##// END OF EJS Templates
reachableroots: fix memleak of integer objects at includepath loop...
Yuya Nishihara -
r26033:9e7d8059 default
parent child Browse files
Show More
@@ -1,2799 +1,2800 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 <ctype.h>
12 12 #include <stddef.h>
13 13 #include <string.h>
14 14
15 15 #include "util.h"
16 16
17 17 static char *versionerrortext = "Python minor version mismatch";
18 18
19 19 static int8_t hextable[256] = {
20 20 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
21 21 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
22 22 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
23 23 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, -1, -1, -1, -1, /* 0-9 */
24 24 -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* A-F */
25 25 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
26 26 -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* a-f */
27 27 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
28 28 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
29 29 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
30 30 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
31 31 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
32 32 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
33 33 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
34 34 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
35 35 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
36 36 };
37 37
38 38 static char lowertable[128] = {
39 39 '\x00', '\x01', '\x02', '\x03', '\x04', '\x05', '\x06', '\x07',
40 40 '\x08', '\x09', '\x0a', '\x0b', '\x0c', '\x0d', '\x0e', '\x0f',
41 41 '\x10', '\x11', '\x12', '\x13', '\x14', '\x15', '\x16', '\x17',
42 42 '\x18', '\x19', '\x1a', '\x1b', '\x1c', '\x1d', '\x1e', '\x1f',
43 43 '\x20', '\x21', '\x22', '\x23', '\x24', '\x25', '\x26', '\x27',
44 44 '\x28', '\x29', '\x2a', '\x2b', '\x2c', '\x2d', '\x2e', '\x2f',
45 45 '\x30', '\x31', '\x32', '\x33', '\x34', '\x35', '\x36', '\x37',
46 46 '\x38', '\x39', '\x3a', '\x3b', '\x3c', '\x3d', '\x3e', '\x3f',
47 47 '\x40',
48 48 '\x61', '\x62', '\x63', '\x64', '\x65', '\x66', '\x67', /* A-G */
49 49 '\x68', '\x69', '\x6a', '\x6b', '\x6c', '\x6d', '\x6e', '\x6f', /* H-O */
50 50 '\x70', '\x71', '\x72', '\x73', '\x74', '\x75', '\x76', '\x77', /* P-W */
51 51 '\x78', '\x79', '\x7a', /* X-Z */
52 52 '\x5b', '\x5c', '\x5d', '\x5e', '\x5f',
53 53 '\x60', '\x61', '\x62', '\x63', '\x64', '\x65', '\x66', '\x67',
54 54 '\x68', '\x69', '\x6a', '\x6b', '\x6c', '\x6d', '\x6e', '\x6f',
55 55 '\x70', '\x71', '\x72', '\x73', '\x74', '\x75', '\x76', '\x77',
56 56 '\x78', '\x79', '\x7a', '\x7b', '\x7c', '\x7d', '\x7e', '\x7f'
57 57 };
58 58
59 59 static char uppertable[128] = {
60 60 '\x00', '\x01', '\x02', '\x03', '\x04', '\x05', '\x06', '\x07',
61 61 '\x08', '\x09', '\x0a', '\x0b', '\x0c', '\x0d', '\x0e', '\x0f',
62 62 '\x10', '\x11', '\x12', '\x13', '\x14', '\x15', '\x16', '\x17',
63 63 '\x18', '\x19', '\x1a', '\x1b', '\x1c', '\x1d', '\x1e', '\x1f',
64 64 '\x20', '\x21', '\x22', '\x23', '\x24', '\x25', '\x26', '\x27',
65 65 '\x28', '\x29', '\x2a', '\x2b', '\x2c', '\x2d', '\x2e', '\x2f',
66 66 '\x30', '\x31', '\x32', '\x33', '\x34', '\x35', '\x36', '\x37',
67 67 '\x38', '\x39', '\x3a', '\x3b', '\x3c', '\x3d', '\x3e', '\x3f',
68 68 '\x40', '\x41', '\x42', '\x43', '\x44', '\x45', '\x46', '\x47',
69 69 '\x48', '\x49', '\x4a', '\x4b', '\x4c', '\x4d', '\x4e', '\x4f',
70 70 '\x50', '\x51', '\x52', '\x53', '\x54', '\x55', '\x56', '\x57',
71 71 '\x58', '\x59', '\x5a', '\x5b', '\x5c', '\x5d', '\x5e', '\x5f',
72 72 '\x60',
73 73 '\x41', '\x42', '\x43', '\x44', '\x45', '\x46', '\x47', /* a-g */
74 74 '\x48', '\x49', '\x4a', '\x4b', '\x4c', '\x4d', '\x4e', '\x4f', /* h-o */
75 75 '\x50', '\x51', '\x52', '\x53', '\x54', '\x55', '\x56', '\x57', /* p-w */
76 76 '\x58', '\x59', '\x5a', /* x-z */
77 77 '\x7b', '\x7c', '\x7d', '\x7e', '\x7f'
78 78 };
79 79
80 80 static inline int hexdigit(const char *p, Py_ssize_t off)
81 81 {
82 82 int8_t val = hextable[(unsigned char)p[off]];
83 83
84 84 if (val >= 0) {
85 85 return val;
86 86 }
87 87
88 88 PyErr_SetString(PyExc_ValueError, "input contains non-hex character");
89 89 return 0;
90 90 }
91 91
92 92 /*
93 93 * Turn a hex-encoded string into binary.
94 94 */
95 95 PyObject *unhexlify(const char *str, int len)
96 96 {
97 97 PyObject *ret;
98 98 char *d;
99 99 int i;
100 100
101 101 ret = PyBytes_FromStringAndSize(NULL, len / 2);
102 102
103 103 if (!ret)
104 104 return NULL;
105 105
106 106 d = PyBytes_AsString(ret);
107 107
108 108 for (i = 0; i < len;) {
109 109 int hi = hexdigit(str, i++);
110 110 int lo = hexdigit(str, i++);
111 111 *d++ = (hi << 4) | lo;
112 112 }
113 113
114 114 return ret;
115 115 }
116 116
117 117 static inline PyObject *_asciitransform(PyObject *str_obj,
118 118 const char table[128],
119 119 PyObject *fallback_fn)
120 120 {
121 121 char *str, *newstr;
122 122 Py_ssize_t i, len;
123 123 PyObject *newobj = NULL;
124 124 PyObject *ret = NULL;
125 125
126 126 str = PyBytes_AS_STRING(str_obj);
127 127 len = PyBytes_GET_SIZE(str_obj);
128 128
129 129 newobj = PyBytes_FromStringAndSize(NULL, len);
130 130 if (!newobj)
131 131 goto quit;
132 132
133 133 newstr = PyBytes_AS_STRING(newobj);
134 134
135 135 for (i = 0; i < len; i++) {
136 136 char c = str[i];
137 137 if (c & 0x80) {
138 138 if (fallback_fn != NULL) {
139 139 ret = PyObject_CallFunctionObjArgs(fallback_fn,
140 140 str_obj, NULL);
141 141 } else {
142 142 PyObject *err = PyUnicodeDecodeError_Create(
143 143 "ascii", str, len, i, (i + 1),
144 144 "unexpected code byte");
145 145 PyErr_SetObject(PyExc_UnicodeDecodeError, err);
146 146 Py_XDECREF(err);
147 147 }
148 148 goto quit;
149 149 }
150 150 newstr[i] = table[(unsigned char)c];
151 151 }
152 152
153 153 ret = newobj;
154 154 Py_INCREF(ret);
155 155 quit:
156 156 Py_XDECREF(newobj);
157 157 return ret;
158 158 }
159 159
160 160 static PyObject *asciilower(PyObject *self, PyObject *args)
161 161 {
162 162 PyObject *str_obj;
163 163 if (!PyArg_ParseTuple(args, "O!:asciilower", &PyBytes_Type, &str_obj))
164 164 return NULL;
165 165 return _asciitransform(str_obj, lowertable, NULL);
166 166 }
167 167
168 168 static PyObject *asciiupper(PyObject *self, PyObject *args)
169 169 {
170 170 PyObject *str_obj;
171 171 if (!PyArg_ParseTuple(args, "O!:asciiupper", &PyBytes_Type, &str_obj))
172 172 return NULL;
173 173 return _asciitransform(str_obj, uppertable, NULL);
174 174 }
175 175
176 176 static inline PyObject *_dict_new_presized(Py_ssize_t expected_size)
177 177 {
178 178 /* _PyDict_NewPresized expects a minused parameter, but it actually
179 179 creates a dictionary that's the nearest power of two bigger than the
180 180 parameter. For example, with the initial minused = 1000, the
181 181 dictionary created has size 1024. Of course in a lot of cases that
182 182 can be greater than the maximum load factor Python's dict object
183 183 expects (= 2/3), so as soon as we cross the threshold we'll resize
184 184 anyway. So create a dictionary that's at least 3/2 the size. */
185 185 return _PyDict_NewPresized(((1 + expected_size) / 2) * 3);
186 186 }
187 187
188 188 static PyObject *dict_new_presized(PyObject *self, PyObject *args)
189 189 {
190 190 Py_ssize_t expected_size;
191 191
192 192 if (!PyArg_ParseTuple(args, "n:make_presized_dict", &expected_size))
193 193 return NULL;
194 194
195 195 return _dict_new_presized(expected_size);
196 196 }
197 197
198 198 static PyObject *make_file_foldmap(PyObject *self, PyObject *args)
199 199 {
200 200 PyObject *dmap, *spec_obj, *normcase_fallback;
201 201 PyObject *file_foldmap = NULL;
202 202 enum normcase_spec spec;
203 203 PyObject *k, *v;
204 204 dirstateTupleObject *tuple;
205 205 Py_ssize_t pos = 0;
206 206 const char *table;
207 207
208 208 if (!PyArg_ParseTuple(args, "O!O!O!:make_file_foldmap",
209 209 &PyDict_Type, &dmap,
210 210 &PyInt_Type, &spec_obj,
211 211 &PyFunction_Type, &normcase_fallback))
212 212 goto quit;
213 213
214 214 spec = (int)PyInt_AS_LONG(spec_obj);
215 215 switch (spec) {
216 216 case NORMCASE_LOWER:
217 217 table = lowertable;
218 218 break;
219 219 case NORMCASE_UPPER:
220 220 table = uppertable;
221 221 break;
222 222 case NORMCASE_OTHER:
223 223 table = NULL;
224 224 break;
225 225 default:
226 226 PyErr_SetString(PyExc_TypeError, "invalid normcasespec");
227 227 goto quit;
228 228 }
229 229
230 230 /* Add some more entries to deal with additions outside this
231 231 function. */
232 232 file_foldmap = _dict_new_presized((PyDict_Size(dmap) / 10) * 11);
233 233 if (file_foldmap == NULL)
234 234 goto quit;
235 235
236 236 while (PyDict_Next(dmap, &pos, &k, &v)) {
237 237 if (!dirstate_tuple_check(v)) {
238 238 PyErr_SetString(PyExc_TypeError,
239 239 "expected a dirstate tuple");
240 240 goto quit;
241 241 }
242 242
243 243 tuple = (dirstateTupleObject *)v;
244 244 if (tuple->state != 'r') {
245 245 PyObject *normed;
246 246 if (table != NULL) {
247 247 normed = _asciitransform(k, table,
248 248 normcase_fallback);
249 249 } else {
250 250 normed = PyObject_CallFunctionObjArgs(
251 251 normcase_fallback, k, NULL);
252 252 }
253 253
254 254 if (normed == NULL)
255 255 goto quit;
256 256 if (PyDict_SetItem(file_foldmap, normed, k) == -1)
257 257 goto quit;
258 258 }
259 259 }
260 260 return file_foldmap;
261 261 quit:
262 262 Py_XDECREF(file_foldmap);
263 263 return NULL;
264 264 }
265 265
266 266 /*
267 267 * This code assumes that a manifest is stitched together with newline
268 268 * ('\n') characters.
