 937 | |
|
932 | 938 | /* This should never happen. */ |
|
933 | 939 | if (baserev <= -2) { |
|
934 | 940 | /* Error should be set by index_deref() */ |
|
935 | 941 | assert(PyErr_Occurred()); |
|
936 | 942 | goto bail; |
|
937 | 943 | } |
|
938 | 944 | } |
|
939 | 945 | |
|
940 | 946 | if (iterrev == stoprev) { |
|
941 | 947 | stopped = 1; |
|
942 | 948 | } |
|
943 | 949 | else { |
|
944 | 950 | PyObject *value = PyInt_FromLong(iterrev); |
|
945 | 951 | if (value == NULL) { |
|
946 | 952 | goto bail; |
|
947 | 953 | } |
|
948 | 954 | if (PyList_Append(chain, value)) { |
|
949 | 955 | Py_DECREF(value); |
|
950 | 956 | goto bail; |
|
951 | 957 | } |
|
952 | 958 | Py_DECREF(value); |
|
953 | 959 | |
|
954 | 960 | stopped = 0; |
|
955 | 961 | } |
|
956 | 962 | |
|
957 | 963 | if (PyList_Reverse(chain)) { |
|
958 | 964 | goto bail; |
|
959 | 965 | } |
|
960 | 966 | |
|
961 | 967 | result = Py_BuildValue("OO", chain, stopped ? Py_True : Py_False); |
|
962 | 968 | Py_DECREF(chain); |
|
963 | 969 | return result; |
|
964 | 970 | |
|
965 | 971 | bail: |
|
966 | 972 | Py_DECREF(chain); |
|
967 | 973 | return NULL; |
|
968 | 974 | } |
|
969 | 975 | |
|
970 | 976 | static inline int nt_level(const char *node, Py_ssize_t level) |
|
971 | 977 | { |
|
972 | 978 | int v = node[level>>1]; |
|
973 | 979 | if (!(level & 1)) |
|
974 | 980 | v >>= 4; |
|
975 | 981 | return v & 0xf; |
|
976 | 982 | } |
|
977 | 983 | |
|
978 | 984 | /* |
|
979 | 985 | * Return values: |
|
980 | 986 | * |
|
981 | 987 | * -4: match is ambiguous (multiple candidates) |
|
982 | 988 | * -2: not found |
|
983 | 989 | * rest: valid rev |
|
984 | 990 | */ |
|
985 | 991 | static int nt_find(nodetree *self, const char *node, Py_ssize_t nodelen, |
|
986 | 992 | int hex) |
|
987 | 993 | { |
|
988 | 994 | int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level; |
|
989 | 995 | int level, maxlevel, off; |
|
990 | 996 | |
|
991 | 997 | if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0) |
|
992 | 998 | return -1; |
|
993 | 999 | |
|
994 | 1000 | if (hex) |
|
995 | 1001 | maxlevel = nodelen > 40 ? 40 : (int)nodelen; |
|
996 | 1002 | else |
|
997 | 1003 | maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2); |
|
998 | 1004 | |
|
999 | 1005 | for (level = off = 0; level < maxlevel; level++) { |
|
1000 | 1006 | int k = getnybble(node, level); |
|
1001 | 1007 | nodetreenode *n = &self->nodes[off]; |
|
1002 | 1008 | int v = n->children[k]; |
|
1003 | 1009 | |
|
1004 | 1010 | if (v < 0) { |
|
1005 | 1011 | const char *n; |
|
1006 | 1012 | Py_ssize_t i; |
|
1007 | 1013 | |
|
1008 | 1014 | v = -(v + 2); |
|
1009 | 1015 | n = index_node(self->index, v); |
|
1010 | 1016 | if (n == NULL) |
|
1011 | 1017 | return -2; |
|
1012 | 1018 | for (i = level; i < maxlevel; i++) |
|
1013 | 1019 | if (getnybble(node, i) != nt_level(n, i)) |
|
1014 | 1020 | return -2; |
|
1015 | 1021 | return v; |
|
1016 | 1022 | } |
|
1017 | 1023 | if (v == 0) |
|
1018 | 1024 | return -2; |
|
1019 | 1025 | off = v; |
|
1020 | 1026 | } |
|
1021 | 1027 | /* multiple matches against an ambiguous prefix */ |
|
1022 | 1028 | return -4; |
|
1023 | 1029 | } |
|
1024 | 1030 | |
|
1025 | 1031 | static int nt_new(nodetree *self) |
|
1026 | 1032 | { |
|
1027 | 1033 | if (self->length == self->capacity) { |
|
1028 | 1034 | unsigned newcapacity; |
|
1029 | 1035 | nodetreenode *newnodes; |
|
1030 | 1036 | newcapacity = self->capacity * 2; |
|
1031 | 1037 | if (newcapacity >= INT_MAX / sizeof(nodetreenode)) { |
|
1032 | 1038 | PyErr_SetString(PyExc_MemoryError, "overflow in nt_new"); |
|
1033 | 1039 | return -1; |
|
1034 | 1040 | } |
|
1035 | 1041 | newnodes = realloc(self->nodes, newcapacity * sizeof(nodetreenode)); |
|
1036 | 1042 | if (newnodes == NULL) { |
|
1037 | 1043 | PyErr_SetString(PyExc_MemoryError, "out of memory"); |
|
1038 | 1044 | return -1; |
|
1039 | 1045 | } |
|
1040 | 1046 | self->capacity = newcapacity; |
|
1041 | 1047 | self->nodes = newnodes; |
|
1042 | 1048 | memset(&self->nodes[self->length], 0, |
|
1043 | 1049 | sizeof(nodetreenode) * (self->capacity - self->length)); |
|
1044 | 1050 | } |
|
1045 | 1051 | return self->length++; |
|
1046 | 1052 | } |
|
1047 | 1053 | |
|
1048 | 1054 | static int nt_insert(nodetree *self, const char *node, int rev) |
|
1049 | 1055 | { |
|
1050 | 1056 | int level = 0; |
|
1051 | 1057 | int off = 0; |
|
1052 | 1058 | |
|
1053 | 1059 | while (level < 40) { |
|
1054 | 1060 | int k = nt_level(node, level); |
|
1055 | 1061 | nodetreenode *n; |
|
1056 | 1062 | int v; |
|
1057 | 1063 | |
|
1058 | 1064 | n = &self->nodes[off]; |
|
1059 | 1065 | v = n->children[k]; |
|
1060 | 1066 | |
|
1061 | 1067 | if (v == 0) { |
|
1062 | 1068 | n->children[k] = -rev - 2; |
|
1063 | 1069 | return 0; |
|
1064 | 1070 | } |
|
1065 | 1071 | if (v < 0) { |
|
1066 | 1072 | const char *oldnode = index_node_existing(self->index, -(v + 2)); |
|
1067 | 1073 | int noff; |
|
1068 | 1074 | |
|
1069 | 1075 | if (oldnode == NULL) |
|
1070 | 1076 | return -1; |
|
1071 | 1077 | if (!memcmp(oldnode, node, 20)) { |
|
1072 | 1078 | n->children[k] = -rev - 2; |
|
1073 | 1079 | return 0; |
|
1074 | 1080 | } |
|
1075 | 1081 | noff = nt_new(self); |
|
1076 | 1082 | if (noff == -1) |
|
1077 | 1083 | return -1; |
|
1078 | 1084 | /* self->nodes may have been changed by realloc */ |
|
1079 | 1085 | self->nodes[off].children[k] = noff; |
|
1080 | 1086 | off = noff; |
|
1081 | 1087 | n = &self->nodes[off]; |
|
1082 | 1088 | n->children[nt_level(oldnode, ++level)] = v; |
|
1083 | 1089 | if (level > self->depth) |
|
1084 | 1090 | self->depth = level; |
|
1085 | 1091 | self->splits += 1; |
|
1086 | 1092 | } else { |
|
1087 | 1093 | level += 1; |
|
1088 | 1094 | off = v; |
|
1089 | 1095 | } |
|
1090 | 1096 | } |
|
1091 | 1097 | |
|
1092 | 1098 | return -1; |
|
1093 | 1099 | } |
|
1094 | 1100 | |
|
1095 | 1101 | static PyObject *ntobj_insert(nodetreeObject *self, PyObject *args) |
|
1096 | 1102 | { |
|
1097 | 1103 | Py_ssize_t rev; |
|
1098 | 1104 | const char *node; |
|
1099 | 1105 | Py_ssize_t length; |
|
1100 | 1106 | if (!PyArg_ParseTuple(args, "n", &rev)) |
|
1101 | 1107 | return NULL; |
|
1102 | 1108 | length = index_length(self->nt.index); |
|
1103 | 1109 | if (rev < 0 || rev >= length) { |
|
1104 | 1110 | PyErr_SetString(PyExc_ValueError, "revlog index out of range"); |
|
1105 | 1111 | return NULL; |
|
1106 | 1112 | } |
|
1107 | 1113 | node = index_node_existing(self->nt.index, rev); |
|
1108 | 1114 | if (nt_insert(&self->nt, node, (int)rev) == -1) |
|
1109 | 1115 | return NULL; |
|
1110 | 1116 | Py_RETURN_NONE; |
|
1111 | 1117 | } |
|
1112 | 1118 | |
|
1113 | 1119 | static int nt_delete_node(nodetree *self, const char *node) |
|
1114 | 1120 | { |
|
1115 | 1121 | /* rev==-2 happens to get encoded as 0, which is interpreted as not set */ |
|
1116 | 1122 | return nt_insert(self, node, -2); |
|
1117 | 1123 | } |
|
1118 | 1124 | |
|
1119 | 1125 | static int nt_init(nodetree *self, indexObject *index, unsigned capacity) |
|
1120 | 1126 | { |
|
1121 | 1127 | /* Initialize before overflow-checking to avoid nt_dealloc() crash. */ |
|
1122 | 1128 | self->nodes = NULL; |
|
1123 | 1129 | |
|
1124 | 1130 | self->index = index; |
|
1125 | 1131 | /* The input capacity is in terms of revisions, while the field is in |
|
1126 | 1132 | * terms of nodetree nodes. */ |
|
1127 | 1133 | self->capacity = (capacity < 4 ? 4 : capacity / 2); |
|
1128 | 1134 | self->depth = 0; |
|
1129 | 1135 | self->splits = 0; |
|
1130 | 1136 | if ((size_t)self->capacity > INT_MAX / sizeof(nodetreenode)) { |
|
1131 | 1137 | PyErr_SetString(PyExc_ValueError, "overflow in init_nt"); |
|
1132 | 1138 | return -1; |
|
1133 | 1139 | } |
|
1134 | 1140 | self->nodes = calloc(self->capacity, sizeof(nodetreenode)); |
|
1135 | 1141 | if (self->nodes == NULL) { |
|
1136 | 1142 | PyErr_NoMemory(); |
|
1137 | 1143 | return -1; |
|
1138 | 1144 | } |
|
1139 | 1145 | self->length = 1; |
|
1140 | 1146 | return 0; |
|
1141 | 1147 | } |
|
1142 | 1148 | |
|
1143 | 1149 | static PyTypeObject indexType; |
|
1144 | 1150 | |
|
1145 | 1151 | static int ntobj_init(nodetreeObject *self, PyObject *args) |
|
1146 | 1152 | { |
|
1147 | 1153 | PyObject *index; |
|
1148 | 1154 | unsigned capacity; |
|
1149 | 1155 | if (!PyArg_ParseTuple(args, "O!I", &indexType, &index, &capacity)) |
|
1150 | 1156 | return -1; |
|
1151 | 1157 | Py_INCREF(index); |
|
1152 | 1158 | return nt_init(&self->nt, (indexObject*)index, capacity); |
|
1153 | 1159 | } |
|
1154 | 1160 | |
|
1155 | 1161 | static int nt_partialmatch(nodetree *self, const char *node, |
|
1156 | 1162 | Py_ssize_t nodelen) |
|
1157 | 1163 | { |
|
1158 | 1164 | return nt_find(self, node, nodelen, 1); |
|
1159 | 1165 | } |
|
1160 | 1166 | |
|
1161 | 1167 | /* |
|