269 269 */
270 270 static PyObject *parse_manifest(PyObject *self, PyObject *args)
271 271 {
272 272 PyObject *mfdict, *fdict;
273 273 char *str, *start, *end;
274 274 int len;
275 275
276 276 if (!PyArg_ParseTuple(args, "O!O!s#:parse_manifest",
277 277 &PyDict_Type, &mfdict,
278 278 &PyDict_Type, &fdict,
279 279 &str, &len))
280 280 goto quit;
281 281
282 282 start = str;
283 283 end = str + len;
284 284 while (start < end) {
285 285 PyObject *file = NULL, *node = NULL;
286 286 PyObject *flags = NULL;
287 287 char *zero = NULL, *newline = NULL;
288 288 ptrdiff_t nlen;
289 289
290 290 zero = memchr(start, '\0', end - start);
291 291 if (!zero) {
292 292 PyErr_SetString(PyExc_ValueError,
293 293 "manifest entry has no separator");
294 294 goto quit;
295 295 }
296 296
297 297 newline = memchr(zero + 1, '\n', end - (zero + 1));
298 298 if (!newline) {
299 299 PyErr_SetString(PyExc_ValueError,
300 300 "manifest contains trailing garbage");
301 301 goto quit;
302 302 }
303 303
304 304 file = PyBytes_FromStringAndSize(start, zero - start);
305 305
306 306 if (!file)
307 307 goto bail;
308 308
309 309 nlen = newline - zero - 1;
310 310
311 311 node = unhexlify(zero + 1, nlen > 40 ? 40 : (int)nlen);
312 312 if (!node)
313 313 goto bail;
314 314
315 315 if (nlen > 40) {
316 316 flags = PyBytes_FromStringAndSize(zero + 41,
317 317 nlen - 40);
318 318 if (!flags)
319 319 goto bail;
320 320
321 321 if (PyDict_SetItem(fdict, file, flags) == -1)
322 322 goto bail;
323 323 }
324 324
325 325 if (PyDict_SetItem(mfdict, file, node) == -1)
326 326 goto bail;
327 327
328 328 start = newline + 1;
329 329
330 330 Py_XDECREF(flags);
331 331 Py_XDECREF(node);
332 332 Py_XDECREF(file);
333 333 continue;
334 334 bail:
335 335 Py_XDECREF(flags);
336 336 Py_XDECREF(node);
337 337 Py_XDECREF(file);
338 338 goto quit;
339 339 }
340 340
341 341 Py_INCREF(Py_None);
342 342 return Py_None;
343 343 quit:
344 344 return NULL;
345 345 }
346 346
347 347 static inline dirstateTupleObject *make_dirstate_tuple(char state, int mode,
348 348 int size, int mtime)
349 349 {
350 350 dirstateTupleObject *t = PyObject_New(dirstateTupleObject,
351 351 &dirstateTupleType);
352 352 if (!t)
353 353 return NULL;
354 354 t->state = state;
355 355 t->mode = mode;
356 356 t->size = size;
357 357 t->mtime = mtime;
358 358 return t;
359 359 }
360 360
361 361 static PyObject *dirstate_tuple_new(PyTypeObject *subtype, PyObject *args,
362 362 PyObject *kwds)
363 363 {
364 364 /* We do all the initialization here and not a tp_init function because
365 365 * dirstate_tuple is immutable. */
366 366 dirstateTupleObject *t;
367 367 char state;
368 368 int size, mode, mtime;
369 369 if (!PyArg_ParseTuple(args, "ciii", &state, &mode, &size, &mtime))
370 370 return NULL;
371 371
372 372 t = (dirstateTupleObject *)subtype->tp_alloc(subtype, 1);
373 373 if (!t)
374 374 return NULL;
375 375 t->state = state;
376 376 t->mode = mode;
377 377 t->size = size;
378 378 t->mtime = mtime;
379 379
380 380 return (PyObject *)t;
381 381 }
382 382
383 383 static void dirstate_tuple_dealloc(PyObject *o)
384 384 {
385 385 PyObject_Del(o);
386 386 }
387 387
388 388 static Py_ssize_t dirstate_tuple_length(PyObject *o)
389 389 {
390 390 return 4;
391 391 }
392 392
393 393 static PyObject *dirstate_tuple_item(PyObject *o, Py_ssize_t i)
394 394 {
395 395 dirstateTupleObject *t = (dirstateTupleObject *)o;
396 396 switch (i) {
397 397 case 0:
398 398 return PyBytes_FromStringAndSize(&t->state, 1);
399 399 case 1:
400 400 return PyInt_FromLong(t->mode);
401 401 case 2:
402 402 return PyInt_FromLong(t->size);
403 403 case 3:
404 404 return PyInt_FromLong(t->mtime);
405 405 default:
406 406 PyErr_SetString(PyExc_IndexError, "index out of range");
407 407 return NULL;
408 408 }
409 409 }
410 410
411 411 static PySequenceMethods dirstate_tuple_sq = {
412 412 dirstate_tuple_length, /* sq_length */
413 413 0, /* sq_concat */
414 414 0, /* sq_repeat */
415 415 dirstate_tuple_item, /* sq_item */
416 416 0, /* sq_ass_item */
417 417 0, /* sq_contains */
418 418 0, /* sq_inplace_concat */
419 419 0 /* sq_inplace_repeat */
420 420 };
421 421
422 422 PyTypeObject dirstateTupleType = {
423 423 PyVarObject_HEAD_INIT(NULL, 0)
424 424 "dirstate_tuple", /* tp_name */
425 425 sizeof(dirstateTupleObject),/* tp_basicsize */
426 426 0, /* tp_itemsize */
427 427 (destructor)dirstate_tuple_dealloc, /* tp_dealloc */
428 428 0, /* tp_print */
429 429 0, /* tp_getattr */
430 430 0, /* tp_setattr */
431 431 0, /* tp_compare */
432 432 0, /* tp_repr */
433 433 0, /* tp_as_number */
434 434 &dirstate_tuple_sq, /* tp_as_sequence */
435 435 0, /* tp_as_mapping */
436 436 0, /* tp_hash */
437 437 0, /* tp_call */
438 438 0, /* tp_str */
439 439 0, /* tp_getattro */
440 440 0, /* tp_setattro */
441 441 0, /* tp_as_buffer */
442 442 Py_TPFLAGS_DEFAULT, /* tp_flags */
443 443 "dirstate tuple", /* tp_doc */
444 444 0, /* tp_traverse */
445 445 0, /* tp_clear */
446 446 0, /* tp_richcompare */
447 447 0, /* tp_weaklistoffset */
448 448 0, /* tp_iter */
449 449 0, /* tp_iternext */
450 450 0, /* tp_methods */
451 451 0, /* tp_members */
452 452 0, /* tp_getset */
453 453 0, /* tp_base */
454 454 0, /* tp_dict */
455 455 0, /* tp_descr_get */
456 456 0, /* tp_descr_set */
457 457 0, /* tp_dictoffset */
458 458 0, /* tp_init */
459 459 0, /* tp_alloc */
460 460 dirstate_tuple_new, /* tp_new */
461 461 };
462 462
463 463 static PyObject *parse_dirstate(PyObject *self, PyObject *args)
464 464 {
465 465 PyObject *dmap, *cmap, *parents = NULL, *ret = NULL;
466 466 PyObject *fname = NULL, *cname = NULL, *entry = NULL;
467 467 char state, *cur, *str, *cpos;
468 468 int mode, size, mtime;
469 469 unsigned int flen, len, pos = 40;
470 470 int readlen;
471 471
472 472 if (!PyArg_ParseTuple(args, "O!O!s#:parse_dirstate",
473 473 &PyDict_Type, &dmap,
474 474 &PyDict_Type, &cmap,
475 475 &str, &readlen))
476 476 goto quit;
477 477
478 478 if (readlen < 0)
479 479 goto quit;
480 480
481 481 len = readlen;
482 482
483 483 /* read parents */
484 484 if (len < 40)
485 485 goto quit;
486 486
487 487 parents = Py_BuildValue("s#s#", str, 20, str + 20, 20);
488 488 if (!parents)
489 489 goto quit;
490 490
491 491 /* read filenames */
492 492 while (pos >= 40 && pos < len) {
493 493 cur = str + pos;
494 494 /* unpack header */
495 495 state = *cur;
496 496 mode = getbe32(cur + 1);
497 497 size = getbe32(cur + 5);
498 498 mtime = getbe32(cur + 9);
499 499 flen = getbe32(cur + 13);
500 500 pos += 17;
501 501 cur += 17;
502 502 if (flen > len - pos) {
503 503 PyErr_SetString(PyExc_ValueError, "overflow in dirstate");
504 504 goto quit;
505 505 }
506 506
507 507 entry = (PyObject *)make_dirstate_tuple(state, mode, size,
508 508 mtime);
509 509 cpos = memchr(cur, 0, flen);
510 510 if (cpos) {
511 511 fname = PyBytes_FromStringAndSize(cur, cpos - cur);
512 512 cname = PyBytes_FromStringAndSize(cpos + 1,
513 513 flen - (cpos - cur) - 1);
514 514 if (!fname || !cname ||
515 515 PyDict_SetItem(cmap, fname, cname) == -1 ||
516 516 PyDict_SetItem(dmap, fname, entry) == -1)
517 517 goto quit;
518 518 Py_DECREF(cname);
519 519 } else {
520 520 fname = PyBytes_FromStringAndSize(cur, flen);
521 521 if (!fname ||
522 522 PyDict_SetItem(dmap, fname, entry) == -1)
523 523 goto quit;
524 524 }
525 525 Py_DECREF(fname);
526 526 Py_DECREF(entry);
527 527 fname = cname = entry = NULL;
528 528 pos += flen;
529 529 }
530 530
531 531 ret = parents;
532 532 Py_INCREF(ret);
533 533 quit:
534 534 Py_XDECREF(fname);
535 535 Py_XDECREF(cname);
536 536 Py_XDECREF(entry);
537 537 Py_XDECREF(parents);
538 538 return ret;
539 539 }
540 540
541 541 /*
542 542 * Efficiently pack a dirstate object into its on-disk format.
543 543 */
544 544 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
545 545 {
546 546 PyObject *packobj = NULL;
547 547 PyObject *map, *copymap, *pl, *mtime_unset = NULL;
548 548 Py_ssize_t nbytes, pos, l;
549 549 PyObject *k, *v = NULL, *pn;
550 550 char *p, *s;
551 551 double now;
552 552
553 553 if (!PyArg_ParseTuple(args, "O!O!Od:pack_dirstate",
554 554 &PyDict_Type, &map, &PyDict_Type, &copymap,
555 555 &pl, &now))
556 556 return NULL;
557 557
558 558 if (!PySequence_Check(pl) || PySequence_Size(pl) != 2) {
559 559 PyErr_SetString(PyExc_TypeError, "expected 2-element sequence");
560 560 return NULL;
561 561 }
562 562
563 563 /* Figure out how much we need to allocate. */
564 564 for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
565 565 PyObject *c;
566 566 if (!PyString_Check(k)) {
567 567 PyErr_SetString(PyExc_TypeError, "expected string key");
568 568 goto bail;
569 569 }
570 570 nbytes += PyString_GET_SIZE(k) + 17;
571 571 c = PyDict_GetItem(copymap, k);
572 572 if (c) {
573 573 if (!PyString_Check(c)) {
574 574 PyErr_SetString(PyExc_TypeError,
575 575 "expected string key");
576 576 goto bail;
577 577 }
578 578 nbytes += PyString_GET_SIZE(c) + 1;
579 579 }
580 580 }
581 581
582 582 packobj = PyString_FromStringAndSize(NULL, nbytes);
583 583 if (packobj == NULL)
584 584 goto bail;
585 585
586 586 p = PyString_AS_STRING(packobj);
587 587
588 588 pn = PySequence_ITEM(pl, 0);
589 589 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
590 590 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
591 591 goto bail;
592 592 }
593 593 memcpy(p, s, l);
594 594 p += 20;
595 595 pn = PySequence_ITEM(pl, 1);
596 596 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
597 597 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
598 598 goto bail;
599 599 }
600 600 memcpy(p, s, l);
601 601 p += 20;
602 602
603 603 for (pos = 0; PyDict_Next(map, &pos, &k, &v); ) {
604 604 dirstateTupleObject *tuple;
605 605 char state;
606 606 uint32_t mode, size, mtime;
607 607 Py_ssize_t len, l;
608 608 PyObject *o;
609 609 char *t;
610 610
611 611 if (!dirstate_tuple_check(v)) {
612 612 PyErr_SetString(PyExc_TypeError,
613 613 "expected a dirstate tuple");
614 614 goto bail;
615 615 }
616 616 tuple = (dirstateTupleObject *)v;
617 617
618 618 state = tuple->state;
619 619 mode = tuple->mode;
620 620 size = tuple->size;
621 621 mtime = tuple->mtime;
622 622 if (state == 'n' && mtime == (uint32_t)now) {
623 623 /* See pure/parsers.py:pack_dirstate for why we do
624 624 * this. */
625 625 mtime = -1;
626 626 mtime_unset = (PyObject *)make_dirstate_tuple(
627 627 state, mode, size, mtime);
628 628 if (!mtime_unset)
629 629 goto bail;
630 630 if (PyDict_SetItem(map, k, mtime_unset) == -1)
631 631 goto bail;
632 632 Py_DECREF(mtime_unset);
633 633 mtime_unset = NULL;
634 634 }
635 635 *p++ = state;
636 636 putbe32(mode, p);
637 637 putbe32(size, p + 4);
638 638 putbe32(mtime, p + 8);
639 639 t = p + 12;
640 640 p += 16;
641 641 len = PyString_GET_SIZE(k);
642 642 memcpy(p, PyString_AS_STRING(k), len);
643 643 p += len;
644 644 o = PyDict_GetItem(copymap, k);
645 645 if (o) {
646 646 *p++ = '\0';
647 647 l = PyString_GET_SIZE(o);
648 648 memcpy(p, PyString_AS_STRING(o), l);
649 649 p += l;
650 650 len += l + 1;
651 651 }
652 652 putbe32((uint32_t)len, t);
653 653 }
654 654
655 655 pos = p - PyString_AS_STRING(packobj);
656 656 if (pos != nbytes) {
657 657 PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
658 658 (long)pos, (long)nbytes);
659 659 goto bail;
660 660 }
661 661
662 662 return packobj;
663 663 bail:
664 664 Py_XDECREF(mtime_unset);
665 665 Py_XDECREF(packobj);
666 666 Py_XDECREF(v);
667 667 return NULL;
668 668 }
669 669
670 670 /*
671 671 * A base-16 trie for fast node->rev mapping.
672 672 *
673 673 * Positive value is index of the next node in the trie
674 674 * Negative value is a leaf: -(rev + 1)
675 675 * Zero is empty
676 676 */
677 677 typedef struct {
678 678 int children[16];
679 679 } nodetree;
680 680
681 681 /*
682 682 * This class has two behaviours.
683 683 *
684 684 * When used in a list-like way (with integer keys), we decode an
685 685 * entry in a RevlogNG index file on demand. Our last entry is a
686 686 * sentinel, always a nullid. We have limited support for
687 687 * integer-keyed insert and delete, only at elements right before the
688 688 * sentinel.
689 689 *
690 690 * With string keys, we lazily perform a reverse mapping from node to
691 691 * rev, using a base-16 trie.
692 692 */
693 693 typedef struct {
694 694 PyObject_HEAD
695 695 /* Type-specific fields go here. */
696 696 PyObject *data; /* raw bytes of index */
697 697 PyObject **cache; /* cached tuples */
698 698 const char **offsets; /* populated on demand */
699 699 Py_ssize_t raw_length; /* original number of elements */
700 700 Py_ssize_t length; /* current number of elements */
701 701 PyObject *added; /* populated on demand */
702 702 PyObject *headrevs; /* cache, invalidated on changes */
703 703 PyObject *filteredrevs;/* filtered revs set */
704 704 nodetree *nt; /* base-16 trie */
705 705 int ntlength; /* # nodes in use */
706 706 int ntcapacity; /* # nodes allocated */
707 707 int ntdepth; /* maximum depth of tree */
708 708 int ntsplits; /* # splits performed */
709 709 int ntrev; /* last rev scanned */
710 710 int ntlookups; /* # lookups */
711 711 int ntmisses; /* # lookups that miss the cache */
712 712 int inlined;
713 713 } indexObject;
714 714
715 715 static Py_ssize_t index_length(const indexObject *self)
716 716 {
717 717 if (self->added == NULL)
718 718 return self->length;
719 719 return self->length + PyList_GET_SIZE(self->added);
720 720 }
721 721
722 722 static PyObject *nullentry;
723 723 static const char nullid[20];
724 724
725 725 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
726 726
727 727 #if LONG_MAX == 0x7fffffffL
728 728 static char *tuple_format = "Kiiiiiis#";
729 729 #else
730 730 static char *tuple_format = "kiiiiiis#";
731 731 #endif
732 732
733 733 /* A RevlogNG v1 index entry is 64 bytes long. */
734 734 static const long v1_hdrsize = 64;
735 735
736 736 /*
737 737 * Return a pointer to the beginning of a RevlogNG record.
738 738 */
739 739 static const char *index_deref(indexObject *self, Py_ssize_t pos)
740 740 {
741 741 if (self->inlined && pos > 0) {
742 742 if (self->offsets == NULL) {
743 743 self->offsets = malloc(self->raw_length *
744 744 sizeof(*self->offsets));
745 745 if (self->offsets == NULL)
746 746 return (const char *)PyErr_NoMemory();
747 747 inline_scan(self, self->offsets);
748 748 }
749 749 return self->offsets[pos];
750 750 }
751 751
752 752 return PyString_AS_STRING(self->data) + pos * v1_hdrsize;
753 753 }
754 754
755 755 static inline int index_get_parents(indexObject *self, Py_ssize_t rev,
756 756 int *ps, int maxrev)
757 757 {
758 758 if (rev >= self->length - 1) {
759 759 PyObject *tuple = PyList_GET_ITEM(self->added,
760 760 rev - self->length + 1);
761 761 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
762 762 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
763 763 } else {
764 764 const char *data = index_deref(self, rev);
765 765 ps[0] = getbe32(data + 24);
766 766 ps[1] = getbe32(data + 28);
767 767 }
768 768 /* If index file is corrupted, ps[] may point to invalid revisions. So
769 769 * there is a risk of buffer overflow to trust them unconditionally. */
770 770 if (ps[0] > maxrev || ps[1] > maxrev) {
771 771 PyErr_SetString(PyExc_ValueError, "parent out of range");
772 772 return -1;
773 773 }
774 774 return 0;
775 775 }
776 776
777 777
778 778 /*
779 779 * RevlogNG format (all in big endian, data may be inlined):
780 780 * 6 bytes: offset
781 781 * 2 bytes: flags
782 782 * 4 bytes: compressed length
783 783 * 4 bytes: uncompressed length
784 784 * 4 bytes: base revision
785 785 * 4 bytes: link revision
786 786 * 4 bytes: parent 1 revision
787 787 * 4 bytes: parent 2 revision
788 788 * 32 bytes: nodeid (only 20 bytes used)
789 789 */
790 790 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
791 791 {
792 792 uint64_t offset_flags;
793 793 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
794 794 const char *c_node_id;
795 795 const char *data;
796 796 Py_ssize_t length = index_length(self);
797 797 PyObject *entry;
798 798
799 799 if (pos < 0)
800 800 pos += length;
801 801
802 802 if (pos < 0 || pos >= length) {
803 803 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
804 804 return NULL;
805 805 }
806 806
807 807 if (pos == length - 1) {
808 808 Py_INCREF(nullentry);
809 809 return nullentry;
810 810 }
811 811
812 812 if (pos >= self->length - 1) {
813 813 PyObject *obj;
814 814 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
815 815 Py_INCREF(obj);
816 816 return obj;
817 817 }
818 818
819 819 if (self->cache) {
820 820 if (self->cache[pos]) {
821 821 Py_INCREF(self->cache[pos]);
822 822 return self->cache[pos];
823 823 }
824 824 } else {
825 825 self->cache = calloc(self->raw_length, sizeof(PyObject *));
826 826 if (self->cache == NULL)
827 827 return PyErr_NoMemory();
828 828 }
829 829
830 830 data = index_deref(self, pos);
831 831 if (data == NULL)
832 832 return NULL;
833 833
834 834 offset_flags = getbe32(data + 4);
835 835 if (pos == 0) /* mask out version number for the first entry */
836 836 offset_flags &= 0xFFFF;
837 837 else {
838 838 uint32_t offset_high = getbe32(data);
839 839 offset_flags |= ((uint64_t)offset_high) << 32;
840 840 }
841 841
842 842 comp_len = getbe32(data + 8);
843 843 uncomp_len = getbe32(data + 12);
844 844 base_rev = getbe32(data + 16);
845 845 link_rev = getbe32(data + 20);
846 846 parent_1 = getbe32(data + 24);
847 847 parent_2 = getbe32(data + 28);
848 848 c_node_id = data + 32;
849 849
850 850 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
851 851 uncomp_len, base_rev, link_rev,
852 852 parent_1, parent_2, c_node_id, 20);
853 853
854 854 if (entry) {
855 855 PyObject_GC_UnTrack(entry);
856 856 Py_INCREF(entry);
857 857 }
858 858
859 859 self->cache[pos] = entry;
860 860
861 861 return entry;
862 862 }
863 863
864 864 /*
865 865 * Return the 20-byte SHA of the node corresponding to the given rev.
866 866 */
867 867 static const char *index_node(indexObject *self, Py_ssize_t pos)
868 868 {
869 869 Py_ssize_t length = index_length(self);
870 870 const char *data;
871 871
872 872 if (pos == length - 1 || pos == INT_MAX)
873 873 return nullid;
874 874
875 875 if (pos >= length)
876 876 return NULL;
877 877
878 878 if (pos >= self->length - 1) {
879 879 PyObject *tuple, *str;
880 880 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
881 881 str = PyTuple_GetItem(tuple, 7);
882 882 return str ? PyString_AS_STRING(str) : NULL;
883 883 }
884 884
885 885 data = index_deref(self, pos);
886 886 return data ? data + 32 : NULL;
887 887 }
888 888
889 889 static int nt_insert(indexObject *self, const char *node, int rev);
890 890
891 891 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
892 892 {
893 893 if (PyString_AsStringAndSize(obj, node, nodelen) == -1)
894 894 return -1;
895 895 if (*nodelen == 20)
896 896 return 0;
897 897 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
898 898 return -1;
899 899 }
900 900
901 901 static PyObject *index_insert(indexObject *self, PyObject *args)
902 902 {
903 903 PyObject *obj;
904 904 char *node;
905 905 int index;
906 906 Py_ssize_t len, nodelen;
907 907
908 908 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
909 909 return NULL;
910 910
911 911 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
912 912 PyErr_SetString(PyExc_TypeError, "8-tuple required");
913 913 return NULL;
914 914 }
915 915
916 916 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
917 917 return NULL;
918 918
919 919 len = index_length(self);
920 920
921 921 if (index < 0)
922 922 index += len;
923 923
924 924 if (index != len - 1) {
925 925 PyErr_SetString(PyExc_IndexError,
926 926 "insert only supported at index -1");
927 927 return NULL;
928 928 }
929 929
930 930 if (self->added == NULL) {
931 931 self->added = PyList_New(0);
932 932 if (self->added == NULL)
933 933 return NULL;
934 934 }
935 935
936 936 if (PyList_Append(self->added, obj) == -1)
937 937 return NULL;
938 938
939 939 if (self->nt)
940 940 nt_insert(self, node, index);
941 941
942 942 Py_CLEAR(self->headrevs);
943 943 Py_RETURN_NONE;
944 944 }
945 945
946 946 static void _index_clearcaches(indexObject *self)
947 947 {
948 948 if (self->cache) {
949 949 Py_ssize_t i;
950 950
951 951 for (i = 0; i < self->raw_length; i++)
952 952 Py_CLEAR(self->cache[i]);
953 953 free(self->cache);
954 954 self->cache = NULL;
955 955 }
956 956 if (self->offsets) {
957 957 free(self->offsets);
958 958 self->offsets = NULL;
959 959 }
960 960 if (self->nt) {
961 961 free(self->nt);
962 962 self->nt = NULL;
963 963 }
964 964 Py_CLEAR(self->headrevs);
965 965 }
966 966
967 967 static PyObject *index_clearcaches(indexObject *self)
968 968 {
969 969 _index_clearcaches(self);
970 970 self->ntlength = self->ntcapacity = 0;
971 971 self->ntdepth = self->ntsplits = 0;
972 972 self->ntrev = -1;
973 973 self->ntlookups = self->ntmisses = 0;
974 974 Py_RETURN_NONE;
975 975 }
976 976
977 977 static PyObject *index_stats(indexObject *self)
978 978 {
979 979 PyObject *obj = PyDict_New();
980 980 PyObject *t = NULL;
981 981
982 982 if (obj == NULL)
983 983 return NULL;
984 984
985 985 #define istat(__n, __d) \
986 986 t = PyInt_FromSsize_t(self->__n); \
987 987 if (!t) \
988 988 goto bail; \
989 989 if (PyDict_SetItemString(obj, __d, t) == -1) \
990 990 goto bail; \
991 991 Py_DECREF(t);
992 992
993 993 if (self->added) {
994 994 Py_ssize_t len = PyList_GET_SIZE(self->added);
995 995 t = PyInt_FromSsize_t(len);
996 996 if (!t)
997 997 goto bail;
998 998 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
999 999 goto bail;
1000 1000 Py_DECREF(t);
1001 1001 }
1002 1002
1003 1003 if (self->raw_length != self->length - 1)
1004 1004 istat(raw_length, "revs on disk");
1005 1005 istat(length, "revs in memory");
1006 1006 istat(ntcapacity, "node trie capacity");
1007 1007 istat(ntdepth, "node trie depth");
1008 1008 istat(ntlength, "node trie count");
1009 1009 istat(ntlookups, "node trie lookups");
1010 1010 istat(ntmisses, "node trie misses");
1011 1011 istat(ntrev, "node trie last rev scanned");
1012 1012 istat(ntsplits, "node trie splits");
1013 1013
1014 1014 #undef istat
1015 1015
1016 1016 return obj;
1017 1017
1018 1018 bail:
1019 1019 Py_XDECREF(obj);
1020 1020 Py_XDECREF(t);
1021 1021 return NULL;
1022 1022 }
1023 1023
1024 1024 /*
1025 1025 * When we cache a list, we want to be sure the caller can't mutate
1026 1026 * the cached copy.