1162 | 1168 | * Find the length of the shortest unique prefix of node. |
|
1163 | 1169 | * |
|
1164 | 1170 | * Return values: |
|
1165 | 1171 | * |
|
1166 | 1172 | * -3: error (exception set) |
|
1167 | 1173 | * -2: not found (no exception set) |
|
1168 | 1174 | * rest: length of shortest prefix |
|
1169 | 1175 | */ |
|
1170 | 1176 | static int nt_shortest(nodetree *self, const char *node) |
|
1171 | 1177 | { |
|
1172 | 1178 | int level, off; |
|
1173 | 1179 | |
|
1174 | 1180 | for (level = off = 0; level < 40; level++) { |
|
1175 | 1181 | int k, v; |
|
1176 | 1182 | nodetreenode *n = &self->nodes[off]; |
|
1177 | 1183 | k = nt_level(node, level); |
|
1178 | 1184 | v = n->children[k]; |
|
1179 | 1185 | if (v < 0) { |
|
1180 | 1186 | const char *n; |
|
1181 | 1187 | v = -(v + 2); |
|
1182 | 1188 | n = index_node_existing(self->index, v); |
|
1183 | 1189 | if (n == NULL) |
|
1184 | 1190 | return -3; |
|
1185 | 1191 | if (memcmp(node, n, 20) != 0) |
|
1186 | 1192 | /* |
|
1187 | 1193 | * Found a unique prefix, but it wasn't for the |
|
1188 | 1194 | * requested node (i.e the requested node does |
|
1189 | 1195 | * not exist). |
|
1190 | 1196 | */ |
|
1191 | 1197 | return -2; |
|
1192 | 1198 | return level + 1; |
|
1193 | 1199 | } |
|
1194 | 1200 | if (v == 0) |
|
1195 | 1201 | return -2; |
|
1196 | 1202 | off = v; |
|
1197 | 1203 | } |
|
1198 | 1204 | /* |
|
1199 | 1205 | * The node was still not unique after 40 hex digits, so this won't |
|
1200 | 1206 | * happen. Also, if we get here, then there's a programming error in |
|
1201 | 1207 | * this file that made us insert a node longer than 40 hex digits. |
|
1202 | 1208 | */ |
|
1203 | 1209 | PyErr_SetString(PyExc_Exception, "broken node tree"); |
|
1204 | 1210 | return -3; |
|
1205 | 1211 | } |
|
1206 | 1212 | |
|
1207 | 1213 | static PyObject *ntobj_shortest(nodetreeObject *self, PyObject *args) |
|
1208 | 1214 | { |
|
1209 | 1215 | PyObject *val; |
|
1210 | 1216 | char *node; |
|
1211 | 1217 | int length; |
|
1212 | 1218 | |
|
1213 | 1219 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
1214 | 1220 | return NULL; |
|
1215 | 1221 | if (node_check(val, &node) == -1) |
|
1216 | 1222 | return NULL; |
|
1217 | 1223 | |
|
1218 | 1224 | length = nt_shortest(&self->nt, node); |
|
1219 | 1225 | if (length == -3) |
|
1220 | 1226 | return NULL; |
|
1221 | 1227 | if (length == -2) { |
|
1222 | 1228 | raise_revlog_error(); |
|
1223 | 1229 | return NULL; |
|
1224 | 1230 | } |
|
1225 | 1231 | return PyInt_FromLong(length); |
|
1226 | 1232 | } |
|
1227 | 1233 | |
|
1228 | 1234 | static void nt_dealloc(nodetree *self) |
|
1229 | 1235 | { |
|
1230 | 1236 | free(self->nodes); |
|
1231 | 1237 | self->nodes = NULL; |
|
1232 | 1238 | } |
|
1233 | 1239 | |
|
1234 | 1240 | static void ntobj_dealloc(nodetreeObject *self) |
|
1235 | 1241 | { |
|
1236 | 1242 | Py_XDECREF(self->nt.index); |
|
1237 | 1243 | nt_dealloc(&self->nt); |
|
1238 | 1244 | PyObject_Del(self); |
|
1239 | 1245 | } |
|
1240 | 1246 | |
|
1241 | 1247 | static PyMethodDef ntobj_methods[] = { |
|
1242 | 1248 | {"insert", (PyCFunction)ntobj_insert, METH_VARARGS, |
|
1243 | 1249 | "insert an index entry"}, |
|
1244 | 1250 | {"shortest", (PyCFunction)ntobj_shortest, METH_VARARGS, |
|
1245 | 1251 | "find length of shortest hex nodeid of a binary ID"}, |
|
1246 | 1252 | {NULL} /* Sentinel */ |
|
1247 | 1253 | }; |
|
1248 | 1254 | |
|
1249 | 1255 | static PyTypeObject nodetreeType = { |
|
1250 | 1256 | PyVarObject_HEAD_INIT(NULL, 0) /* header */ |
|
1251 | 1257 | "parsers.nodetree", /* tp_name */ |
|
1252 | 1258 | sizeof(nodetreeObject) , /* tp_basicsize */ |
|
1253 | 1259 | 0, /* tp_itemsize */ |
|
1254 | 1260 | (destructor)ntobj_dealloc, /* tp_dealloc */ |
|
1255 | 1261 | 0, /* tp_print */ |
|
1256 | 1262 | 0, /* tp_getattr */ |
|
1257 | 1263 | 0, /* tp_setattr */ |
|
1258 | 1264 | 0, /* tp_compare */ |
|
1259 | 1265 | 0, /* tp_repr */ |
|
1260 | 1266 | 0, /* tp_as_number */ |
|
1261 | 1267 | 0, /* tp_as_sequence */ |
|
1262 | 1268 | 0, /* tp_as_mapping */ |
|
1263 | 1269 | 0, /* tp_hash */ |
|
1264 | 1270 | 0, /* tp_call */ |
|
1265 | 1271 | 0, /* tp_str */ |
|
1266 | 1272 | 0, /* tp_getattro */ |
|
1267 | 1273 | 0, /* tp_setattro */ |
|
1268 | 1274 | 0, /* tp_as_buffer */ |
|
1269 | 1275 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
1270 | 1276 | "nodetree", /* tp_doc */ |
|
1271 | 1277 | 0, /* tp_traverse */ |
|
1272 | 1278 | 0, /* tp_clear */ |
|
1273 | 1279 | 0, /* tp_richcompare */ |
|
1274 | 1280 | 0, /* tp_weaklistoffset */ |
|
1275 | 1281 | 0, /* tp_iter */ |
|
1276 | 1282 | 0, /* tp_iternext */ |
|
1277 | 1283 | ntobj_methods, /* tp_methods */ |
|
1278 | 1284 | 0, /* tp_members */ |
|
1279 | 1285 | 0, /* tp_getset */ |
|
1280 | 1286 | 0, /* tp_base */ |
|
1281 | 1287 | 0, /* tp_dict */ |
|
1282 | 1288 | 0, /* tp_descr_get */ |
|
1283 | 1289 | 0, /* tp_descr_set */ |
|
1284 | 1290 | 0, /* tp_dictoffset */ |
|
1285 | 1291 | (initproc)ntobj_init, /* tp_init */ |
|
1286 | 1292 | 0, /* tp_alloc */ |
|
1287 | 1293 | }; |
|
1288 | 1294 | |
|
1289 | 1295 | static int index_init_nt(indexObject *self) |
|
1290 | 1296 | { |
|
1291 | 1297 | if (!self->ntinitialized) { |
|
1292 | 1298 | if (nt_init(&self->nt, self, (int)self->raw_length) == -1) { |
|
1293 | 1299 | nt_dealloc(&self->nt); |
|
1294 | 1300 | return -1; |
|
1295 | 1301 | } |
|
1296 | 1302 | if (nt_insert(&self->nt, nullid, -1) == -1) { |
|
1297 | 1303 | nt_dealloc(&self->nt); |
|
1298 | 1304 | return -1; |
|
1299 | 1305 | } |
|
1300 | 1306 | self->ntinitialized = 1; |
|
1301 | 1307 | self->ntrev = (int)index_length(self); |
|
1302 | 1308 | self->ntlookups = 1; |
|
1303 | 1309 | self->ntmisses = 0; |
|
1304 | 1310 | } |
|
1305 | 1311 | return 0; |
|
1306 | 1312 | } |
|
1307 | 1313 | |
|
1308 | 1314 | /* |
|
1309 | 1315 | * Return values: |
|
1310 | 1316 | * |
|
1311 | 1317 | * -3: error (exception set) |
|
1312 | 1318 | * -2: not found (no exception set) |
|
1313 | 1319 | * rest: valid rev |
|
1314 | 1320 | */ |
|
1315 | 1321 | static int index_find_node(indexObject *self, |
|
1316 | 1322 | const char *node, Py_ssize_t nodelen) |
|
1317 | 1323 | { |
|
1318 | 1324 | int rev; |
|
1319 | 1325 | |
|
1320 | 1326 | if (index_init_nt(self) == -1) |
|
1321 | 1327 | return -3; |
|
1322 | 1328 | |
|
1323 | 1329 | self->ntlookups++; |
|
1324 | 1330 | rev = nt_find(&self->nt, node, nodelen, 0); |
|
1325 | 1331 | if (rev >= -1) |
|
1326 | 1332 | return rev; |
|
1327 | 1333 | |
|
1328 | 1334 | /* |
|
1329 | 1335 | * For the first handful of lookups, we scan the entire index, |
|
1330 | 1336 | * and cache only the matching nodes. This optimizes for cases |
|
1331 | 1337 | * like "hg tip", where only a few nodes are accessed. |
|
1332 | 1338 | * |
|
1333 | 1339 | * After that, we cache every node we visit, using a single |
|
1334 | 1340 | * scan amortized over multiple lookups. This gives the best |
|
1335 | 1341 | * bulk performance, e.g. for "hg log". |
|
1336 | 1342 | */ |
|
1337 | 1343 | if (self->ntmisses++ < 4) { |
|
1338 | 1344 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
1339 | 1345 | const char *n = index_node_existing(self, rev); |
|
1340 | 1346 | if (n == NULL) |
|
1341 | 1347 | return -3; |
|
1342 | 1348 | if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) { |
|
1343 | 1349 | if (nt_insert(&self->nt, n, rev) == -1) |
|
1344 | 1350 | return -3; |
|
1345 | 1351 | break; |
|
1346 | 1352 | } |
|
1347 | 1353 | } |
|
1348 | 1354 | } else { |
|
1349 | 1355 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
1350 | 1356 | const char *n = index_node_existing(self, rev); |
|
1351 | 1357 | if (n == NULL) |
|
1352 | 1358 | return -3; |
|
1353 | 1359 | if (nt_insert(&self->nt, n, rev) == -1) { |
|
1354 | 1360 | self->ntrev = rev + 1; |
|
1355 | 1361 | return -3; |
|
1356 | 1362 | } |
|
1357 | 1363 | if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) { |
|
1358 | 1364 | break; |
|
1359 | 1365 | } |
|
1360 | 1366 | } |
|
1361 | 1367 | self->ntrev = rev; |
|
1362 | 1368 | } |
|
1363 | 1369 | |
|
1364 | 1370 | if (rev >= 0) |
|
1365 | 1371 | return rev; |
|
1366 | 1372 | return -2; |
|
1367 | 1373 | } |
|
1368 | 1374 | |
|
1369 | 1375 | static PyObject *index_getitem(indexObject *self, PyObject *value) |
|
1370 | 1376 | { |
|
1371 | 1377 | char *node; |
|
1372 | 1378 | int rev; |
|
1373 | 1379 | |
|
1374 | 1380 | if (PyInt_Check(value)) |
|
1375 | 1381 | return index_get(self, PyInt_AS_LONG(value)); |
|
1376 | 1382 | |
|
1377 | 1383 | if (node_check(value, &node) == -1) |
|
1378 | 1384 | return NULL; |
|
1379 | 1385 | rev = index_find_node(self, node, 20); |
|
1380 | 1386 | if (rev >= -1) |
|
1381 | 1387 | return PyInt_FromLong(rev); |
|
1382 | 1388 | if (rev == -2) |
|
1383 | 1389 | raise_revlog_error(); |
|
1384 | 1390 | return NULL; |
|
1385 | 1391 | } |
|
1386 | 1392 | |
|
1387 | 1393 | /* |
|
1388 | 1394 | * Fully populate the radix tree. |
|
1389 | 1395 | */ |
|
1390 | 1396 | static int index_populate_nt(indexObject *self) { |
|
1391 | 1397 | int rev; |
|
1392 | 1398 | if (self->ntrev > 0) { |
|
1393 | 1399 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
1394 | 1400 | const char *n = index_node_existing(self, rev); |
|
1395 | 1401 | if (n == NULL) |
|
1396 | 1402 | return -1; |
|
1397 | 1403 | if (nt_insert(&self->nt, n, rev) == -1) |
|
1398 | 1404 | return -1; |
|
1399 | 1405 | } |
|
1400 | 1406 | self->ntrev = -1; |
|
1401 | 1407 | } |
|
1402 | 1408 | return 0; |
|
1403 | 1409 | } |
|
1404 | 1410 | |
|
1405 | 1411 | static PyObject *index_partialmatch(indexObject *self, PyObject *args) |
|
1406 | 1412 | { |
|
1407 | 1413 | const char *fullnode; |
|
1408 | 1414 | int nodelen; |
|
1409 | 1415 | char *node; |
|
1410 | 1416 | int rev, i; |
|
1411 | 1417 | |
|
1412 | 1418 | if (!PyArg_ParseTuple(args, PY23("s#", "y#"), &node, &nodelen)) |
|
1413 | 1419 | return NULL; |
|
1414 | 1420 | |
|
1415 | 1421 | if (nodelen < 1) { |
|
1416 | 1422 | PyErr_SetString(PyExc_ValueError, "key too short"); |
|
1417 | 1423 | return NULL; |
|
1418 | 1424 | } |
|
1419 | 1425 | |
|
1420 | 1426 | if (nodelen > 40) { |
|
1421 | 1427 | PyErr_SetString(PyExc_ValueError, "key too long"); |
|
1422 | 1428 | return NULL; |
|
1423 | 1429 | } |
|
1424 | 1430 | |
|
1425 | 1431 | for (i = 0; i < nodelen; i++) |
|
1426 | 1432 | hexdigit(node, i); |
|
1427 | 1433 | if (PyErr_Occurred()) { |
|
1428 | 1434 | /* input contains non-hex characters */ |
|
1429 | 1435 | PyErr_Clear(); |
|
1430 | 1436 | Py_RETURN_NONE; |
|
1431 | 1437 | } |
|
1432 | 1438 | |
|
1433 | 1439 | if (index_init_nt(self) == -1) |
|
1434 | 1440 | return NULL; |
|
1435 | 1441 | if (index_populate_nt(self) == -1) |
|
1436 | 1442 | return NULL; |
|
1437 | 1443 | rev = nt_partialmatch(&self->nt, node, nodelen); |
|
1438 | 1444 | |
|
1439 | 1445 | switch (rev) { |
|
1440 | 1446 | case -4: |
|
1441 | 1447 | raise_revlog_error(); |
|
1442 | 1448 | return NULL; |
|
1443 | 1449 | case -2: |
|
1444 | 1450 | Py_RETURN_NONE; |
|
1445 | 1451 | case -1: |
|
1446 | 1452 | return PyBytes_FromStringAndSize(nullid, 20); |
|
1447 | 1453 | } |
|
1448 | 1454 | |
|
1449 | 1455 | fullnode = index_node_existing(self, rev); |
|
1450 | 1456 | if (fullnode == NULL) { |
|
1451 | 1457 | return NULL; |
|
1452 | 1458 | } |
|
1453 | 1459 | return PyBytes_FromStringAndSize(fullnode, 20); |
|
1454 | 1460 | } |
|
1455 | 1461 | |
|
1456 | 1462 | static PyObject *index_shortest(indexObject *self, PyObject *args) |
|
1457 | 1463 | { |
|
1458 | 1464 | PyObject *val; |
|
1459 | 1465 | char *node; |
|
1460 | 1466 | int length; |
|
1461 | 1467 | |
|
1462 | 1468 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
1463 | 1469 | return NULL; |
|
1464 | 1470 | if (node_check(val, &node) == -1) |
|
1465 | 1471 | return NULL; |
|
1466 | 1472 | |
|
1467 | 1473 | self->ntlookups++; |
|
1468 | 1474 | if (index_init_nt(self) == -1) |
|
1469 | 1475 | return NULL; |
|
1470 | 1476 | if (index_populate_nt(self) == -1) |
|
1471 | 1477 | return NULL; |
|
1472 | 1478 | length = nt_shortest(&self->nt, node); |
|
1473 | 1479 | if (length == -3) |
|
1474 | 1480 | return NULL; |
|
1475 | 1481 | if (length == -2) { |
|
1476 | 1482 | raise_revlog_error(); |
|
1477 | 1483 | return NULL; |
|
1478 | 1484 | } |
|
1479 | 1485 | return PyInt_FromLong(length); |
|
1480 | 1486 | } |
|
1481 | 1487 | |
|
1482 | 1488 | static PyObject *index_m_get(indexObject *self, PyObject *args) |
|
1483 | 1489 | { |
|
1484 | 1490 | PyObject *val; |
|
1485 | 1491 | char *node; |
|
1486 | 1492 | int rev; |
|
1487 | 1493 | |
|
1488 | 1494 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
1489 | 1495 | return NULL; |
|
1490 | 1496 | if (node_check(val, &node) == -1) |
|
1491 | 1497 | return NULL; |
|
1492 | 1498 | rev = index_find_node(self, node, 20); |
|
1493 | 1499 | if (rev == -3) |
|
1494 | 1500 | return NULL; |
|
1495 | 1501 | if (rev == -2) |
|
1496 | 1502 | Py_RETURN_NONE; |
|
1497 | 1503 | return PyInt_FromLong(rev); |
|
1498 | 1504 | } |
|
1499 | 1505 | |
|
1500 | 1506 | static int index_contains(indexObject *self, PyObject *value) |
|
1501 | 1507 | { |
|
1502 | 1508 | char *node; |
|
1503 | 1509 | |
|
1504 | 1510 | if (PyInt_Check(value)) { |
|
1505 | 1511 | long rev = PyInt_AS_LONG(value); |
|
1506 | 1512 | return rev >= -1 && rev < index_length(self); |
|
1507 | 1513 | } |
|
1508 | 1514 | |
|
1509 | 1515 | if (node_check(value, &node) == -1) |
|
1510 | 1516 | return -1; |
|
1511 | 1517 | |
|
1512 | 1518 | switch (index_find_node(self, node, 20)) { |
|
1513 | 1519 | case -3: |
|
1514 | 1520 | return -1; |
|
1515 | 1521 | case -2: |
|
1516 | 1522 | return 0; |
|
1517 | 1523 | default: |
|
1518 | 1524 | return 1; |
|
1519 | 1525 | } |
|
1520 | 1526 | } |
|
1521 | 1527 | |
|
1522 | 1528 | typedef uint64_t bitmask; |
|
1523 | 1529 | |
|
1524 | 1530 | /* |
|
1525 | 1531 | * Given a disjoint set of revs, return all candidates for the |
|
1526 | 1532 | * greatest common ancestor. In revset notation, this is the set |
|
1527 | 1533 | * "heads(::a and ::b and ...)" |
|
1528 | 1534 | */ |
|
1529 | 1535 | static PyObject *find_gca_candidates(indexObject *self, const int *revs, |
|
1530 | 1536 | int revcount) |
|
1531 | 1537 | { |
|
1532 | 1538 | const bitmask allseen = (1ull << revcount) - 1; |
|
1533 | 1539 | const bitmask poison = 1ull << revcount; |
|
1534 | 1540 | PyObject *gca = PyList_New(0); |
|
1535 | 1541 | int i, v, interesting; |
|
1536 | 1542 | int maxrev = -1; |
|
1537 | 1543 | bitmask sp; |
|
1538 | 1544 | bitmask *seen; |
|
1539 | 1545 | |
|
1540 | 1546 | if (gca == NULL) |
|
1541 | 1547 | return PyErr_NoMemory(); |
|
1542 | 1548 | |
|
1543 | 1549 | for (i = 0; i < revcount; i++) { |
|
1544 | 1550 | if (revs[i] > maxrev) |
|
1545 | 1551 | maxrev = revs[i]; |
|
1546 | 1552 | } |
|
1547 | 1553 | |
|
1548 | 1554 | seen = calloc(sizeof(*seen), maxrev + 1); |
|
1549 | 1555 | if (seen == NULL) { |
|
1550 | 1556 | Py_DECREF(gca); |
|
1551 | 1557 | return PyErr_NoMemory(); |
|
1552 | 1558 | } |
|
1553 | 1559 | |
|
1554 | 1560 | for (i = 0; i < revcount; i++) |
|
1555 | 1561 | seen[revs[i]] = 1ull << i; |
|
1556 | 1562 | |
|
1557 | 1563 | interesting = revcount; |
|
1558 | 1564 | |
|
1559 | 1565 | for (v = maxrev; v >= 0 && interesting; v--) { |
|
1560 | 1566 | bitmask sv = seen[v]; |
|
1561 | 1567 | int parents[2]; |
|
1562 | 1568 | |
|
1563 | 1569 | if (!sv) |
|
1564 | 1570 | continue; |
|
1565 | 1571 | |
|
1566 | 1572 | if (sv < poison) { |
|
1567 | 1573 | interesting -= 1; |
|
1568 | 1574 | if (sv == allseen) { |
|
1569 | 1575 | PyObject *obj = PyInt_FromLong(v); |
|
1570 | 1576 | if (obj == NULL) |
|
1571 | 1577 | goto bail; |
|
1572 | 1578 | if (PyList_Append(gca, obj) == -1) { |
|
1573 | 1579 | Py_DECREF(obj); |
|
1574 | 1580 | goto bail; |
|
1575 | 1581 | } |
|
1576 | 1582 | sv |= poison; |
|
1577 | 1583 | for (i = 0; i < revcount; i++) { |
|
1578 | 1584 | if (revs[i] == v) |
|
1579 | 1585 | goto done; |
|
1580 | 1586 | } |
|
1581 | 1587 | } |
|
1582 | 1588 | } |
|
1583 | 1589 | if (index_get_parents(self, v, parents, maxrev) < 0) |
|
1584 | 1590 | goto bail; |
|
1585 | 1591 | |
|
1586 | 1592 | for (i = 0; i < 2; i++) { |
|
1587 | 1593 | int p = parents[i]; |
|
1588 | 1594 | if (p == -1) |
|
1589 | 1595 | continue; |
|
1590 | 1596 | sp = seen[p]; |
|
1591 | 1597 | if (sv < poison) { |
|
1592 | 1598 | if (sp == 0) { |
|
1593 | 1599 | seen[p] = sv; |
|
1594 | 1600 | interesting++; |
|
1595 | 1601 | } |
|
1596 | 1602 | else if (sp != sv) |
|
1597 | 1603 | seen[p] |= sv; |
|
1598 | 1604 | } else { |
|
1599 | 1605 | if (sp && sp < poison) |
|
1600 | 1606 | interesting--; |
|
1601 | 1607 | seen[p] = sv; |
|
1602 | 1608 | } |
|
1603 | 1609 | } |
|
1604 | 1610 | } |
|
1605 | 1611 | |
|
1606 | 1612 | done: |
|
1607 | 1613 | free(seen); |
|
1608 | 1614 | return gca; |
|
1609 | 1615 | bail: |
|
1610 | 1616 | free(seen); |
|
1611 | 1617 | Py_XDECREF(gca); |
|
1612 | 1618 | return NULL; |
|
1613 | 1619 | } |
|
1614 | 1620 | |
|
1615 | 1621 | /* |
|
1616 | 1622 | * Given a disjoint set of revs, return the subset with the longest |
|
1617 | 1623 | * path to the root. |
|
1618 | 1624 | */ |
|
1619 | 1625 | static PyObject *find_deepest(indexObject *self, PyObject *revs) |
|
1620 | 1626 | { |
|
1621 | 1627 | const Py_ssize_t revcount = PyList_GET_SIZE(revs); |
|
1622 | 1628 | static const Py_ssize_t capacity = 24; |
|
1623 | 1629 | int *depth, *interesting = NULL; |
|
1624 | 1630 | int i, j, v, ninteresting; |
|
1625 | 1631 | PyObject *dict = NULL, *keys = NULL; |
|
1626 | 1632 | long *seen = NULL; |
|