1027 1027 */
1028 1028 static PyObject *list_copy(PyObject *list)
1029 1029 {
1030 1030 Py_ssize_t len = PyList_GET_SIZE(list);
1031 1031 PyObject *newlist = PyList_New(len);
1032 1032 Py_ssize_t i;
1033 1033
1034 1034 if (newlist == NULL)
1035 1035 return NULL;
1036 1036
1037 1037 for (i = 0; i < len; i++) {
1038 1038 PyObject *obj = PyList_GET_ITEM(list, i);
1039 1039 Py_INCREF(obj);
1040 1040 PyList_SET_ITEM(newlist, i, obj);
1041 1041 }
1042 1042
1043 1043 return newlist;
1044 1044 }
1045 1045
1046 1046 /* arg should be Py_ssize_t but Python 2.4 do not support the n format */
1047 1047 static int check_filter(PyObject *filter, unsigned long arg) {
1048 1048 if (filter) {
1049 1049 PyObject *arglist, *result;
1050 1050 int isfiltered;
1051 1051
1052 1052 arglist = Py_BuildValue("(k)", arg);
1053 1053 if (!arglist) {
1054 1054 return -1;
1055 1055 }
1056 1056
1057 1057 result = PyEval_CallObject(filter, arglist);
1058 1058 Py_DECREF(arglist);
1059 1059 if (!result) {
1060 1060 return -1;
1061 1061 }
1062 1062
1063 1063 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
1064 1064 * same as this function, so we can just return it directly.*/
1065 1065 isfiltered = PyObject_IsTrue(result);
1066 1066 Py_DECREF(result);
1067 1067 return isfiltered;
1068 1068 } else {
1069 1069 return 0;
1070 1070 }
1071 1071 }
1072 1072
1073 1073 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
1074 1074 Py_ssize_t marker, char *phases)
1075 1075 {
1076 1076 PyObject *iter = NULL;
1077 1077 PyObject *iter_item = NULL;
1078 1078 Py_ssize_t min_idx = index_length(self) + 1;
1079 1079 long iter_item_long;
1080 1080
1081 1081 if (PyList_GET_SIZE(list) != 0) {
1082 1082 iter = PyObject_GetIter(list);
1083 1083 if (iter == NULL)
1084 1084 return -2;
1085 1085 while ((iter_item = PyIter_Next(iter)))
1086 1086 {
1087 1087 iter_item_long = PyInt_AS_LONG(iter_item);
1088 1088 Py_DECREF(iter_item);
1089 1089 if (iter_item_long < min_idx)
1090 1090 min_idx = iter_item_long;
1091 1091 phases[iter_item_long] = marker;
1092 1092 }
1093 1093 Py_DECREF(iter);
1094 1094 }
1095 1095
1096 1096 return min_idx;
1097 1097 }
1098 1098
1099 1099 static inline void set_phase_from_parents(char *phases, int parent_1,
1100 1100 int parent_2, Py_ssize_t i)
1101 1101 {
1102 1102 if (parent_1 >= 0 && phases[parent_1] > phases[i])
1103 1103 phases[i] = phases[parent_1];
1104 1104 if (parent_2 >= 0 && phases[parent_2] > phases[i])
1105 1105 phases[i] = phases[parent_2];
1106 1106 }
1107 1107
1108 1108 static PyObject *reachableroots(indexObject *self, PyObject *args)
1109 1109 {
1110 1110
1111 1111 /* Input */
1112 1112 long minroot;
1113 1113 PyObject *includepatharg = NULL;
1114 1114 int includepath = 0;
1115 1115 /* heads is a list */
1116 1116 PyObject *heads = NULL;
1117 1117 Py_ssize_t numheads;
1118 1118 /* roots is a set */
1119 1119 PyObject *roots = NULL;
1120 1120 PyObject *reachable = NULL;
1121 1121
1122 1122 PyObject *val;
1123 1123 Py_ssize_t len = index_length(self) - 1;
1124 1124 long revnum;
1125 1125 Py_ssize_t k;
1126 1126 Py_ssize_t i;
1127 1127 int r;
1128 1128 int minidx;
1129 1129 int parents[2];
1130 1130
1131 1131 /* Internal data structure:
1132 1132 * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit
1133 1133 * seen: array of length len+1 (all revs + nullrev) 0: not seen, 1 seen*/
1134 1134 int *tovisit = NULL;
1135 1135 long lentovisit = 0;
1136 1136 char *seen = NULL;
1137 1137
1138 1138 /* Get arguments */
1139 1139 if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
1140 1140 &PySet_Type, &roots, &PyBool_Type, &includepatharg))
1141 1141 goto bail;
1142 1142
1143 1143 if (includepatharg == Py_True)
1144 1144 includepath = 1;
1145 1145
1146 1146 /* Initialize return set */
1147 1147 reachable = PySet_New(NULL);
1148 1148 if (reachable == NULL) {
1149 1149 PyErr_NoMemory();
1150 1150 goto bail;
1151 1151 }
1152 1152
1153 1153 /* Initialize internal datastructures */
1154 1154 tovisit = (int *)malloc((len + 1) * sizeof(int));
1155 1155 if (tovisit == NULL) {
1156 1156 PyErr_NoMemory();
1157 1157 goto bail;
1158 1158 }
1159 1159
1160 1160 seen = (char *)calloc(len+1, 1);
1161 1161 if (seen == NULL) {
1162 1162 PyErr_NoMemory();
1163 1163 goto bail;
1164 1164 }
1165 1165
1166 1166 /* Populate tovisit with all the heads */
1167 1167 numheads = PyList_GET_SIZE(heads);
1168 1168 for (i = 0; i < numheads; i++) {
1169 1169 revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
1170 1170 if (revnum == -1 && PyErr_Occurred())
1171 1171 goto bail;
1172 1172 if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
1173 1173 PyErr_SetString(PyExc_IndexError, "head out of range");
1174 1174 goto bail;
1175 1175 }
1176 1176 if (seen[revnum+1] == 0) {
1177 1177 tovisit[lentovisit++] = revnum;
1178 1178 seen[revnum+1]=1;
1179 1179 }
1180 1180 }
1181 1181
1182 1182 /* Visit the tovisit list and find the reachable roots */
1183 1183 k = 0;
1184 1184 while (k < lentovisit) {
1185 1185 /* Add the node to reachable if it is a root*/
1186 1186 revnum = tovisit[k++];
1187 1187 val = PyInt_FromLong(revnum);
1188 1188 if (val == NULL)
1189 1189 goto bail;
1190 1190 if (PySet_Contains(roots, val) == 1) {
1191 1191 PySet_Add(reachable, val);
1192 1192 if (includepath == 0) {
1193 1193 Py_DECREF(val);
1194 1194 continue;
1195 1195 }
1196 1196 }
1197 1197 Py_DECREF(val);
1198 1198
1199 1199 /* Add its parents to the list of nodes to visit */
1200 1200 if (revnum != -1) {
1201 1201 r = index_get_parents(self, revnum, parents, (int)len - 1);
1202 1202 if (r < 0)
1203 1203 goto bail;
1204 1204
1205 1205 for (i = 0; i < 2; i++) {
1206 1206 if (seen[parents[i] + 1] == 0 && parents[i] >= minroot) {
1207 1207 tovisit[lentovisit++] = parents[i];
1208 1208 seen[parents[i] + 1] = 1;
1209 1209 }
1210 1210 }
1211 1211 }
1212 1212 }
1213 1213
1214 1214 /* Find all the nodes in between the roots we found and the heads
1215 1215 * and add them to the reachable set */
1216 1216 if (includepath == 1) {
1217 1217 minidx = minroot;
1218 1218 if (minidx < 0)
1219 1219 minidx = 0;
1220 1220 for (i = minidx; i < len; i++) {
1221 1221 if (seen[i + 1] == 1) {
1222 1222 r = index_get_parents(self, i, parents, (int)len - 1);
1223 1223 /* Corrupted index file, error is set from index_get_parents */
1224 1224 if (r < 0)
1225 1225 goto bail;
1226 1226 for (k = 0; k < 2; k++) {
1227 1227 PyObject *p = PyInt_FromLong(parents[k]);
1228 1228 if (p == NULL)
1229 1229 goto bail;
1230 1230 if (PySet_Contains(reachable, p) == 1) {
1231 1231 val = PyInt_FromLong(i);
1232 1232 if (val == NULL)
1233 1233 goto bail;
1234 1234 PySet_Add(reachable, val);
1235 Py_DECREF(val);
1235 1236 }
1236 1237 Py_DECREF(p);
1237 1238 }
1238 1239 }
1239 1240 }
1240 1241 }
1241 1242
1242 1243 free(seen);
1243 1244 free(tovisit);
1244 1245 return reachable;
1245 1246 bail:
1246 1247 Py_XDECREF(reachable);
1247 1248 free(seen);
1248 1249 free(tovisit);
1249 1250 return NULL;
1250 1251 }
1251 1252
1252 1253 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
1253 1254 {
1254 1255 PyObject *roots = Py_None;
1255 1256 PyObject *ret = NULL;
1256 1257 PyObject *phaseslist = NULL;
1257 1258 PyObject *phaseroots = NULL;
1258 1259 PyObject *phaseset = NULL;
1259 1260 PyObject *phasessetlist = NULL;
1260 1261 PyObject *rev = NULL;
1261 1262 Py_ssize_t len = index_length(self) - 1;
1262 1263 Py_ssize_t numphase = 0;
1263 1264 Py_ssize_t minrevallphases = 0;
1264 1265 Py_ssize_t minrevphase = 0;
1265 1266 Py_ssize_t i = 0;
1266 1267 char *phases = NULL;
1267 1268 long phase;
1268 1269
1269 1270 if (!PyArg_ParseTuple(args, "O", &roots))
1270 1271 goto release_none;
1271 1272 if (roots == NULL || !PyList_Check(roots))
1272 1273 goto release_none;
1273 1274
1274 1275 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
1275 1276 if (phases == NULL)
1276 1277 goto release_none;
1277 1278 /* Put the phase information of all the roots in phases */
1278 1279 numphase = PyList_GET_SIZE(roots)+1;
1279 1280 minrevallphases = len + 1;
1280 1281 phasessetlist = PyList_New(numphase);
1281 1282 if (phasessetlist == NULL)
1282 1283 goto release_none;
1283 1284
1284 1285 PyList_SET_ITEM(phasessetlist, 0, Py_None);
1285 1286 Py_INCREF(Py_None);
1286 1287
1287 1288 for (i = 0; i < numphase-1; i++) {
1288 1289 phaseroots = PyList_GET_ITEM(roots, i);
1289 1290 phaseset = PySet_New(NULL);
1290 1291 if (phaseset == NULL)
1291 1292 goto release_phasesetlist;
1292 1293 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
1293 1294 if (!PyList_Check(phaseroots))
1294 1295 goto release_phasesetlist;
1295 1296 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
1296 1297 if (minrevphase == -2) /* Error from add_roots_get_min */
1297 1298 goto release_phasesetlist;
1298 1299 minrevallphases = MIN(minrevallphases, minrevphase);
1299 1300 }
1300 1301 /* Propagate the phase information from the roots to the revs */
1301 1302 if (minrevallphases != -1) {
1302 1303 int parents[2];
1303 1304 for (i = minrevallphases; i < len; i++) {
1304 1305 if (index_get_parents(self, i, parents,
1305 1306 (int)len - 1) < 0)
1306 1307 goto release_phasesetlist;
1307 1308 set_phase_from_parents(phases, parents[0], parents[1], i);