1627 | 1633 | int maxrev = -1; |
|
1628 | 1634 | long final; |
|
1629 | 1635 | |
|
1630 | 1636 | if (revcount > capacity) { |
|
1631 | 1637 | PyErr_Format(PyExc_OverflowError, |
|
1632 | 1638 | "bitset size (%ld) > capacity (%ld)", |
|
1633 | 1639 | (long)revcount, (long)capacity); |
|
1634 | 1640 | return NULL; |
|
1635 | 1641 | } |
|
1636 | 1642 | |
|
1637 | 1643 | for (i = 0; i < revcount; i++) { |
|
1638 | 1644 | int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i)); |
|
1639 | 1645 | if (n > maxrev) |
|
1640 | 1646 | maxrev = n; |
|
1641 | 1647 | } |
|
1642 | 1648 | |
|
1643 | 1649 | depth = calloc(sizeof(*depth), maxrev + 1); |
|
1644 | 1650 | if (depth == NULL) |
|
1645 | 1651 | return PyErr_NoMemory(); |
|
1646 | 1652 | |
|
1647 | 1653 | seen = calloc(sizeof(*seen), maxrev + 1); |
|
1648 | 1654 | if (seen == NULL) { |
|
1649 | 1655 | PyErr_NoMemory(); |
|
1650 | 1656 | goto bail; |
|
1651 | 1657 | } |
|
1652 | 1658 | |
|
1653 | 1659 | interesting = calloc(sizeof(*interesting), ((size_t)1) << revcount); |
|
1654 | 1660 | if (interesting == NULL) { |
|
1655 | 1661 | PyErr_NoMemory(); |
|
1656 | 1662 | goto bail; |
|
1657 | 1663 | } |
|
1658 | 1664 | |
|
1659 | 1665 | if (PyList_Sort(revs) == -1) |
|
1660 | 1666 | goto bail; |
|
1661 | 1667 | |
|
1662 | 1668 | for (i = 0; i < revcount; i++) { |
|
1663 | 1669 | int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i)); |
|
1664 | 1670 | long b = 1l << i; |
|
1665 | 1671 | depth[n] = 1; |
|
1666 | 1672 | seen[n] = b; |
|
1667 | 1673 | interesting[b] = 1; |
|
1668 | 1674 | } |
|
1669 | 1675 | |
|
1670 | 1676 | /* invariant: ninteresting is the number of non-zero entries in |
|
1671 | 1677 | * interesting. */ |
|
1672 | 1678 | ninteresting = (int)revcount; |
|
1673 | 1679 | |
|
1674 | 1680 | for (v = maxrev; v >= 0 && ninteresting > 1; v--) { |
|
1675 | 1681 | int dv = depth[v]; |
|
1676 | 1682 | int parents[2]; |
|
1677 | 1683 | long sv; |
|
1678 | 1684 | |
|
1679 | 1685 | if (dv == 0) |
|
1680 | 1686 | continue; |
|
1681 | 1687 | |
|
1682 | 1688 | sv = seen[v]; |
|
1683 | 1689 | if (index_get_parents(self, v, parents, maxrev) < 0) |
|
1684 | 1690 | goto bail; |
|
1685 | 1691 | |
|
1686 | 1692 | for (i = 0; i < 2; i++) { |
|
1687 | 1693 | int p = parents[i]; |
|
1688 | 1694 | long sp; |
|
1689 | 1695 | int dp; |
|
1690 | 1696 | |
|
1691 | 1697 | if (p == -1) |
|
1692 | 1698 | continue; |
|
1693 | 1699 | |
|
1694 | 1700 | dp = depth[p]; |
|
1695 | 1701 | sp = seen[p]; |
|
1696 | 1702 | if (dp <= dv) { |
|
1697 | 1703 | depth[p] = dv + 1; |
|
1698 | 1704 | if (sp != sv) { |
|
1699 | 1705 | interesting[sv] += 1; |
|
1700 | 1706 | seen[p] = sv; |
|
1701 | 1707 | if (sp) { |
|
1702 | 1708 | interesting[sp] -= 1; |
|
1703 | 1709 | if (interesting[sp] == 0) |
|
1704 | 1710 | ninteresting -= 1; |
|
1705 | 1711 | } |
|
1706 | 1712 | } |
|
1707 | 1713 | } |
|
1708 | 1714 | else if (dv == dp - 1) { |
|
1709 | 1715 | long nsp = sp | sv; |
|
1710 | 1716 | if (nsp == sp) |
|
1711 | 1717 | continue; |
|
1712 | 1718 | seen[p] = nsp; |
|
1713 | 1719 | interesting[sp] -= 1; |
|
1714 | 1720 | if (interesting[sp] == 0) |
|
1715 | 1721 | ninteresting -= 1; |
|
1716 | 1722 | if (interesting[nsp] == 0) |
|
1717 | 1723 | ninteresting += 1; |
|
1718 | 1724 | interesting[nsp] += 1; |
|
1719 | 1725 | } |
|
1720 | 1726 | } |
|
1721 | 1727 | interesting[sv] -= 1; |
|
1722 | 1728 | if (interesting[sv] == 0) |
|
1723 | 1729 | ninteresting -= 1; |
|
1724 | 1730 | } |
|
1725 | 1731 | |
|
1726 | 1732 | final = 0; |
|
1727 | 1733 | j = ninteresting; |
|
1728 | 1734 | for (i = 0; i < (int)(2 << revcount) && j > 0; i++) { |
|
1729 | 1735 | if (interesting[i] == 0) |
|
1730 | 1736 | continue; |
|
1731 | 1737 | final |= i; |
|
1732 | 1738 | j -= 1; |
|
1733 | 1739 | } |
|
1734 | 1740 | if (final == 0) { |
|
1735 | 1741 | keys = PyList_New(0); |
|
1736 | 1742 | goto bail; |
|
1737 | 1743 | } |
|
1738 | 1744 | |
|
1739 | 1745 | dict = PyDict_New(); |
|
1740 | 1746 | if (dict == NULL) |
|
1741 | 1747 | goto bail; |
|
1742 | 1748 | |
|
1743 | 1749 | for (i = 0; i < revcount; i++) { |
|
1744 | 1750 | PyObject *key; |
|
1745 | 1751 | |
|
1746 | 1752 | if ((final & (1 << i)) == 0) |
|
1747 | 1753 | continue; |
|
1748 | 1754 | |
|
1749 | 1755 | key = PyList_GET_ITEM(revs, i); |
|
1750 | 1756 | Py_INCREF(key); |
|
1751 | 1757 | Py_INCREF(Py_None); |
|
1752 | 1758 | if (PyDict_SetItem(dict, key, Py_None) == -1) { |
|
1753 | 1759 | Py_DECREF(key); |
|
1754 | 1760 | Py_DECREF(Py_None); |
|
1755 | 1761 | goto bail; |
|
1756 | 1762 | } |
|
1757 | 1763 | } |
|
1758 | 1764 | |
|
1759 | 1765 | keys = PyDict_Keys(dict); |
|
1760 | 1766 | |
|
1761 | 1767 | bail: |
|
1762 | 1768 | free(depth); |
|
1763 | 1769 | free(seen); |
|
1764 | 1770 | free(interesting); |
|
1765 | 1771 | Py_XDECREF(dict); |
|
1766 | 1772 | |
|
1767 | 1773 | return keys; |
|
1768 | 1774 | } |
|
1769 | 1775 | |
|
1770 | 1776 | /* |
|
1771 | 1777 | * Given a (possibly overlapping) set of revs, return all the |
|
1772 | 1778 | * common ancestors heads: heads(::args[0] and ::a[1] and ...) |
|
1773 | 1779 | */ |
|
1774 | 1780 | static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args) |
|
1775 | 1781 | { |
|
1776 | 1782 | PyObject *ret = NULL; |
|
1777 | 1783 | Py_ssize_t argcount, i, len; |
|
1778 | 1784 | bitmask repeat = 0; |
|
1779 | 1785 | int revcount = 0; |
|
1780 | 1786 | int *revs; |
|
1781 | 1787 | |
|
1782 | 1788 | argcount = PySequence_Length(args); |
|
1783 | 1789 | revs = PyMem_Malloc(argcount * sizeof(*revs)); |
|
1784 | 1790 | if (argcount > 0 && revs == NULL) |
|
1785 | 1791 | return PyErr_NoMemory(); |
|
1786 | 1792 | len = index_length(self); |
|
1787 | 1793 | |
|
1788 | 1794 | for (i = 0; i < argcount; i++) { |
|
1789 | 1795 | static const int capacity = 24; |
|
1790 | 1796 | PyObject *obj = PySequence_GetItem(args, i); |
|
1791 | 1797 | bitmask x; |
|
1792 | 1798 | long val; |
|
1793 | 1799 | |
|
1794 | 1800 | if (!PyInt_Check(obj)) { |
|
1795 | 1801 | PyErr_SetString(PyExc_TypeError, |
|
1796 | 1802 | "arguments must all be ints"); |
|
1797 | 1803 | Py_DECREF(obj); |
|
1798 | 1804 | goto bail; |
|
1799 | 1805 | } |
|
1800 | 1806 | val = PyInt_AsLong(obj); |
|
1801 | 1807 | Py_DECREF(obj); |
|
1802 | 1808 | if (val == -1) { |
|
1803 | 1809 | ret = PyList_New(0); |
|
1804 | 1810 | goto done; |
|
1805 | 1811 | } |
|
1806 | 1812 | if (val < 0 || val >= len) { |
|
1807 | 1813 | PyErr_SetString(PyExc_IndexError, |
|
1808 | 1814 | "index out of range"); |
|
1809 | 1815 | goto bail; |
|
1810 | 1816 | } |
|
1811 | 1817 | /* this cheesy bloom filter lets us avoid some more |
|
1812 | 1818 | * expensive duplicate checks in the common set-is-disjoint |
|
1813 | 1819 | * case */ |
|
1814 | 1820 | x = 1ull << (val & 0x3f); |
|
1815 | 1821 | if (repeat & x) { |
|
1816 | 1822 | int k; |
|
1817 | 1823 | for (k = 0; k < revcount; k++) { |
|
1818 | 1824 | if (val == revs[k]) |
|
1819 | 1825 | goto duplicate; |
|
1820 | 1826 | } |
|
1821 | 1827 | } |
|
1822 | 1828 | else repeat |= x; |
|
1823 | 1829 | if (revcount >= capacity) { |
|
1824 | 1830 | PyErr_Format(PyExc_OverflowError, |
|
1825 | 1831 | "bitset size (%d) > capacity (%d)", |
|
1826 | 1832 | revcount, capacity); |
|
1827 | 1833 | goto bail; |
|
1828 | 1834 | } |
|
1829 | 1835 | revs[revcount++] = (int)val; |
|
1830 | 1836 | duplicate:; |
|
1831 | 1837 | } |
|
1832 | 1838 | |
|
1833 | 1839 | if (revcount == 0) { |
|
1834 | 1840 | ret = PyList_New(0); |
|
1835 | 1841 | goto done; |
|
1836 | 1842 | } |
|
1837 | 1843 | if (revcount == 1) { |
|
1838 | 1844 | PyObject *obj; |
|
1839 | 1845 | ret = PyList_New(1); |
|
1840 | 1846 | if (ret == NULL) |
|
1841 | 1847 | goto bail; |
|
1842 | 1848 | obj = PyInt_FromLong(revs[0]); |
|
1843 | 1849 | if (obj == NULL) |
|
1844 | 1850 | goto bail; |
|
1845 | 1851 | PyList_SET_ITEM(ret, 0, obj); |
|
1846 | 1852 | goto done; |
|
1847 | 1853 | } |
|
1848 | 1854 | |
|
1849 | 1855 | ret = find_gca_candidates(self, revs, revcount); |
|
1850 | 1856 | if (ret == NULL) |
|
1851 | 1857 | goto bail; |
|
1852 | 1858 | |
|
1853 | 1859 | done: |
|
1854 | 1860 | PyMem_Free(revs); |
|
1855 | 1861 | return ret; |
|
1856 | 1862 | |
|
1857 | 1863 | bail: |
|
1858 | 1864 | PyMem_Free(revs); |
|
1859 | 1865 | Py_XDECREF(ret); |
|
1860 | 1866 | return NULL; |
|
1861 | 1867 | } |
|
1862 | 1868 | |
|
1863 | 1869 | /* |
|
1864 | 1870 | * Given a (possibly overlapping) set of revs, return the greatest |
|
1865 | 1871 | * common ancestors: those with the longest path to the root. |
|
1866 | 1872 | */ |
|
1867 | 1873 | static PyObject *index_ancestors(indexObject *self, PyObject *args) |
|
1868 | 1874 | { |
|
1869 | 1875 | PyObject *ret; |
|
1870 | 1876 | PyObject *gca = index_commonancestorsheads(self, args); |
|
1871 | 1877 | if (gca == NULL) |
|
1872 | 1878 | return NULL; |
|
1873 | 1879 | |
|
1874 | 1880 | if (PyList_GET_SIZE(gca) <= 1) { |
|
1875 | 1881 | return gca; |
|
1876 | 1882 | } |
|
1877 | 1883 | |
|
1878 | 1884 | ret = find_deepest(self, gca); |
|
1879 | 1885 | Py_DECREF(gca); |
|
1880 | 1886 | return ret; |
|
1881 | 1887 | } |
|
1882 | 1888 | |
|
1883 | 1889 | /* |
|
1884 | 1890 | * Invalidate any trie entries introduced by added revs. |
|
1885 | 1891 | */ |
|
1886 | 1892 | static void index_invalidate_added(indexObject *self, Py_ssize_t start) |
|
1887 | 1893 | { |
|
1888 | 1894 | Py_ssize_t i, len = PyList_GET_SIZE(self->added); |
|
1889 | 1895 | |
|
1890 | 1896 | for (i = start; i < len; i++) { |
|
1891 | 1897 | PyObject *tuple = PyList_GET_ITEM(self->added, i); |
|
1892 | 1898 | PyObject *node = PyTuple_GET_ITEM(tuple, 7); |
|
1893 | 1899 | |
|
1894 | 1900 | nt_delete_node(&self->nt, PyBytes_AS_STRING(node)); |
|
1895 | 1901 | } |
|
1896 | 1902 | |
|
1897 | 1903 | if (start == 0) |
|
1898 | 1904 | Py_CLEAR(self->added); |
|
1899 | 1905 | } |
|
1900 | 1906 | |
|
1901 | 1907 | /* |
|
1902 | 1908 | * Delete a numeric range of revs, which must be at the end of the |
|
1903 | 1909 | * range, but exclude the sentinel nullid entry. |
|
1904 | 1910 | */ |
|
1905 | 1911 | static int index_slice_del(indexObject *self, PyObject *item) |
|
1906 | 1912 | { |
|
1907 | 1913 | Py_ssize_t start, stop, step, slicelength; |
|
1908 | 1914 | Py_ssize_t length = index_length(self) + 1; |
|
1909 | 1915 | int ret = 0; |
|
1910 | 1916 | |
|
1911 | 1917 | /* Argument changed from PySliceObject* to PyObject* in Python 3. */ |
|
1912 | 1918 | #ifdef IS_PY3K |
|
1913 | 1919 | if (PySlice_GetIndicesEx(item, length, |
|
1914 | 1920 | &start, &stop, &step, &slicelength) < 0) |
|
1915 | 1921 | #else |
|
1916 | 1922 | if (PySlice_GetIndicesEx((PySliceObject*)item, length, |
|
1917 | 1923 | &start, &stop, &step, &slicelength) < 0) |
|
1918 | 1924 | #endif |
|
1919 | 1925 | return -1; |
|
1920 | 1926 | |
|
1921 | 1927 | if (slicelength <= 0) |
|
1922 | 1928 | return 0; |
|
1923 | 1929 | |
|
1924 | 1930 | if ((step < 0 && start < stop) || (step > 0 && start > stop)) |
|
1925 | 1931 | stop = start; |
|
1926 | 1932 | |
|
1927 | 1933 | if (step < 0) { |
|
1928 | 1934 | stop = start + 1; |
|
1929 | 1935 | start = stop + step*(slicelength - 1) - 1; |
|
1930 | 1936 | step = -step; |
|
1931 | 1937 | } |
|
1932 | 1938 | |
|
1933 | 1939 | if (step != 1) { |
|
1934 | 1940 | PyErr_SetString(PyExc_ValueError, |
|
1935 | 1941 | "revlog index delete requires step size of 1"); |
|
1936 | 1942 | return -1; |
|
1937 | 1943 | } |
|
1938 | 1944 | |
|
1939 | 1945 | if (stop != length - 1) { |
|
1940 | 1946 | PyErr_SetString(PyExc_IndexError, |
|
1941 | 1947 | "revlog index deletion indices are invalid"); |
|
1942 | 1948 | return -1; |
|
1943 | 1949 | } |
|
1944 | 1950 | |
|
1945 | 1951 | if (start < self->length) { |
|
1946 | 1952 | if (self->ntinitialized) { |
|
1947 | 1953 | Py_ssize_t i; |
|
1948 | 1954 | |
|
1949 | 1955 | for (i = start + 1; i < self->length; i++) { |
|
1950 | 1956 | const char *node = index_node_existing(self, i); |
|
1951 | 1957 | if (node == NULL) |
|
1952 | 1958 | return -1; |
|
1953 | 1959 | |
|
1954 | 1960 | nt_delete_node(&self->nt, node); |
|
1955 | 1961 | } |
|
1956 | 1962 | if (self->added) |
|
1957 | 1963 | index_invalidate_added(self, 0); |
|
1958 | 1964 | if (self->ntrev > start) |
|
1959 | 1965 | self->ntrev = (int)start; |
|
1960 | 1966 | } |
|
1961 | 1967 | self->length = start; |
|
1962 | 1968 | if (start < self->raw_length) { |
|
1963 | 1969 | if (self->cache) { |
|
1964 | 1970 | Py_ssize_t i; |
|
1965 | 1971 | for (i = start; i < self->raw_length; i++) |
|
1966 | 1972 | Py_CLEAR(self->cache[i]); |
|
1967 | 1973 | } |
|
1968 | 1974 | self->raw_length = start; |
|
1969 | 1975 | } |
|
1970 | 1976 | goto done; |
|
1971 | 1977 | } |
|
1972 | 1978 | |
|
1973 | 1979 | if (self->ntinitialized) { |
|
1974 | 1980 | index_invalidate_added(self, start - self->length); |
|
1975 | 1981 | if (self->ntrev > start) |
|
1976 | 1982 | self->ntrev = (int)start; |
|
1977 | 1983 | } |
|
1978 | 1984 | if (self->added) |
|
1979 | 1985 | ret = PyList_SetSlice(self->added, start - self->length, |
|
1980 | 1986 | PyList_GET_SIZE(self->added), NULL); |
|
1981 | 1987 | done: |
|
1982 | 1988 | Py_CLEAR(self->headrevs); |
|
1983 | 1989 | return ret; |
|
1984 | 1990 | } |
|
1985 | 1991 | |
|
1986 | 1992 | /* |
|
1987 | 1993 | * Supported ops: |
|
1988 | 1994 | * |
|
1989 | 1995 | * slice deletion |
|
1990 | 1996 | * string assignment (extend node->rev mapping) |
|
1991 | 1997 | * string deletion (shrink node->rev mapping) |
|
1992 | 1998 | */ |
|
1993 | 1999 | static int index_assign_subscript(indexObject *self, PyObject *item, |
|
1994 | 2000 | PyObject *value) |
|
1995 | 2001 | { |
|
1996 | 2002 | char *node; |
|
1997 | 2003 | long rev; |
|
1998 | 2004 | |
|
1999 | 2005 | if (PySlice_Check(item) && value == NULL) |
|
2000 | 2006 | return index_slice_del(self, item); |
|
2001 | 2007 | |
|
2002 | 2008 | if (node_check(item, &node) == -1) |
|
2003 | 2009 | return -1; |
|
2004 | 2010 | |
|
2005 | 2011 | if (value == NULL) |
|
2006 | 2012 | return self->ntinitialized ? nt_delete_node(&self->nt, node) : 0; |
|
2007 | 2013 | rev = PyInt_AsLong(value); |
|
2008 | 2014 | if (rev > INT_MAX || rev < 0) { |
|
2009 | 2015 | if (!PyErr_Occurred()) |
|
2010 | 2016 | PyErr_SetString(PyExc_ValueError, "rev out of range"); |
|
2011 | 2017 | return -1; |
|
2012 | 2018 | } |
|
2013 | 2019 | |
|
2014 | 2020 | if (index_init_nt(self) == -1) |
|
2015 | 2021 | return -1; |
|
2016 | 2022 | return nt_insert(&self->nt, node, (int)rev); |
|
2017 | 2023 | } |
|
2018 | 2024 | |
|
2019 | 2025 | /* |
|
2020 | 2026 | * Find all RevlogNG entries in an index that has inline data. Update |
|
2021 | 2027 | * the optional "offsets" table with those entries. |
|
2022 | 2028 | */ |
|
2023 | 2029 | static Py_ssize_t inline_scan(indexObject *self, const char **offsets) |
|
2024 | 2030 | { |
|
2025 | 2031 | const char *data = (const char *)self->buf.buf; |
|
2026 | 2032 | Py_ssize_t pos = 0; |
|
2027 | 2033 | Py_ssize_t end = self->buf.len; |
|
2028 | 2034 | long incr = v1_hdrsize; |
|
2029 | 2035 | Py_ssize_t len = 0; |
|
2030 | 2036 | |
|
2031 | 2037 | while (pos + v1_hdrsize <= end && pos >= 0) { |
|
2032 | 2038 | uint32_t comp_len; |
|
2033 | 2039 | /* 3rd element of header is length of compressed inline data */ |
|
2034 | 2040 | comp_len = getbe32(data + pos + 8); |
|
2035 | 2041 | incr = v1_hdrsize + comp_len; |
|
2036 | 2042 | if (offsets) |
|
2037 | 2043 | offsets[len] = data + pos; |
|
2038 | 2044 | len++; |
|
2039 | 2045 | pos += incr; |
|
2040 | 2046 | } |
|
2041 | 2047 | |
|
2042 | 2048 | if (pos != end) { |
|
2043 | 2049 | if (!PyErr_Occurred()) |
|
2044 | 2050 | PyErr_SetString(PyExc_ValueError, "corrupt index file"); |
|
2045 | 2051 | return -1; |
|
2046 | 2052 | } |
|
2047 | 2053 | |
|
2048 | 2054 | return len; |
|
2049 | 2055 | } |
|
2050 | 2056 | |
|
2051 | 2057 | static int index_init(indexObject *self, PyObject *args) |
|
2052 | 2058 | { |
|
2053 | 2059 | PyObject *data_obj, *inlined_obj; |
|
2054 | 2060 | Py_ssize_t size; |
|
2055 | 2061 | |
|
2056 | 2062 | /* Initialize before argument-checking to avoid index_dealloc() crash. */ |
|
2057 | 2063 | self->raw_length = 0; |
|
2058 | 2064 | self->added = NULL; |
|
2059 | 2065 | self->cache = NULL; |
|
2060 | 2066 | self->data = NULL; |
|
2061 | 2067 | memset(&self->buf, 0, sizeof(self->buf)); |
|
2062 | 2068 | self->headrevs = NULL; |
|
2063 | 2069 | self->filteredrevs = Py_None; |
|
2064 | 2070 | Py_INCREF(Py_None); |
|
2065 | 2071 | self->ntinitialized = 0; |
|
2066 | 2072 | self->offsets = NULL; |
|
2067 | 2073 | |
|
2068 | 2074 | if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj)) |
|
2069 | 2075 | return -1; |
|
2070 | 2076 | if (!PyObject_CheckBuffer(data_obj)) { |
|
2071 | 2077 | PyErr_SetString(PyExc_TypeError, |
|
2072 | 2078 | "data does not support buffer interface"); |
|
2073 | 2079 | return -1; |
|
2074 | 2080 | } |
|
2075 | 2081 | |
|
2076 | 2082 | if (PyObject_GetBuffer(data_obj, &self->buf, PyBUF_SIMPLE) == -1) |
|
2077 | 2083 | return -1; |
|
2078 | 2084 | size = self->buf.len; |
|
2079 | 2085 | |
|
2080 | 2086 | self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj); |
|
2081 | 2087 | self->data = data_obj; |
|
2082 | 2088 | |
|
2083 | 2089 | self->ntlookups = self->ntmisses = 0; |
|
2084 | 2090 | self->ntrev = -1; |
|
2085 | 2091 | Py_INCREF(self->data); |
|
2086 | 2092 | |
|
2087 | 2093 | if (self->inlined) { |
|
2088 | 2094 | Py_ssize_t len = inline_scan(self, NULL); |
|
2089 | 2095 | if (len == -1) |
|
2090 | 2096 | goto bail; |
|
2091 | 2097 | self->raw_length = len; |