1308 1309 }
1309 1310 }
1310 1311 /* Transform phase list to a python list */
1311 1312 phaseslist = PyList_New(len);
1312 1313 if (phaseslist == NULL)
1313 1314 goto release_phasesetlist;
1314 1315 for (i = 0; i < len; i++) {
1315 1316 phase = phases[i];
1316 1317 /* We only store the sets of phase for non public phase, the public phase
1317 1318 * is computed as a difference */
1318 1319 if (phase != 0) {
1319 1320 phaseset = PyList_GET_ITEM(phasessetlist, phase);
1320 1321 rev = PyInt_FromLong(i);
1321 1322 PySet_Add(phaseset, rev);
1322 1323 Py_XDECREF(rev);
1323 1324 }
1324 1325 PyList_SET_ITEM(phaseslist, i, PyInt_FromLong(phase));
1325 1326 }
1326 1327 ret = PyList_New(2);
1327 1328 if (ret == NULL)
1328 1329 goto release_phaseslist;
1329 1330
1330 1331 PyList_SET_ITEM(ret, 0, phaseslist);
1331 1332 PyList_SET_ITEM(ret, 1, phasessetlist);
1332 1333 /* We don't release phaseslist and phasessetlist as we return them to
1333 1334 * python */
1334 1335 goto release_phases;
1335 1336
1336 1337 release_phaseslist:
1337 1338 Py_XDECREF(phaseslist);
1338 1339 release_phasesetlist:
1339 1340 Py_XDECREF(phasessetlist);
1340 1341 release_phases:
1341 1342 free(phases);
1342 1343 release_none:
1343 1344 return ret;
1344 1345 }
1345 1346
1346 1347 static PyObject *index_headrevs(indexObject *self, PyObject *args)
1347 1348 {
1348 1349 Py_ssize_t i, j, len;
1349 1350 char *nothead = NULL;
1350 1351 PyObject *heads = NULL;
1351 1352 PyObject *filter = NULL;
1352 1353 PyObject *filteredrevs = Py_None;
1353 1354
1354 1355 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
1355 1356 return NULL;
1356 1357 }
1357 1358
1358 1359 if (self->headrevs && filteredrevs == self->filteredrevs)
1359 1360 return list_copy(self->headrevs);
1360 1361
1361 1362 Py_DECREF(self->filteredrevs);
1362 1363 self->filteredrevs = filteredrevs;
1363 1364 Py_INCREF(filteredrevs);
1364 1365
1365 1366 if (filteredrevs != Py_None) {
1366 1367 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
1367 1368 if (!filter) {
1368 1369 PyErr_SetString(PyExc_TypeError,
1369 1370 "filteredrevs has no attribute __contains__");
1370 1371 goto bail;
1371 1372 }
1372 1373 }
1373 1374
1374 1375 len = index_length(self) - 1;
1375 1376 heads = PyList_New(0);
1376 1377 if (heads == NULL)
1377 1378 goto bail;
1378 1379 if (len == 0) {
1379 1380 PyObject *nullid = PyInt_FromLong(-1);
1380 1381 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
1381 1382 Py_XDECREF(nullid);
1382 1383 goto bail;
1383 1384 }
1384 1385 goto done;
1385 1386 }
1386 1387
1387 1388 nothead = calloc(len, 1);
1388 1389 if (nothead == NULL)
1389 1390 goto bail;
1390 1391
1391 1392 for (i = 0; i < len; i++) {
1392 1393 int isfiltered;
1393 1394 int parents[2];
1394 1395
1395 1396 isfiltered = check_filter(filter, i);
1396 1397 if (isfiltered == -1) {
1397 1398 PyErr_SetString(PyExc_TypeError,
1398 1399 "unable to check filter");
1399 1400 goto bail;
1400 1401 }
1401 1402
1402 1403 if (isfiltered) {
1403 1404 nothead[i] = 1;
1404 1405 continue;
1405 1406 }
1406 1407
1407 1408 if (index_get_parents(self, i, parents, (int)len - 1) < 0)
1408 1409 goto bail;
1409 1410 for (j = 0; j < 2; j++) {
1410 1411 if (parents[j] >= 0)
1411 1412 nothead[parents[j]] = 1;
1412 1413 }
1413 1414 }
1414 1415
1415 1416 for (i = 0; i < len; i++) {
1416 1417 PyObject *head;
1417 1418
1418 1419 if (nothead[i])
1419 1420 continue;
1420 1421 head = PyInt_FromSsize_t(i);
1421 1422 if (head == NULL || PyList_Append(heads, head) == -1) {
1422 1423 Py_XDECREF(head);
1423 1424 goto bail;
1424 1425 }
1425 1426 }
1426 1427
1427 1428 done:
1428 1429 self->headrevs = heads;
1429 1430 Py_XDECREF(filter);
1430 1431 free(nothead);
1431 1432 return list_copy(self->headrevs);
1432 1433 bail:
1433 1434 Py_XDECREF(filter);
1434 1435 Py_XDECREF(heads);
1435 1436 free(nothead);
1436 1437 return NULL;
1437 1438 }
1438 1439
1439 1440 static inline int nt_level(const char *node, Py_ssize_t level)
1440 1441 {
1441 1442 int v = node[level>>1];
1442 1443 if (!(level & 1))
1443 1444 v >>= 4;
1444 1445 return v & 0xf;
1445 1446 }
1446 1447
1447 1448 /*
1448 1449 * Return values:
1449 1450 *
1450 1451 * -4: match is ambiguous (multiple candidates)
1451 1452 * -2: not found
1452 1453 * rest: valid rev
1453 1454 */
1454 1455 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
1455 1456 int hex)
1456 1457 {
1457 1458 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
1458 1459 int level, maxlevel, off;
1459 1460
1460 1461 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
1461 1462 return -1;
1462 1463
1463 1464 if (self->nt == NULL)
1464 1465 return -2;
1465 1466
1466 1467 if (hex)
1467 1468 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
1468 1469 else
1469 1470 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
1470 1471
1471 1472 for (level = off = 0; level < maxlevel; level++) {
1472 1473 int k = getnybble(node, level);
1473 1474 nodetree *n = &self->nt[off];
1474 1475 int v = n->children[k];
1475 1476
1476 1477 if (v < 0) {
1477 1478 const char *n;
1478 1479 Py_ssize_t i;
1479 1480
1480 1481 v = -(v + 1);
1481 1482 n = index_node(self, v);
1482 1483 if (n == NULL)
1483 1484 return -2;
1484 1485 for (i = level; i < maxlevel; i++)
1485 1486 if (getnybble(node, i) != nt_level(n, i))
1486 1487 return -2;
1487 1488 return v;
1488 1489 }
1489 1490 if (v == 0)
1490 1491 return -2;
1491 1492 off = v;
1492 1493 }
1493 1494 /* multiple matches against an ambiguous prefix */
1494 1495 return -4;
1495 1496 }
1496 1497
1497 1498 static int nt_new(indexObject *self)
1498 1499 {
1499 1500 if (self->ntlength == self->ntcapacity) {
1500 1501 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1501 1502 PyErr_SetString(PyExc_MemoryError,
1502 1503 "overflow in nt_new");
1503 1504 return -1;
1504 1505 }
1505 1506 self->ntcapacity *= 2;
1506 1507 self->nt = realloc(self->nt,
1507 1508 self->ntcapacity * sizeof(nodetree));
1508 1509 if (self->nt == NULL) {
1509 1510 PyErr_SetString(PyExc_MemoryError, "out of memory");
1510 1511 return -1;
1511 1512 }
1512 1513 memset(&self->nt[self->ntlength], 0,
1513 1514 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1514 1515 }
1515 1516 return self->ntlength++;
1516 1517 }
1517 1518
1518 1519 static int nt_insert(indexObject *self, const char *node, int rev)
1519 1520 {
1520 1521 int level = 0;
1521 1522 int off = 0;
1522 1523
1523 1524 while (level < 40) {
1524 1525 int k = nt_level(node, level);
1525 1526 nodetree *n;
1526 1527 int v;
1527 1528
1528 1529 n = &self->nt[off];
1529 1530 v = n->children[k];
1530 1531
1531 1532 if (v == 0) {
1532 1533 n->children[k] = -rev - 1;
1533 1534 return 0;
1534 1535 }
1535 1536 if (v < 0) {
1536 1537 const char *oldnode = index_node(self, -(v + 1));
1537 1538 int noff;
1538 1539
1539 1540 if (!oldnode || !memcmp(oldnode, node, 20)) {
1540 1541 n->children[k] = -rev - 1;
1541 1542 return 0;
1542 1543 }
1543 1544 noff = nt_new(self);
1544 1545 if (noff == -1)
1545 1546 return -1;
1546 1547 /* self->nt may have been changed by realloc */
1547 1548 self->nt[off].children[k] = noff;
1548 1549 off = noff;
1549 1550 n = &self->nt[off];
1550 1551 n->children[nt_level(oldnode, ++level)] = v;
1551 1552 if (level > self->ntdepth)
1552 1553 self->ntdepth = level;
1553 1554 self->ntsplits += 1;
1554 1555 } else {
1555 1556 level += 1;
1556 1557 off = v;
1557 1558 }
1558 1559 }
1559 1560
1560 1561 return -1;
1561 1562 }
1562 1563
1563 1564 static int nt_init(indexObject *self)
1564 1565 {
1565 1566 if (self->nt == NULL) {
1566 1567 if (self->raw_length > INT_MAX / sizeof(nodetree)) {
1567 1568 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1568 1569 return -1;
1569 1570 }
1570 1571 self->ntcapacity = self->raw_length < 4
1571 1572 ? 4 : (int)self->raw_length / 2;
1572 1573
1573 1574 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1574 1575 if (self->nt == NULL) {
1575 1576 PyErr_NoMemory();
1576 1577 return -1;
1577 1578 }
1578 1579 self->ntlength = 1;
1579 1580 self->ntrev = (int)index_length(self) - 1;
1580 1581 self->ntlookups = 1;
1581 1582 self->ntmisses = 0;
1582 1583 if (nt_insert(self, nullid, INT_MAX) == -1)
1583 1584 return -1;
1584 1585 }
1585 1586 return 0;
1586 1587 }
1587 1588
1588 1589 /*
1589 1590 * Return values:
1590 1591 *
1591 1592 * -3: error (exception set)
1592 1593 * -2: not found (no exception set)
1593 1594 * rest: valid rev
1594 1595 */
1595 1596 static int index_find_node(indexObject *self,
1596 1597 const char *node, Py_ssize_t nodelen)
1597 1598 {
1598 1599 int rev;
1599 1600
1600 1601 self->ntlookups++;
1601 1602 rev = nt_find(self, node, nodelen, 0);
1602 1603 if (rev >= -1)
1603 1604 return rev;
1604 1605
1605 1606 if (nt_init(self) == -1)
1606 1607 return -3;
1607 1608
1608 1609 /*
1609 1610 * For the first handful of lookups, we scan the entire index,
1610 1611 * and cache only the matching nodes. This optimizes for cases
1611 1612 * like "hg tip", where only a few nodes are accessed.
1612 1613 *
1613 1614 * After that, we cache every node we visit, using a single
1614 1615 * scan amortized over multiple lookups. This gives the best
1615 1616 * bulk performance, e.g. for "hg log".