|
2092 | 2098 | self->length = len; |
|
2093 | 2099 | } else { |
|
2094 | 2100 | if (size % v1_hdrsize) { |
|
2095 | 2101 | PyErr_SetString(PyExc_ValueError, "corrupt index file"); |
|
2096 | 2102 | goto bail; |
|
2097 | 2103 | } |
|
2098 | 2104 | self->raw_length = size / v1_hdrsize; |
|
2099 | 2105 | self->length = self->raw_length; |
|
2100 | 2106 | } |
|
2101 | 2107 | |
|
2102 | 2108 | return 0; |
|
2103 | 2109 | bail: |
|
2104 | 2110 | return -1; |
|
2105 | 2111 | } |
|
2106 | 2112 | |
|
2107 | 2113 | static PyObject *index_nodemap(indexObject *self) |
|
2108 | 2114 | { |
|
2109 | 2115 | Py_INCREF(self); |
|
2110 | 2116 | return (PyObject *)self; |
|
2111 | 2117 | } |
|
2112 | 2118 | |
|
2113 | 2119 | static void _index_clearcaches(indexObject *self) |
|
2114 | 2120 | { |
|
2115 | 2121 | if (self->cache) { |
|
2116 | 2122 | Py_ssize_t i; |
|
2117 | 2123 | |
|
2118 | 2124 | for (i = 0; i < self->raw_length; i++) |
|
2119 | 2125 | Py_CLEAR(self->cache[i]); |
|
2120 | 2126 | free(self->cache); |
|
2121 | 2127 | self->cache = NULL; |
|
2122 | 2128 | } |
|
2123 | 2129 | if (self->offsets) { |
|
2124 | 2130 | PyMem_Free((void *)self->offsets); |
|
2125 | 2131 | self->offsets = NULL; |
|
2126 | 2132 | } |
|
2127 | 2133 | if (self->ntinitialized) { |
|
2128 | 2134 | nt_dealloc(&self->nt); |
|
2129 | 2135 | } |
|
2130 | 2136 | self->ntinitialized = 0; |
|
2131 | 2137 | Py_CLEAR(self->headrevs); |
|
2132 | 2138 | } |
|
2133 | 2139 | |
|
2134 | 2140 | static PyObject *index_clearcaches(indexObject *self) |
|
2135 | 2141 | { |
|
2136 | 2142 | _index_clearcaches(self); |
|
2137 | 2143 | self->ntrev = -1; |
|
2138 | 2144 | self->ntlookups = self->ntmisses = 0; |
|
2139 | 2145 | Py_RETURN_NONE; |
|
2140 | 2146 | } |
|
2141 | 2147 | |
|
2142 | 2148 | static void index_dealloc(indexObject *self) |
|
2143 | 2149 | { |
|
2144 | 2150 | _index_clearcaches(self); |
|
2145 | 2151 | Py_XDECREF(self->filteredrevs); |
|
2146 | 2152 | if (self->buf.buf) { |
|
2147 | 2153 | PyBuffer_Release(&self->buf); |
|
2148 | 2154 | memset(&self->buf, 0, sizeof(self->buf)); |
|
2149 | 2155 | } |
|
2150 | 2156 | Py_XDECREF(self->data); |
|
2151 | 2157 | Py_XDECREF(self->added); |
|
2152 | 2158 | PyObject_Del(self); |
|
2153 | 2159 | } |
|
2154 | 2160 | |
|
2155 | 2161 | static PySequenceMethods index_sequence_methods = { |
|
2156 | 2162 | (lenfunc)index_length, /* sq_length */ |
|
2157 | 2163 | 0, /* sq_concat */ |
|
2158 | 2164 | 0, /* sq_repeat */ |
|
2159 | 2165 | (ssizeargfunc)index_get, /* sq_item */ |
|
2160 | 2166 | 0, /* sq_slice */ |
|
2161 | 2167 | 0, /* sq_ass_item */ |
|
2162 | 2168 | 0, /* sq_ass_slice */ |
|
2163 | 2169 | (objobjproc)index_contains, /* sq_contains */ |
|
2164 | 2170 | }; |
|
2165 | 2171 | |
|
2166 | 2172 | static PyMappingMethods index_mapping_methods = { |
|
2167 | 2173 | (lenfunc)index_length, /* mp_length */ |
|
2168 | 2174 | (binaryfunc)index_getitem, /* mp_subscript */ |
|
2169 | 2175 | (objobjargproc)index_assign_subscript, /* mp_ass_subscript */ |
|
2170 | 2176 | }; |
|
2171 | 2177 | |
|
2172 | 2178 | static PyMethodDef index_methods[] = { |
|
2173 | 2179 | {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS, |
|
2174 | 2180 | "return the gca set of the given revs"}, |
|
2175 | 2181 | {"commonancestorsheads", (PyCFunction)index_commonancestorsheads, |
|
2176 | 2182 | METH_VARARGS, |
|
2177 | 2183 | "return the heads of the common ancestors of the given revs"}, |
|
2178 | 2184 | {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS, |
|
2179 | 2185 | "clear the index caches"}, |
|
2180 | 2186 | {"get", (PyCFunction)index_m_get, METH_VARARGS, |
|
2181 | 2187 | "get an index entry"}, |
|
2182 | 2188 | {"computephasesmapsets", (PyCFunction)compute_phases_map_sets, |
|
2183 | 2189 | METH_VARARGS, "compute phases"}, |
|
2184 | 2190 | {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS, |
|
2185 | 2191 | "reachableroots"}, |
|
2186 | 2192 | {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS, |
|
2187 | 2193 | "get head revisions"}, /* Can do filtering since 3.2 */ |
|
2188 | 2194 | {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS, |
|
2189 | 2195 | "get filtered head revisions"}, /* Can always do filtering */ |
|
2190 | 2196 | {"deltachain", (PyCFunction)index_deltachain, METH_VARARGS, |
|
2191 | 2197 | "determine revisions with deltas to reconstruct fulltext"}, |
|
2192 | 2198 | {"append", (PyCFunction)index_append, METH_O, |
|
2193 | 2199 | "append an index entry"}, |
|
2194 | 2200 | {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS, |
|
2195 | 2201 | "match a potentially ambiguous node ID"}, |
|
2196 | 2202 | {"shortest", (PyCFunction)index_shortest, METH_VARARGS, |
|
2197 | 2203 | "find length of shortest hex nodeid of a binary ID"}, |
|
2198 | 2204 | {"stats", (PyCFunction)index_stats, METH_NOARGS, |
|
2199 | 2205 | "stats for the index"}, |
|
2200 | 2206 | {NULL} /* Sentinel */ |
|
2201 | 2207 | }; |
|
2202 | 2208 | |
|
2203 | 2209 | static PyGetSetDef index_getset[] = { |
|
2204 | 2210 | {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL}, |
|
2205 | 2211 | {NULL} /* Sentinel */ |
|
2206 | 2212 | }; |
|
2207 | 2213 | |
|
2208 | 2214 | static PyTypeObject indexType = { |
|
2209 | 2215 | PyVarObject_HEAD_INIT(NULL, 0) /* header */ |
|
2210 | 2216 | "parsers.index", /* tp_name */ |
|
2211 | 2217 | sizeof(indexObject), /* tp_basicsize */ |
|
2212 | 2218 | 0, /* tp_itemsize */ |
|
2213 | 2219 | (destructor)index_dealloc, /* tp_dealloc */ |
|
2214 | 2220 | 0, /* tp_print */ |
|
2215 | 2221 | 0, /* tp_getattr */ |
|
2216 | 2222 | 0, /* tp_setattr */ |
|
2217 | 2223 | 0, /* tp_compare */ |
|
2218 | 2224 | 0, /* tp_repr */ |
|
2219 | 2225 | 0, /* tp_as_number */ |
|
2220 | 2226 | &index_sequence_methods, /* tp_as_sequence */ |
|
2221 | 2227 | &index_mapping_methods, /* tp_as_mapping */ |
|
2222 | 2228 | 0, /* tp_hash */ |
|
2223 | 2229 | 0, /* tp_call */ |
|
2224 | 2230 | 0, /* tp_str */ |
|
2225 | 2231 | 0, /* tp_getattro */ |
|
2226 | 2232 | 0, /* tp_setattro */ |
|
2227 | 2233 | 0, /* tp_as_buffer */ |
|
2228 | 2234 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
2229 | 2235 | "revlog index", /* tp_doc */ |
|
2230 | 2236 | 0, /* tp_traverse */ |
|
2231 | 2237 | 0, /* tp_clear */ |
|
2232 | 2238 | 0, /* tp_richcompare */ |
|
2233 | 2239 | 0, /* tp_weaklistoffset */ |
|
2234 | 2240 | 0, /* tp_iter */ |
|
2235 | 2241 | 0, /* tp_iternext */ |
|
2236 | 2242 | index_methods, /* tp_methods */ |
|
2237 | 2243 | 0, /* tp_members */ |
|
2238 | 2244 | index_getset, /* tp_getset */ |
|
2239 | 2245 | 0, /* tp_base */ |
|
2240 | 2246 | 0, /* tp_dict */ |
|
2241 | 2247 | 0, /* tp_descr_get */ |
|
2242 | 2248 | 0, /* tp_descr_set */ |
|
2243 | 2249 | 0, /* tp_dictoffset */ |
|
2244 | 2250 | (initproc)index_init, /* tp_init */ |
|
2245 | 2251 | 0, /* tp_alloc */ |
|
2246 | 2252 | }; |
|
2247 | 2253 | |
|
2248 | 2254 | /* |
|
2249 | 2255 | * returns a tuple of the form (index, index, cache) with elements as |
|
2250 | 2256 | * follows: |
|
2251 | 2257 | * |
|
2252 | 2258 | * index: an index object that lazily parses RevlogNG records |
|
2253 | 2259 | * cache: if data is inlined, a tuple (0, index_file_content), else None |
|
2254 | 2260 | * index_file_content could be a string, or a buffer |
|
2255 | 2261 | * |
|
2256 | 2262 | * added complications are for backwards compatibility |
|
2257 | 2263 | */ |
|
2258 | 2264 | PyObject *parse_index2(PyObject *self, PyObject *args) |
|
2259 | 2265 | { |
|
2260 | 2266 | PyObject *tuple = NULL, *cache = NULL; |
|
2261 | 2267 | indexObject *idx; |
|
2262 | 2268 | int ret; |
|
2263 | 2269 | |
|
2264 | 2270 | idx = PyObject_New(indexObject, &indexType); |
|
2265 | 2271 | if (idx == NULL) |
|
2266 | 2272 | goto bail; |
|
2267 | 2273 | |
|
2268 | 2274 | ret = index_init(idx, args); |
|
2269 | 2275 | if (ret == -1) |
|
2270 | 2276 | goto bail; |
|
2271 | 2277 | |
|
2272 | 2278 | if (idx->inlined) { |
|
2273 | 2279 | cache = Py_BuildValue("iO", 0, idx->data); |
|
2274 | 2280 | if (cache == NULL) |
|
2275 | 2281 | goto bail; |
|
2276 | 2282 | } else { |
|
2277 | 2283 | cache = Py_None; |
|
2278 | 2284 | Py_INCREF(cache); |
|
2279 | 2285 | } |
|
2280 | 2286 | |
|
2281 | 2287 | tuple = Py_BuildValue("NN", idx, cache); |
|
2282 | 2288 | if (!tuple) |
|
2283 | 2289 | goto bail; |
|
2284 | 2290 | return tuple; |
|
2285 | 2291 | |
|
2286 | 2292 | bail: |
|
2287 | 2293 | Py_XDECREF(idx); |
|
2288 | 2294 | Py_XDECREF(cache); |
|
2289 | 2295 | Py_XDECREF(tuple); |
|
2290 | 2296 | return NULL; |
|
2291 | 2297 | } |
|
2292 | 2298 | |
|
2293 | 2299 | #ifdef WITH_RUST |
|
2294 | 2300 | |
|
2295 | 2301 | /* rustlazyancestors: iteration over ancestors implemented in Rust |
|
2296 | 2302 | * |
|
2297 | 2303 | * This class holds a reference to an index and to the Rust iterator. |
|
2298 | 2304 | */ |
|
2299 | 2305 | typedef struct rustlazyancestorsObjectStruct rustlazyancestorsObject; |
|
2300 | 2306 | |
|
2301 | 2307 | struct rustlazyancestorsObjectStruct { |
|
2302 | 2308 | PyObject_HEAD |
|
2303 | 2309 | /* Type-specific fields go here. */ |
|
2304 | 2310 | indexObject *index; /* Ref kept to avoid GC'ing the index */ |
|
2305 | 2311 | void *iter; /* Rust iterator */ |
|
2306 | 2312 | }; |
|
2307 | 2313 | |
|
2308 | 2314 | /* FFI exposed from Rust code */ |
|
2309 | 2315 | rustlazyancestorsObject *rustlazyancestors_init( |
|
2310 | 2316 | indexObject *index, |
|
2311 | 2317 | /* to pass index_get_parents() */ |
|
2312 | 2318 | int (*)(indexObject *, Py_ssize_t, int*, int), |
|
2313 | 2319 | /* intrevs vector */ |
|
2314 | 2320 | Py_ssize_t initrevslen, long *initrevs, |
|
2315 | 2321 | long stoprev, |
|
2316 | 2322 | int inclusive); |
|
2317 | 2323 | void rustlazyancestors_drop(rustlazyancestorsObject *self); |
|
2318 | 2324 | int rustlazyancestors_next(rustlazyancestorsObject *self); |
|
2319 | 2325 | int rustlazyancestors_contains(rustlazyancestorsObject *self, long rev); |
|
2320 | 2326 | |
|
2321 | 2327 | /* CPython instance methods */ |
|
2322 | 2328 | static int rustla_init(rustlazyancestorsObject *self, |
|
2323 | 2329 | PyObject *args) { |
|
2324 | 2330 | PyObject *initrevsarg = NULL; |
|
2325 | 2331 | PyObject *inclusivearg = NULL; |
|
2326 | 2332 | long stoprev = 0; |
|
2327 | 2333 | long *initrevs = NULL; |
|
2328 | 2334 | int inclusive = 0; |
|
2329 | 2335 | Py_ssize_t i; |
|
2330 | 2336 | |
|
2331 | 2337 | indexObject *index; |
|
2332 | 2338 | if (!PyArg_ParseTuple(args, "O!O!lO!", |
|
2333 | 2339 | &indexType, &index, |
|
2334 | 2340 | &PyList_Type, &initrevsarg, |
|
2335 | 2341 | &stoprev, |
|
2336 | 2342 | &PyBool_Type, &inclusivearg)) |
|
2337 | 2343 | return -1; |
|
2338 | 2344 | |
|
2339 | 2345 | Py_INCREF(index); |
|
2340 | 2346 | self->index = index; |
|
2341 | 2347 | |
|
2342 | 2348 | if (inclusivearg == Py_True) |
|
2343 | 2349 | inclusive = 1; |
|
2344 | 2350 | |
|
2345 | 2351 | Py_ssize_t linit = PyList_GET_SIZE(initrevsarg); |
|
2346 | 2352 | |
|
2347 | 2353 | initrevs = (long*)calloc(linit, sizeof(long)); |
|
2348 | 2354 | |
|
2349 | 2355 | if (initrevs == NULL) { |
|
2350 | 2356 | PyErr_NoMemory(); |
|
2351 | 2357 | goto bail; |
|
2352 | 2358 | } |
|
2353 | 2359 | |
|
2354 | 2360 | for (i=0; i<linit; i++) { |
|
2355 | 2361 | initrevs[i] = PyInt_AsLong(PyList_GET_ITEM(initrevsarg, i)); |
|
2356 | 2362 | } |
|
2357 | 2363 | if (PyErr_Occurred()) |
|
2358 | 2364 | goto bail; |
|
2359 | 2365 | |
|
2360 | 2366 | self->iter = rustlazyancestors_init(index, |
|
2361 | 2367 | index_get_parents, |
|
2362 | 2368 | linit, initrevs, |
|
2363 | 2369 | stoprev, inclusive); |
|
2364 | 2370 | if (self->iter == NULL) { |
|
2365 | 2371 | /* if this is because of GraphError::ParentOutOfRange |
|
2366 | 2372 | * index_get_parents() has already set the proper ValueError */ |
|
2367 | 2373 | goto bail; |
|
2368 | 2374 | } |
|
2369 | 2375 | |
|
2370 | 2376 | free(initrevs); |
|
2371 | 2377 | return 0; |
|
2372 | 2378 | |
|
2373 | 2379 | bail: |
|
2374 | 2380 | free(initrevs); |
|
2375 | 2381 | return -1; |
|
2376 | 2382 | }; |
|
2377 | 2383 | |
|
2378 | 2384 | static void rustla_dealloc(rustlazyancestorsObject *self) |
|
2379 | 2385 | { |
|
2380 | 2386 | Py_XDECREF(self->index); |
|
2381 | 2387 | if (self->iter != NULL) { /* can happen if rustla_init failed */ |
|
2382 | 2388 | rustlazyancestors_drop(self->iter); |
|
2383 | 2389 | } |
|
2384 | 2390 | PyObject_Del(self); |
|
2385 | 2391 | } |
|
2386 | 2392 | |
|
2387 | 2393 | static PyObject *rustla_next(rustlazyancestorsObject *self) { |
|
2388 | 2394 | int res = rustlazyancestors_next(self->iter); |
|
2389 | 2395 | if (res == -1) { |
|
2390 | 2396 | /* Setting an explicit exception seems unnecessary |
|
2391 | 2397 | * as examples from Python source code (Objects/rangeobjets.c and |
|
2392 | 2398 | * Modules/_io/stringio.c) seem to demonstrate. |
|
2393 | 2399 | */ |
|
2394 | 2400 | return NULL; |
|
2395 | 2401 | } |
|
2396 | 2402 | return PyInt_FromLong(res); |
|
2397 | 2403 | } |
|
2398 | 2404 | |
|
2399 | 2405 | static int rustla_contains(rustlazyancestorsObject *self, PyObject *rev) { |
|
2400 | 2406 | if (!(PyInt_Check(rev))) { |
|
2401 | 2407 | return 0; |
|
2402 | 2408 | } |
|
2403 | 2409 | return rustlazyancestors_contains(self->iter, PyInt_AS_LONG(rev)); |
|
2404 | 2410 | } |
|
2405 | 2411 | |
|
2406 | 2412 | static PySequenceMethods rustla_sequence_methods = { |
|
2407 | 2413 | 0, /* sq_length */ |
|
2408 | 2414 | 0, /* sq_concat */ |
|
2409 | 2415 | 0, /* sq_repeat */ |
|
2410 | 2416 | 0, /* sq_item */ |
|
2411 | 2417 | 0, /* sq_slice */ |
|
2412 | 2418 | 0, /* sq_ass_item */ |
|
2413 | 2419 | 0, /* sq_ass_slice */ |
|
2414 | 2420 | (objobjproc)rustla_contains, /* sq_contains */ |
|
2415 | 2421 | }; |
|
2416 | 2422 | |
|
2417 | 2423 | static PyTypeObject rustlazyancestorsType = { |
|
2418 | 2424 | PyVarObject_HEAD_INIT(NULL, 0) /* header */ |
|
2419 | 2425 | "parsers.rustlazyancestors", /* tp_name */ |
|
2420 | 2426 | sizeof(rustlazyancestorsObject), /* tp_basicsize */ |
|
2421 | 2427 | 0, /* tp_itemsize */ |
|
2422 | 2428 | (destructor)rustla_dealloc, /* tp_dealloc */ |
|
2423 | 2429 | 0, /* tp_print */ |
|
2424 | 2430 | 0, /* tp_getattr */ |
|
2425 | 2431 | 0, /* tp_setattr */ |
|
2426 | 2432 | 0, /* tp_compare */ |
|
2427 | 2433 | 0, /* tp_repr */ |
|
2428 | 2434 | 0, /* tp_as_number */ |
|
2429 | 2435 | &rustla_sequence_methods, /* tp_as_sequence */ |
|
2430 | 2436 | 0, /* tp_as_mapping */ |
|
2431 | 2437 | 0, /* tp_hash */ |
|
2432 | 2438 | 0, /* tp_call */ |
|
2433 | 2439 | 0, /* tp_str */ |
|
2434 | 2440 | 0, /* tp_getattro */ |
|
2435 | 2441 | 0, /* tp_setattro */ |
|
2436 | 2442 | 0, /* tp_as_buffer */ |
|
2437 | 2443 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
2438 | 2444 | "Iterator over ancestors, implemented in Rust", /* tp_doc */ |
|
2439 | 2445 | 0, /* tp_traverse */ |
|
2440 | 2446 | 0, /* tp_clear */ |
|
2441 | 2447 | 0, /* tp_richcompare */ |
|
2442 | 2448 | 0, /* tp_weaklistoffset */ |
|
2443 | 2449 | 0, /* tp_iter */ |
|
2444 | 2450 | (iternextfunc)rustla_next, /* tp_iternext */ |
|
2445 | 2451 | 0, /* tp_methods */ |
|
2446 | 2452 | 0, /* tp_members */ |
|
2447 | 2453 | 0, /* tp_getset */ |
|
2448 | 2454 | 0, /* tp_base */ |
|
2449 | 2455 | 0, /* tp_dict */ |
|
2450 | 2456 | 0, /* tp_descr_get */ |
|
2451 | 2457 | 0, /* tp_descr_set */ |
|
2452 | 2458 | 0, /* tp_dictoffset */ |
|
2453 | 2459 | (initproc)rustla_init, /* tp_init */ |
|
2454 | 2460 | 0, /* tp_alloc */ |
|
2455 | 2461 | }; |
|
2456 | 2462 | #endif /* WITH_RUST */ |
|
2457 | 2463 | |
|
2458 | 2464 | void revlog_module_init(PyObject *mod) |
|
2459 | 2465 | { |
|
2460 | 2466 | indexType.tp_new = PyType_GenericNew; |
|
2461 | 2467 | if (PyType_Ready(&indexType) < 0) |
|
2462 | 2468 | return; |
|
2463 | 2469 | Py_INCREF(&indexType); |
|
2464 | 2470 | PyModule_AddObject(mod, "index", (PyObject *)&indexType); |
|
2465 | 2471 | |
|
2466 | 2472 | nodetreeType.tp_new = PyType_GenericNew; |
|
2467 | 2473 | if (PyType_Ready(&nodetreeType) < 0) |
|
2468 | 2474 | return; |
|
2469 | 2475 | Py_INCREF(&nodetreeType); |
|
2470 | 2476 | PyModule_AddObject(mod, "nodetree", (PyObject *)&nodetreeType); |
|
2471 | 2477 | |
|
2472 | 2478 | if (!nullentry) { |
|
2473 | 2479 | nullentry = Py_BuildValue(PY23("iiiiiiis#", "iiiiiiiy#"), 0, 0, 0, |
|
2474 | 2480 | -1, -1, -1, -1, nullid, 20); |
|
2475 | 2481 | } |
|
2476 | 2482 | if (nullentry) |
|
2477 | 2483 | PyObject_GC_UnTrack(nullentry); |
|
2478 | 2484 | |
|
2479 | 2485 | #ifdef WITH_RUST |
|
2480 | 2486 | rustlazyancestorsType.tp_new = PyType_GenericNew; |
|
2481 | 2487 | if (PyType_Ready(&rustlazyancestorsType) < 0) |
|
2482 | 2488 | return; |
|
2483 | 2489 | Py_INCREF(&rustlazyancestorsType); |
|
2484 | 2490 | PyModule_AddObject(mod, "rustlazyancestors", |
|
2485 | 2491 | (PyObject *)&rustlazyancestorsType); |
|
2486 | 2492 | #endif |
|
2487 | 2493 | |
|
2488 | 2494 | } |
@@ -1,205 +1,222 b'' | |||
|
1 | 1 | revlog.parseindex must be able to parse the index file even if |
|
2 | 2 | an index entry is split between two 64k blocks. The ideal test |
|
3 | 3 | would be to create an index file with inline data where |
|
4 | 4 | 64k < size < 64k + 64 (64k is the size of the read buffer, 64 is |
|
5 | 5 | the size of an index entry) and with an index entry starting right |
|
6 | 6 | before the 64k block boundary, and try to read it. |
|
7 | 7 | We approximate that by reducing the read buffer to 1 byte. |
|
8 | 8 | |
|
9 | 9 | $ hg init a |
|
10 | 10 | $ cd a |
|
11 | 11 | $ echo abc > foo |
|