1616 1617 */
1617 1618 if (self->ntmisses++ < 4) {
1618 1619 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1619 1620 const char *n = index_node(self, rev);
1620 1621 if (n == NULL)
1621 1622 return -2;
1622 1623 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1623 1624 if (nt_insert(self, n, rev) == -1)
1624 1625 return -3;
1625 1626 break;
1626 1627 }
1627 1628 }
1628 1629 } else {
1629 1630 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1630 1631 const char *n = index_node(self, rev);
1631 1632 if (n == NULL) {
1632 1633 self->ntrev = rev + 1;
1633 1634 return -2;
1634 1635 }
1635 1636 if (nt_insert(self, n, rev) == -1) {
1636 1637 self->ntrev = rev + 1;
1637 1638 return -3;
1638 1639 }
1639 1640 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1640 1641 break;
1641 1642 }
1642 1643 }
1643 1644 self->ntrev = rev;
1644 1645 }
1645 1646
1646 1647 if (rev >= 0)
1647 1648 return rev;
1648 1649 return -2;
1649 1650 }
1650 1651
1651 1652 static void raise_revlog_error(void)
1652 1653 {
1653 1654 PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
1654 1655
1655 1656 mod = PyImport_ImportModule("mercurial.error");
1656 1657 if (mod == NULL) {
1657 1658 goto cleanup;
1658 1659 }
1659 1660
1660 1661 dict = PyModule_GetDict(mod);
1661 1662 if (dict == NULL) {
1662 1663 goto cleanup;
1663 1664 }
1664 1665 Py_INCREF(dict);
1665 1666
1666 1667 errclass = PyDict_GetItemString(dict, "RevlogError");
1667 1668 if (errclass == NULL) {
1668 1669 PyErr_SetString(PyExc_SystemError,
1669 1670 "could not find RevlogError");
1670 1671 goto cleanup;
1671 1672 }
1672 1673
1673 1674 /* value of exception is ignored by callers */
1674 1675 PyErr_SetString(errclass, "RevlogError");
1675 1676
1676 1677 cleanup:
1677 1678 Py_XDECREF(dict);
1678 1679 Py_XDECREF(mod);
1679 1680 }
1680 1681
1681 1682 static PyObject *index_getitem(indexObject *self, PyObject *value)
1682 1683 {
1683 1684 char *node;
1684 1685 Py_ssize_t nodelen;
1685 1686 int rev;
1686 1687
1687 1688 if (PyInt_Check(value))
1688 1689 return index_get(self, PyInt_AS_LONG(value));
1689 1690
1690 1691 if (node_check(value, &node, &nodelen) == -1)
1691 1692 return NULL;
1692 1693 rev = index_find_node(self, node, nodelen);
1693 1694 if (rev >= -1)
1694 1695 return PyInt_FromLong(rev);
1695 1696 if (rev == -2)
1696 1697 raise_revlog_error();
1697 1698 return NULL;
1698 1699 }
1699 1700
1700 1701 static int nt_partialmatch(indexObject *self, const char *node,
1701 1702 Py_ssize_t nodelen)
1702 1703 {
1703 1704 int rev;
1704 1705
1705 1706 if (nt_init(self) == -1)
1706 1707 return -3;
1707 1708
1708 1709 if (self->ntrev > 0) {
1709 1710 /* ensure that the radix tree is fully populated */
1710 1711 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1711 1712 const char *n = index_node(self, rev);
1712 1713 if (n == NULL)
1713 1714 return -2;
1714 1715 if (nt_insert(self, n, rev) == -1)
1715 1716 return -3;
1716 1717 }
1717 1718 self->ntrev = rev;
1718 1719 }
1719 1720
1720 1721 return nt_find(self, node, nodelen, 1);
1721 1722 }
1722 1723
1723 1724 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1724 1725 {
1725 1726 const char *fullnode;
1726 1727 int nodelen;
1727 1728 char *node;
1728 1729 int rev, i;
1729 1730
1730 1731 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1731 1732 return NULL;
1732 1733
1733 1734 if (nodelen < 4) {
1734 1735 PyErr_SetString(PyExc_ValueError, "key too short");
1735 1736 return NULL;
1736 1737 }
1737 1738
1738 1739 if (nodelen > 40) {
1739 1740 PyErr_SetString(PyExc_ValueError, "key too long");
1740 1741 return NULL;
1741 1742 }
1742 1743
1743 1744 for (i = 0; i < nodelen; i++)
1744 1745 hexdigit(node, i);
1745 1746 if (PyErr_Occurred()) {
1746 1747 /* input contains non-hex characters */
1747 1748 PyErr_Clear();
1748 1749 Py_RETURN_NONE;
1749 1750 }
1750 1751
1751 1752 rev = nt_partialmatch(self, node, nodelen);
1752 1753
1753 1754 switch (rev) {
1754 1755 case -4:
1755 1756 raise_revlog_error();
1756 1757 case -3:
1757 1758 return NULL;
1758 1759 case -2:
1759 1760 Py_RETURN_NONE;
1760 1761 case -1:
1761 1762 return PyString_FromStringAndSize(nullid, 20);
1762 1763 }
1763 1764
1764 1765 fullnode = index_node(self, rev);
1765 1766 if (fullnode == NULL) {
1766 1767 PyErr_Format(PyExc_IndexError,
1767 1768 "could not access rev %d", rev);
1768 1769 return NULL;
1769 1770 }
1770 1771 return PyString_FromStringAndSize(fullnode, 20);
1771 1772 }
1772 1773
1773 1774 static PyObject *index_m_get(indexObject *self, PyObject *args)
1774 1775 {
1775 1776 Py_ssize_t nodelen;
1776 1777 PyObject *val;
1777 1778 char *node;
1778 1779 int rev;
1779 1780
1780 1781 if (!PyArg_ParseTuple(args, "O", &val))
1781 1782 return NULL;
1782 1783 if (node_check(val, &node, &nodelen) == -1)
1783 1784 return NULL;
1784 1785 rev = index_find_node(self, node, nodelen);
1785 1786 if (rev == -3)
1786 1787 return NULL;
1787 1788 if (rev == -2)
1788 1789 Py_RETURN_NONE;
1789 1790 return PyInt_FromLong(rev);
1790 1791 }
1791 1792
1792 1793 static int index_contains(indexObject *self, PyObject *value)
1793 1794 {
1794 1795 char *node;
1795 1796 Py_ssize_t nodelen;
1796 1797
1797 1798 if (PyInt_Check(value)) {
1798 1799 long rev = PyInt_AS_LONG(value);
1799 1800 return rev >= -1 && rev < index_length(self);
1800 1801 }
1801 1802
1802 1803 if (node_check(value, &node, &nodelen) == -1)
1803 1804 return -1;
1804 1805
1805 1806 switch (index_find_node(self, node, nodelen)) {
1806 1807 case -3:
1807 1808 return -1;
1808 1809 case -2:
1809 1810 return 0;
1810 1811 default:
1811 1812 return 1;
1812 1813 }
1813 1814 }
1814 1815
1815 1816 typedef uint64_t bitmask;
1816 1817
1817 1818 /*
1818 1819 * Given a disjoint set of revs, return all candidates for the
1819 1820 * greatest common ancestor. In revset notation, this is the set
1820 1821 * "heads(::a and ::b and ...)"
1821 1822 */
1822 1823 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1823 1824 int revcount)
1824 1825 {
1825 1826 const bitmask allseen = (1ull << revcount) - 1;
1826 1827 const bitmask poison = 1ull << revcount;
1827 1828 PyObject *gca = PyList_New(0);
1828 1829 int i, v, interesting;
1829 1830 int maxrev = -1;
1830 1831 bitmask sp;
1831 1832 bitmask *seen;
1832 1833
1833 1834 if (gca == NULL)
1834 1835 return PyErr_NoMemory();
1835 1836
1836 1837 for (i = 0; i < revcount; i++) {
1837 1838 if (revs[i] > maxrev)
1838 1839 maxrev = revs[i];
1839 1840 }
1840 1841
1841 1842 seen = calloc(sizeof(*seen), maxrev + 1);
1842 1843 if (seen == NULL) {
1843 1844 Py_DECREF(gca);
1844 1845 return PyErr_NoMemory();
1845 1846 }
1846 1847
1847 1848 for (i = 0; i < revcount; i++)
1848 1849 seen[revs[i]] = 1ull << i;
1849 1850
1850 1851 interesting = revcount;
1851 1852
1852 1853 for (v = maxrev; v >= 0 && interesting; v--) {
1853 1854 bitmask sv = seen[v];
1854 1855 int parents[2];
1855 1856
1856 1857 if (!sv)
1857 1858 continue;
1858 1859
1859 1860 if (sv < poison) {
1860 1861 interesting -= 1;
1861 1862 if (sv == allseen) {
1862 1863 PyObject *obj = PyInt_FromLong(v);
1863 1864 if (obj == NULL)
1864 1865 goto bail;
1865 1866 if (PyList_Append(gca, obj) == -1) {
1866 1867 Py_DECREF(obj);
1867 1868 goto bail;
1868 1869 }
1869 1870 sv |= poison;
1870 1871 for (i = 0; i < revcount; i++) {
1871 1872 if (revs[i] == v)
1872 1873 goto done;
1873 1874 }
1874 1875 }
1875 1876 }
1876 1877 if (index_get_parents(self, v, parents, maxrev) < 0)
1877 1878 goto bail;
1878 1879
1879 1880 for (i = 0; i < 2; i++) {
1880 1881 int p = parents[i];
1881 1882 if (p == -1)
1882 1883 continue;
1883 1884 sp = seen[p];
1884 1885 if (sv < poison) {
1885 1886 if (sp == 0) {
1886 1887 seen[p] = sv;
1887 1888 interesting++;
1888 1889 }
1889 1890 else if (sp != sv)
1890 1891 seen[p] |= sv;
1891 1892 } else {
1892 1893 if (sp && sp < poison)
1893 1894 interesting--;
1894 1895 seen[p] = sv;
1895 1896 }
1896 1897 }
1897 1898 }
1898 1899
1899 1900 done:
1900 1901 free(seen);
1901 1902 return gca;
1902 1903 bail:
1903 1904 free(seen);
1904 1905 Py_XDECREF(gca);
1905 1906 return NULL;
1906 1907 }
1907 1908
1908 1909 /*
1909 1910 * Given a disjoint set of revs, return the subset with the longest
1910 1911 * path to the root.
1911 1912 */
1912 1913 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1913 1914 {
1914 1915 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1915 1916 static const Py_ssize_t capacity = 24;
1916 1917 int *depth, *interesting = NULL;
1917 1918 int i, j, v, ninteresting;
1918 1919 PyObject *dict = NULL, *keys = NULL;
1919 1920 long *seen = NULL;
1920 1921 int maxrev = -1;
1921 1922 long final;
1922 1923
1923 1924 if (revcount > capacity) {
1924 1925 PyErr_Format(PyExc_OverflowError,
1925 1926 "bitset size (%ld) > capacity (%ld)",
1926 1927 (long)revcount, (long)capacity);
1927 1928 return NULL;
1928 1929 }
1929 1930
1930 1931 for (i = 0; i < revcount; i++) {
1931 1932 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1932 1933 if (n > maxrev)
1933 1934 maxrev = n;
1934 1935 }
1935 1936
1936 1937 depth = calloc(sizeof(*depth), maxrev + 1);
1937 1938 if (depth == NULL)
1938 1939 return PyErr_NoMemory();
1939 1940
1940 1941 seen = calloc(sizeof(*seen), maxrev + 1);
1941 1942 if (seen == NULL) {
1942 1943 PyErr_NoMemory();
1943 1944 goto bail;
1944 1945 }
1945 1946
1946 1947 interesting = calloc(sizeof(*interesting), 2 << revcount);
1947 1948 if (interesting == NULL) {
1948 1949 PyErr_NoMemory();
1949 1950 goto bail;
1950 1951 }
1951 1952
1952 1953 if (PyList_Sort(revs) == -1)
1953 1954 goto bail;
1954 1955
1955 1956 for (i = 0; i < revcount; i++) {
1956 1957 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1957 1958 long b = 1l << i;
1958 1959 depth[n] = 1;
1959 1960 seen[n] = b;
1960 1961 interesting[b] = 1;
1961 1962 }
1962 1963
1963 1964 ninteresting = (int)revcount;
1964 1965
1965 1966 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
1966 1967 int dv = depth[v];
1967 1968 int parents[2];
1968 1969 long sv;
1969 1970
1970 1971 if (dv == 0)
1971 1972 continue;
1972 1973
1973 1974 sv = seen[v];
1974 1975 if (index_get_parents(self, v, parents, maxrev) < 0)
1975 1976 goto bail;
1976 1977
1977 1978 for (i = 0; i < 2; i++) {
1978 1979 int p = parents[i];
1979 1980 long nsp, sp;
1980 1981 int dp;
1981 1982
1982 1983 if (p == -1)
1983 1984 continue;
1984 1985
1985 1986 dp = depth[p];
1986 1987 nsp = sp = seen[p];
1987 1988 if (dp <= dv) {
1988 1989 depth[p] = dv + 1;
1989 1990 if (sp != sv) {
1990 1991 interesting[sv] += 1;
1991 1992 nsp = seen[p] = sv;
1992 1993 if (sp) {
1993 1994 interesting[sp] -= 1;
1994 1995 if (interesting[sp] == 0)
1995 1996 ninteresting -= 1;
1996 1997 }
1997 1998 }
1998 1999 }
1999 2000 else if (dv == dp - 1) {
2000 2001 nsp = sp | sv;
2001 2002 if (nsp == sp)
2002 2003 continue;
2003 2004 seen[p] = nsp;
2004 2005 interesting[sp] -= 1;
2005 2006 if (interesting[sp] == 0 && interesting[nsp] > 0)
2006 2007 ninteresting -= 1;
2007 2008 interesting[nsp] += 1;
2008 2009 }
2009 2010 }
2010 2011 interesting[sv] -= 1;
2011 2012 if (interesting[sv] == 0)
2012 2013 ninteresting -= 1;
2013 2014 }
2014 2015
2015 2016 final = 0;
2016 2017 j = ninteresting;
2017 2018 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
2018 2019 if (interesting[i] == 0)
2019 2020 continue;
2020 2021 final |= i;
2021 2022 j -= 1;
2022 2023 }
2023 2024 if (final == 0) {
2024 2025 keys = PyList_New(0);
2025 2026 goto bail;
2026 2027 }
2027 2028
2028 2029 dict = PyDict_New();
2029 2030 if (dict == NULL)
2030 2031 goto bail;
2031 2032
2032 2033 for (i = 0; i < revcount; i++) {
2033 2034 PyObject *key;
2034 2035
2035 2036 if ((final & (1 << i)) == 0)
2036 2037 continue;
2037 2038
2038 2039 key = PyList_GET_ITEM(revs, i);
2039 2040 Py_INCREF(key);
2040 2041 Py_INCREF(Py_None);
2041 2042 if (PyDict_SetItem(dict, key, Py_None) == -1) {
2042 2043 Py_DECREF(key);
2043 2044 Py_DECREF(Py_None);
2044 2045 goto bail;
2045 2046 }
2046 2047 }
2047 2048
2048 2049 keys = PyDict_Keys(dict);
2049 2050
2050 2051 bail:
2051 2052 free(depth);
2052 2053 free(seen);
2053 2054 free(interesting);
2054 2055 Py_XDECREF(dict);
2055 2056
2056 2057 return keys;
2057 2058 }
2058 2059
2059 2060 /*
2060 2061 * Given a (possibly overlapping) set of revs, return all the
2061 2062 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
2062 2063 */
2063 2064 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
2064 2065 {
2065 2066 PyObject *ret = NULL;
2066 2067 Py_ssize_t argcount, i, len;
2067 2068 bitmask repeat = 0;
2068 2069 int revcount = 0;
2069 2070 int *revs;
2070 2071
2071 2072 argcount = PySequence_Length(args);
2072 2073 revs = malloc(argcount * sizeof(*revs));
2073 2074 if (argcount > 0 && revs == NULL)
2074 2075 return PyErr_NoMemory();
2075 2076 len = index_length(self) - 1;
2076 2077
2077 2078 for (i = 0; i < argcount; i++) {
2078 2079 static const int capacity = 24;
2079 2080 PyObject *obj = PySequence_GetItem(args, i);
2080 2081 bitmask x;
2081 2082 long val;
2082 2083
2083 2084 if (!PyInt_Check(obj)) {
2084 2085 PyErr_SetString(PyExc_TypeError,
2085 2086 "arguments must all be ints");
2086 2087 Py_DECREF(obj);
2087 2088 goto bail;
2088 2089 }
2089 2090 val = PyInt_AsLong(obj);
2090 2091 Py_DECREF(obj);
2091 2092 if (val == -1) {
2092 2093 ret = PyList_New(0);
2093 2094 goto done;
2094 2095 }
2095 2096 if (val < 0 || val >= len) {
2096 2097 PyErr_SetString(PyExc_IndexError,
2097 2098 "index out of range");
2098 2099 goto bail;
2099 2100 }
2100 2101 /* this cheesy bloom filter lets us avoid some more
2101 2102 * expensive duplicate checks in the common set-is-disjoint
2102 2103 * case */
2103 2104 x = 1ull << (val & 0x3f);
2104 2105 if (repeat & x) {
2105 2106 int k;
2106 2107 for (k = 0; k < revcount; k++) {
2107 2108 if (val == revs[k])
2108 2109 goto duplicate;
2109 2110 }
2110 2111 }
2111 2112 else repeat |= x;
2112 2113 if (revcount >= capacity) {
2113 2114 PyErr_Format(PyExc_OverflowError,
2114 2115 "bitset size (%d) > capacity (%d)",
2115 2116 revcount, capacity);
2116 2117 goto bail;
2117 2118 }
2118 2119 revs[revcount++] = (int)val;
2119 2120 duplicate:;
2120 2121 }
2121 2122
2122 2123 if (revcount == 0) {
2123 2124 ret = PyList_New(0);
2124 2125 goto done;
2125 2126 }
2126 2127 if (revcount == 1) {
2127 2128 PyObject *obj;
2128 2129 ret = PyList_New(1);
2129 2130 if (ret == NULL)
2130 2131 goto bail;
2131 2132 obj = PyInt_FromLong(revs[0]);
2132 2133 if (obj == NULL)
2133 2134 goto bail;
2134 2135 PyList_SET_ITEM(ret, 0, obj);
2135 2136 goto done;
2136 2137 }
2137 2138
2138 2139 ret = find_gca_candidates(self, revs, revcount);
2139 2140 if (ret == NULL)
2140 2141 goto bail;
2141 2142
2142 2143 done:
2143 2144 free(revs);
2144 2145 return ret;
2145 2146
2146 2147 bail:
2147 2148 free(revs);
2148 2149 Py_XDECREF(ret);
2149 2150 return NULL;
2150 2151 }
2151 2152
2152 2153 /*
2153 2154 * Given a (possibly overlapping) set of revs, return the greatest
2154 2155 * common ancestors: those with the longest path to the root.
2155 2156 */
2156 2157 static PyObject *index_ancestors(indexObject *self, PyObject *args)
2157 2158 {
2158 2159 PyObject *gca = index_commonancestorsheads(self, args);
2159 2160 if (gca == NULL)
2160 2161 return NULL;
2161 2162
2162 2163 if (PyList_GET_SIZE(gca) <= 1) {
2163 2164 Py_INCREF(gca);
2164 2165 return gca;
2165 2166 }
2166 2167
2167 2168 return find_deepest(self, gca);
2168 2169 }
2169 2170
2170 2171 /*
2171 2172 * Invalidate any trie entries introduced by added revs.
2172 2173 */
2173 2174 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
2174 2175 {
2175 2176 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
2176 2177
2177 2178 for (i = start; i < len; i++) {
2178 2179 PyObject *tuple = PyList_GET_ITEM(self->added, i);
2179 2180 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
2180 2181
2181 2182 nt_insert(self, PyString_AS_STRING(node), -1);
2182 2183 }
2183 2184
2184 2185 if (start == 0)
2185 2186 Py_CLEAR(self->added);
2186 2187 }
2187 2188
2188 2189 /*
2189 2190 * Delete a numeric range of revs, which must be at the end of the
2190 2191 * range, but exclude the sentinel nullid entry.
2191 2192 */
2192 2193 static int index_slice_del(indexObject *self, PyObject *item)
2193 2194 {
2194 2195 Py_ssize_t start, stop, step, slicelength;
2195 2196 Py_ssize_t length = index_length(self);
2196 2197 int ret = 0;
2197 2198
2198 2199 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
2199 2200 &start, &stop, &step, &slicelength) < 0)
2200 2201 return -1;
2201 2202
2202 2203 if (slicelength <= 0)
2203 2204 return 0;
2204 2205
2205 2206 if ((step < 0 && start < stop) || (step > 0 && start > stop))
2206 2207 stop = start;
2207 2208
2208 2209 if (step < 0) {
2209 2210 stop = start + 1;
2210 2211 start = stop + step*(slicelength - 1) - 1;
2211 2212 step = -step;
2212 2213 }
2213 2214
2214 2215 if (step != 1) {
2215 2216 PyErr_SetString(PyExc_ValueError,
2216 2217 "revlog index delete requires step size of 1");
2217 2218 return -1;
2218 2219 }
2219 2220
2220 2221 if (stop != length - 1) {
2221 2222 PyErr_SetString(PyExc_IndexError,
2222 2223 "revlog index deletion indices are invalid");
2223 2224 return -1;
2224 2225 }
2225 2226
2226 2227 if (start < self->length - 1) {
2227 2228 if (self->nt) {
2228 2229 Py_ssize_t i;
2229 2230
2230 2231 for (i = start + 1; i < self->length - 1; i++) {
2231 2232 const char *node = index_node(self, i);
2232 2233
2233 2234 if (node)
2234 2235 nt_insert(self, node, -1);
2235 2236 }
2236 2237 if (self->added)
2237 2238 nt_invalidate_added(self, 0);
2238 2239 if (self->ntrev > start)
2239 2240 self->ntrev = (int)start;
2240 2241 }
2241 2242 self->length = start + 1;
2242 2243 if (start < self->raw_length) {
2243 2244 if (self->cache) {
2244 2245 Py_ssize_t i;
2245 2246 for (i = start; i < self->raw_length; i++)
2246 2247 Py_CLEAR(self->cache[i]);
2247 2248 }
2248 2249 self->raw_length = start;
2249 2250 }
2250 2251 goto done;
2251 2252 }
2252 2253
2253 2254 if (self->nt) {
2254 2255 nt_invalidate_added(self, start - self->length + 1);
2255 2256 if (self->ntrev > start)
2256 2257 self->ntrev = (int)start;
2257 2258 }
2258 2259 if (self->added)
2259 2260 ret = PyList_SetSlice(self->added, start - self->length + 1,
2260 2261 PyList_GET_SIZE(self->added), NULL);
2261 2262 done:
2262 2263 Py_CLEAR(self->headrevs);
2263 2264 return ret;
2264 2265 }
2265 2266
2266 2267 /*
2267 2268 * Supported ops:
2268 2269 *
2269 2270 * slice deletion
2270 2271 * string assignment (extend node->rev mapping)
2271 2272 * string deletion (shrink node->rev mapping)
2272 2273 */
2273 2274 static int index_assign_subscript(indexObject *self, PyObject *item,
2274 2275 PyObject *value)
2275 2276 {
2276 2277 char *node;
2277 2278 Py_ssize_t nodelen;
2278 2279 long rev;
2279 2280
2280 2281 if (PySlice_Check(item) && value == NULL)
2281 2282 return index_slice_del(self, item);
2282 2283
2283 2284 if (node_check(item, &node, &nodelen) == -1)
2284 2285 return -1;
2285 2286
2286 2287 if (value == NULL)
2287 2288 return self->nt ? nt_insert(self, node, -1) : 0;
2288 2289 rev = PyInt_AsLong(value);
2289 2290 if (rev > INT_MAX || rev < 0) {
2290 2291 if (!PyErr_Occurred())
2291 2292 PyErr_SetString(PyExc_ValueError, "rev out of range");
2292 2293 return -1;
2293 2294 }
2294 2295
2295 2296 if (nt_init(self) == -1)
2296 2297 return -1;
2297 2298 return nt_insert(self, node, (int)rev);
2298 2299 }
2299 2300
2300 2301 /*
2301 2302 * Find all RevlogNG entries in an index that has inline data. Update
2302 2303 * the optional "offsets" table with those entries.