12 | 12 | $ hg add foo |
|
13 | 13 | $ hg commit -m 'add foo' |
|
14 | 14 | $ echo >> foo |
|
15 | 15 | $ hg commit -m 'change foo' |
|
16 | 16 | $ hg log -r 0: |
|
17 | 17 | changeset: 0:7c31755bf9b5 |
|
18 | 18 | user: test |
|
19 | 19 | date: Thu Jan 01 00:00:00 1970 +0000 |
|
20 | 20 | summary: add foo |
|
21 | 21 | |
|
22 | 22 | changeset: 1:26333235a41c |
|
23 | 23 | tag: tip |
|
24 | 24 | user: test |
|
25 | 25 | date: Thu Jan 01 00:00:00 1970 +0000 |
|
26 | 26 | summary: change foo |
|
27 | 27 | |
|
28 | 28 | $ cat >> test.py << EOF |
|
29 | 29 | > from __future__ import print_function |
|
30 | 30 | > from mercurial import changelog, node, vfs |
|
31 | 31 | > |
|
32 | 32 | > class singlebyteread(object): |
|
33 | 33 | > def __init__(self, real): |
|
34 | 34 | > self.real = real |
|
35 | 35 | > |
|
36 | 36 | > def read(self, size=-1): |
|
37 | 37 | > if size == 65536: |
|
38 | 38 | > size = 1 |
|
39 | 39 | > return self.real.read(size) |
|
40 | 40 | > |
|
41 | 41 | > def __getattr__(self, key): |
|
42 | 42 | > return getattr(self.real, key) |
|
43 | 43 | > |
|
44 | 44 | > def __enter__(self): |
|
45 | 45 | > self.real.__enter__() |
|
46 | 46 | > return self |
|
47 | 47 | > |
|
48 | 48 | > def __exit__(self, *args, **kwargs): |
|
49 | 49 | > return self.real.__exit__(*args, **kwargs) |
|
50 | 50 | > |
|
51 | 51 | > def opener(*args): |
|
52 | 52 | > o = vfs.vfs(*args) |
|
53 | 53 | > def wrapper(*a, **kwargs): |
|
54 | 54 | > f = o(*a, **kwargs) |
|
55 | 55 | > return singlebyteread(f) |
|
56 | 56 | > return wrapper |
|
57 | 57 | > |
|
58 | 58 | > cl = changelog.changelog(opener('.hg/store')) |
|
59 | 59 | > print(len(cl), 'revisions:') |
|
60 | 60 | > for r in cl: |
|
61 | 61 | > print(node.short(cl.node(r))) |
|
62 | 62 | > EOF |
|
63 | 63 | $ "$PYTHON" test.py |
|
64 | 64 | 2 revisions: |
|
65 | 65 | 7c31755bf9b5 |
|
66 | 66 | 26333235a41c |
|
67 | 67 | |
|
68 | 68 | $ cd .. |
|
69 | 69 | |
|
70 | 70 | #if no-pure |
|
71 | 71 | |
|
72 | 72 | Test SEGV caused by bad revision passed to reachableroots() (issue4775): |
|
73 | 73 | |
|
74 | 74 | $ cd a |
|
75 | 75 | |
|
76 | 76 | $ "$PYTHON" <<EOF |
|
77 | 77 | > from __future__ import print_function |
|
78 | 78 | > from mercurial import changelog, vfs |
|
79 | 79 | > cl = changelog.changelog(vfs.vfs('.hg/store')) |
|
80 | 80 | > print('good heads:') |
|
81 | 81 | > for head in [0, len(cl) - 1, -1]: |
|
82 | 82 | > print('%s: %r' % (head, cl.reachableroots(0, [head], [0]))) |
|
83 | 83 | > print('bad heads:') |
|
84 | 84 | > for head in [len(cl), 10000, -2, -10000, None]: |
|
85 | 85 | > print('%s:' % head, end=' ') |
|
86 | 86 | > try: |
|
87 | 87 | > cl.reachableroots(0, [head], [0]) |
|
88 | 88 | > print('uncaught buffer overflow?') |
|
89 | 89 | > except (IndexError, TypeError) as inst: |
|
90 | 90 | > print(inst) |
|
91 | 91 | > print('good roots:') |
|
92 | 92 | > for root in [0, len(cl) - 1, -1]: |
|
93 | 93 | > print('%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root]))) |
|
94 | 94 | > print('out-of-range roots are ignored:') |
|
95 | 95 | > for root in [len(cl), 10000, -2, -10000]: |
|
96 | 96 | > print('%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root]))) |
|
97 | 97 | > print('bad roots:') |
|
98 | 98 | > for root in [None]: |
|
99 | 99 | > print('%s:' % root, end=' ') |
|
100 | 100 | > try: |
|
101 | 101 | > cl.reachableroots(root, [len(cl) - 1], [root]) |
|
102 | 102 | > print('uncaught error?') |
|
103 | 103 | > except TypeError as inst: |
|
104 | 104 | > print(inst) |
|
105 | 105 | > EOF |
|
106 | 106 | good heads: |
|
107 | 107 | 0: [0] |
|
108 | 108 | 1: [0] |
|
109 | 109 | -1: [] |
|
110 | 110 | bad heads: |
|
111 | 111 | 2: head out of range |
|
112 | 112 | 10000: head out of range |
|
113 | 113 | -2: head out of range |
|
114 | 114 | -10000: head out of range |
|
115 | 115 | None: an integer is required |
|
116 | 116 | good roots: |
|
117 | 117 | 0: [0] |
|
118 | 118 | 1: [1] |
|
119 | 119 | -1: [-1] |
|
120 | 120 | out-of-range roots are ignored: |
|
121 | 121 | 2: [] |
|
122 | 122 | 10000: [] |
|
123 | 123 | -2: [] |
|
124 | 124 | -10000: [] |
|
125 | 125 | bad roots: |
|
126 | 126 | None: an integer is required |
|
127 | 127 | |
|
128 | 128 | $ cd .. |
|
129 | 129 | |
|
130 | 130 | Test corrupted p1/p2 fields that could cause SEGV at parsers.c: |
|
131 | 131 | |
|
132 | 132 | $ mkdir invalidparent |
|
133 | 133 | $ cd invalidparent |
|
134 | 134 | |
|
135 | 135 | $ hg clone --pull -q --config phases.publish=False ../a limit |
|
136 | $ hg clone --pull -q --config phases.publish=False ../a neglimit | |
|
136 | 137 | $ hg clone --pull -q --config phases.publish=False ../a segv |
|
137 | $ rm -R limit/.hg/cache segv/.hg/cache | |
|
138 | $ rm -R limit/.hg/cache neglimit/.hg/cache segv/.hg/cache | |
|
138 | 139 | |
|
139 | 140 | $ "$PYTHON" <<EOF |
|
140 | 141 | > data = open("limit/.hg/store/00changelog.i", "rb").read() |
|
141 | > for n, p in [(b'limit', b'\0\0\0\x02'), (b'segv', b'\0\x01\0\0')]: | |
|
142 | > poisons = [ | |
|
143 | > (b'limit', b'\0\0\0\x02'), | |
|
144 | > (b'neglimit', b'\xff\xff\xff\xfe'), | |
|
145 | > (b'segv', b'\0\x01\0\0'), | |
|
146 | > ] | |
|
147 | > for n, p in poisons: | |
|
142 | 148 | > # corrupt p1 at rev0 and p2 at rev1 |
|
143 | 149 | > d = data[:24] + p + data[28:127 + 28] + p + data[127 + 32:] |
|
144 | 150 | > open(n + b"/.hg/store/00changelog.i", "wb").write(d) |
|
145 | 151 | > EOF |
|
146 | 152 | |
|
147 | 153 | $ hg -R limit debugrevlogindex -f1 -c |
|
148 | 154 | rev flag size link p1 p2 nodeid |
|
149 | 155 | 0 0000 62 0 2 -1 7c31755bf9b5 |
|
150 | 156 | 1 0000 65 1 0 2 26333235a41c |
|
151 | 157 | |
|
152 | 158 | $ hg -R limit debugdeltachain -c |
|
153 | 159 | rev chain# chainlen prev delta size rawsize chainsize ratio lindist extradist extraratio |
|
154 | 160 | 0 1 1 -1 base 63 62 63 1.01613 63 0 0.00000 |
|
155 | 161 | 1 2 1 -1 base 66 65 66 1.01538 66 0 0.00000 |
|
156 | 162 | |
|
163 | $ hg -R neglimit debugrevlogindex -f1 -c | |
|
164 | rev flag size link p1 p2 nodeid | |
|
165 | 0 0000 62 0 -2 -1 7c31755bf9b5 | |
|
166 | 1 0000 65 1 0 -2 26333235a41c | |
|
167 | ||
|
157 | 168 |
$ |
|
158 | 169 | rev flag size link p1 p2 nodeid |
|
159 | 170 | 0 0000 62 0 65536 -1 7c31755bf9b5 |
|
160 | 171 | 1 0000 65 1 0 65536 26333235a41c |
|
161 | 172 | |
|
162 | 173 | $ hg -R segv debugdeltachain -c |
|
163 | 174 | rev chain# chainlen prev delta size rawsize chainsize ratio lindist extradist extraratio |
|
164 | 175 | 0 1 1 -1 base 63 62 63 1.01613 63 0 0.00000 |
|
165 | 176 | 1 2 1 -1 base 66 65 66 1.01538 66 0 0.00000 |
|
166 | 177 | |
|
167 | 178 | $ cat <<EOF > test.py |
|
168 | 179 | > from __future__ import print_function |
|
169 | 180 | > import sys |
|
170 | 181 | > from mercurial import changelog, vfs |
|
171 | 182 | > cl = changelog.changelog(vfs.vfs(sys.argv[1])) |
|
172 | 183 | > n0, n1 = cl.node(0), cl.node(1) |
|
173 | 184 | > ops = [ |
|
174 | 185 | > ('reachableroots', |
|
175 | 186 | > lambda: cl.index.reachableroots2(0, [1], [0], False)), |
|
176 | 187 | > ('compute_phases_map_sets', lambda: cl.computephases([[0], []])), |
|
177 | 188 | > ('index_headrevs', lambda: cl.headrevs()), |
|
178 | 189 | > ('find_gca_candidates', lambda: cl.commonancestorsheads(n0, n1)), |
|
179 | 190 | > ('find_deepest', lambda: cl.ancestor(n0, n1)), |
|
180 | 191 | > ] |
|
181 | 192 | > for l, f in ops: |
|
182 | 193 | > print(l + ':', end=' ') |
|
183 | 194 | > try: |
|
184 | 195 | > f() |
|
185 | 196 | > print('uncaught buffer overflow?') |
|
186 | 197 | > except ValueError as inst: |
|
187 | 198 | > print(inst) |
|
188 | 199 | > EOF |
|
189 | 200 | |
|
190 | 201 | $ "$PYTHON" test.py limit/.hg/store |
|
191 | 202 | reachableroots: parent out of range |
|
192 | 203 | compute_phases_map_sets: parent out of range |
|
193 | 204 | index_headrevs: parent out of range |
|
194 | 205 | find_gca_candidates: parent out of range |
|
195 | 206 | find_deepest: parent out of range |
|
207 | $ "$PYTHON" test.py neglimit/.hg/store | |
|
208 | reachableroots: parent out of range | |
|
209 | compute_phases_map_sets: parent out of range | |
|
210 | index_headrevs: parent out of range | |
|
211 | find_gca_candidates: parent out of range | |
|
212 | find_deepest: parent out of range | |
|
196 | 213 | $ "$PYTHON" test.py segv/.hg/store |
|
197 | 214 | reachableroots: parent out of range |
|
198 | 215 | compute_phases_map_sets: parent out of range |
|
199 | 216 | index_headrevs: parent out of range |
|
200 | 217 | find_gca_candidates: parent out of range |
|
201 | 218 | find_deepest: parent out of range |
|
202 | 219 | |
|
203 | 220 | $ cd .. |
|
204 | 221 | |
|
205 | 222 | #endif |
General Comments 0
You need to be logged in to leave comments.
Login now