2303 2304 */
2304 2305 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
2305 2306 {
2306 2307 const char *data = PyString_AS_STRING(self->data);
2307 2308 Py_ssize_t pos = 0;
2308 2309 Py_ssize_t end = PyString_GET_SIZE(self->data);
2309 2310 long incr = v1_hdrsize;
2310 2311 Py_ssize_t len = 0;
2311 2312
2312 2313 while (pos + v1_hdrsize <= end && pos >= 0) {
2313 2314 uint32_t comp_len;
2314 2315 /* 3rd element of header is length of compressed inline data */
2315 2316 comp_len = getbe32(data + pos + 8);
2316 2317 incr = v1_hdrsize + comp_len;
2317 2318 if (offsets)
2318 2319 offsets[len] = data + pos;
2319 2320 len++;
2320 2321 pos += incr;
2321 2322 }
2322 2323
2323 2324 if (pos != end) {
2324 2325 if (!PyErr_Occurred())
2325 2326 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2326 2327 return -1;
2327 2328 }
2328 2329
2329 2330 return len;
2330 2331 }
2331 2332
2332 2333 static int index_init(indexObject *self, PyObject *args)
2333 2334 {
2334 2335 PyObject *data_obj, *inlined_obj;
2335 2336 Py_ssize_t size;
2336 2337
2337 2338 /* Initialize before argument-checking to avoid index_dealloc() crash. */
2338 2339 self->raw_length = 0;
2339 2340 self->added = NULL;
2340 2341 self->cache = NULL;
2341 2342 self->data = NULL;
2342 2343 self->headrevs = NULL;
2343 2344 self->filteredrevs = Py_None;
2344 2345 Py_INCREF(Py_None);
2345 2346 self->nt = NULL;
2346 2347 self->offsets = NULL;
2347 2348
2348 2349 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
2349 2350 return -1;
2350 2351 if (!PyString_Check(data_obj)) {
2351 2352 PyErr_SetString(PyExc_TypeError, "data is not a string");
2352 2353 return -1;
2353 2354 }
2354 2355 size = PyString_GET_SIZE(data_obj);
2355 2356
2356 2357 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
2357 2358 self->data = data_obj;
2358 2359
2359 2360 self->ntlength = self->ntcapacity = 0;
2360 2361 self->ntdepth = self->ntsplits = 0;
2361 2362 self->ntlookups = self->ntmisses = 0;
2362 2363 self->ntrev = -1;
2363 2364 Py_INCREF(self->data);
2364 2365
2365 2366 if (self->inlined) {
2366 2367 Py_ssize_t len = inline_scan(self, NULL);
2367 2368 if (len == -1)
2368 2369 goto bail;
2369 2370 self->raw_length = len;
2370 2371 self->length = len + 1;
2371 2372 } else {
2372 2373 if (size % v1_hdrsize) {
2373 2374 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2374 2375 goto bail;
2375 2376 }
2376 2377 self->raw_length = size / v1_hdrsize;
2377 2378 self->length = self->raw_length + 1;
2378 2379 }
2379 2380
2380 2381 return 0;
2381 2382 bail:
2382 2383 return -1;
2383 2384 }
2384 2385
2385 2386 static PyObject *index_nodemap(indexObject *self)
2386 2387 {
2387 2388 Py_INCREF(self);
2388 2389 return (PyObject *)self;
2389 2390 }
2390 2391
2391 2392 static void index_dealloc(indexObject *self)
2392 2393 {
2393 2394 _index_clearcaches(self);
2394 2395 Py_XDECREF(self->filteredrevs);
2395 2396 Py_XDECREF(self->data);
2396 2397 Py_XDECREF(self->added);
2397 2398 PyObject_Del(self);
2398 2399 }
2399 2400
2400 2401 static PySequenceMethods index_sequence_methods = {
2401 2402 (lenfunc)index_length, /* sq_length */
2402 2403 0, /* sq_concat */
2403 2404 0, /* sq_repeat */
2404 2405 (ssizeargfunc)index_get, /* sq_item */
2405 2406 0, /* sq_slice */
2406 2407 0, /* sq_ass_item */
2407 2408 0, /* sq_ass_slice */
2408 2409 (objobjproc)index_contains, /* sq_contains */
2409 2410 };
2410 2411
2411 2412 static PyMappingMethods index_mapping_methods = {
2412 2413 (lenfunc)index_length, /* mp_length */
2413 2414 (binaryfunc)index_getitem, /* mp_subscript */
2414 2415 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
2415 2416 };
2416 2417
2417 2418 static PyMethodDef index_methods[] = {
2418 2419 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
2419 2420 "return the gca set of the given revs"},
2420 2421 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
2421 2422 METH_VARARGS,
2422 2423 "return the heads of the common ancestors of the given revs"},
2423 2424 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
2424 2425 "clear the index caches"},
2425 2426 {"get", (PyCFunction)index_m_get, METH_VARARGS,
2426 2427 "get an index entry"},
2427 2428 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
2428 2429 METH_VARARGS, "compute phases"},
2429 2430 {"reachableroots", (PyCFunction)reachableroots, METH_VARARGS,
2430 2431 "reachableroots"},
2431 2432 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
2432 2433 "get head revisions"}, /* Can do filtering since 3.2 */
2433 2434 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
2434 2435 "get filtered head revisions"}, /* Can always do filtering */
2435 2436 {"insert", (PyCFunction)index_insert, METH_VARARGS,
2436 2437 "insert an index entry"},
2437 2438 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
2438 2439 "match a potentially ambiguous node ID"},
2439 2440 {"stats", (PyCFunction)index_stats, METH_NOARGS,
2440 2441 "stats for the index"},
2441 2442 {NULL} /* Sentinel */
2442 2443 };
2443 2444
2444 2445 static PyGetSetDef index_getset[] = {
2445 2446 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
2446 2447 {NULL} /* Sentinel */
2447 2448 };
2448 2449
2449 2450 static PyTypeObject indexType = {
2450 2451 PyObject_HEAD_INIT(NULL)
2451 2452 0, /* ob_size */
2452 2453 "parsers.index", /* tp_name */
2453 2454 sizeof(indexObject), /* tp_basicsize */
2454 2455 0, /* tp_itemsize */
2455 2456 (destructor)index_dealloc, /* tp_dealloc */
2456 2457 0, /* tp_print */
2457 2458 0, /* tp_getattr */
2458 2459 0, /* tp_setattr */
2459 2460 0, /* tp_compare */
2460 2461 0, /* tp_repr */
2461 2462 0, /* tp_as_number */
2462 2463 &index_sequence_methods, /* tp_as_sequence */
2463 2464 &index_mapping_methods, /* tp_as_mapping */
2464 2465 0, /* tp_hash */
2465 2466 0, /* tp_call */
2466 2467 0, /* tp_str */
2467 2468 0, /* tp_getattro */
2468 2469 0, /* tp_setattro */
2469 2470 0, /* tp_as_buffer */
2470 2471 Py_TPFLAGS_DEFAULT, /* tp_flags */
2471 2472 "revlog index", /* tp_doc */
2472 2473 0, /* tp_traverse */
2473 2474 0, /* tp_clear */
2474 2475 0, /* tp_richcompare */
2475 2476 0, /* tp_weaklistoffset */
2476 2477 0, /* tp_iter */
2477 2478 0, /* tp_iternext */
2478 2479 index_methods, /* tp_methods */
2479 2480 0, /* tp_members */
2480 2481 index_getset, /* tp_getset */
2481 2482 0, /* tp_base */
2482 2483 0, /* tp_dict */
2483 2484 0, /* tp_descr_get */
2484 2485 0, /* tp_descr_set */
2485 2486 0, /* tp_dictoffset */
2486 2487 (initproc)index_init, /* tp_init */
2487 2488 0, /* tp_alloc */
2488 2489 };
2489 2490
2490 2491 /*
2491 2492 * returns a tuple of the form (index, index, cache) with elements as
2492 2493 * follows:
2493 2494 *
2494 2495 * index: an index object that lazily parses RevlogNG records
2495 2496 * cache: if data is inlined, a tuple (index_file_content, 0), else None
2496 2497 *
2497 2498 * added complications are for backwards compatibility
2498 2499 */
2499 2500 static PyObject *parse_index2(PyObject *self, PyObject *args)
2500 2501 {
2501 2502 PyObject *tuple = NULL, *cache = NULL;
2502 2503 indexObject *idx;
2503 2504 int ret;
2504 2505
2505 2506 idx = PyObject_New(indexObject, &indexType);
2506 2507 if (idx == NULL)
2507 2508 goto bail;
2508 2509
2509 2510 ret = index_init(idx, args);
2510 2511 if (ret == -1)
2511 2512 goto bail;
2512 2513
2513 2514 if (idx->inlined) {
2514 2515 cache = Py_BuildValue("iO", 0, idx->data);
2515 2516 if (cache == NULL)
2516 2517 goto bail;
2517 2518 } else {
2518 2519 cache = Py_None;
2519 2520 Py_INCREF(cache);
2520 2521 }
2521 2522
2522 2523 tuple = Py_BuildValue("NN", idx, cache);
2523 2524 if (!tuple)
2524 2525 goto bail;
2525 2526 return tuple;
2526 2527
2527 2528 bail:
2528 2529 Py_XDECREF(idx);
2529 2530 Py_XDECREF(cache);
2530 2531 Py_XDECREF(tuple);
2531 2532 return NULL;
2532 2533 }
2533 2534
2534 2535 #define BUMPED_FIX 1
2535 2536 #define USING_SHA_256 2
2536 2537
2537 2538 static PyObject *readshas(
2538 2539 const char *source, unsigned char num, Py_ssize_t hashwidth)
2539 2540 {
2540 2541 int i;
2541 2542 PyObject *list = PyTuple_New(num);
2542 2543 if (list == NULL) {
2543 2544 return NULL;
2544 2545 }
2545 2546 for (i = 0; i < num; i++) {
2546 2547 PyObject *hash = PyString_FromStringAndSize(source, hashwidth);
2547 2548 if (hash == NULL) {
2548 2549 Py_DECREF(list);
2549 2550 return NULL;
2550 2551 }
2551 2552 PyTuple_SetItem(list, i, hash);
2552 2553 source += hashwidth;
2553 2554 }
2554 2555 return list;
2555 2556 }
2556 2557
2557 2558 static PyObject *fm1readmarker(const char *data, uint32_t *msize)
2558 2559 {
2559 2560 const char *meta;
2560 2561
2561 2562 double mtime;
2562 2563 int16_t tz;
2563 2564 uint16_t flags;
2564 2565 unsigned char nsuccs, nparents, nmetadata;
2565 2566 Py_ssize_t hashwidth = 20;
2566 2567
2567 2568 PyObject *prec = NULL, *parents = NULL, *succs = NULL;
2568 2569 PyObject *metadata = NULL, *ret = NULL;
2569 2570 int i;
2570 2571
2571 2572 *msize = getbe32(data);
2572 2573 data += 4;
2573 2574 mtime = getbefloat64(data);
2574 2575 data += 8;
2575 2576 tz = getbeint16(data);
2576 2577 data += 2;
2577 2578 flags = getbeuint16(data);
2578 2579 data += 2;
2579 2580
2580 2581 if (flags & USING_SHA_256) {
2581 2582 hashwidth = 32;
2582 2583 }
2583 2584
2584 2585 nsuccs = (unsigned char)(*data++);
2585 2586 nparents = (unsigned char)(*data++);
2586 2587 nmetadata = (unsigned char)(*data++);
2587 2588
2588 2589 prec = PyString_FromStringAndSize(data, hashwidth);
2589 2590 data += hashwidth;
2590 2591 if (prec == NULL) {
2591 2592 goto bail;
2592 2593 }
2593 2594
2594 2595 succs = readshas(data, nsuccs, hashwidth);
2595 2596 if (succs == NULL) {
2596 2597 goto bail;
2597 2598 }
2598 2599 data += nsuccs * hashwidth;
2599 2600
2600 2601 if (nparents == 1 || nparents == 2) {
2601 2602 parents = readshas(data, nparents, hashwidth);
2602 2603 if (parents == NULL) {
2603 2604 goto bail;
2604 2605 }
2605 2606 data += nparents * hashwidth;
2606 2607 } else {
2607 2608 parents = Py_None;
2608 2609 }
2609 2610
2610 2611 meta = data + (2 * nmetadata);
2611 2612 metadata = PyTuple_New(nmetadata);
2612 2613 if (metadata == NULL) {
2613 2614 goto bail;
2614 2615 }
2615 2616 for (i = 0; i < nmetadata; i++) {
2616 2617 PyObject *tmp, *left = NULL, *right = NULL;
2617 2618 Py_ssize_t metasize = (unsigned char)(*data++);
2618 2619 left = PyString_FromStringAndSize(meta, metasize);
2619 2620 meta += metasize;
2620 2621 metasize = (unsigned char)(*data++);
2621 2622 right = PyString_FromStringAndSize(meta, metasize);
2622 2623 meta += metasize;
2623 2624 if (!left || !right) {
2624 2625 Py_XDECREF(left);
2625 2626 Py_XDECREF(right);
2626 2627 goto bail;
2627 2628 }
2628 2629 tmp = PyTuple_Pack(2, left, right);
2629 2630 Py_DECREF(left);
2630 2631 Py_DECREF(right);
2631 2632 if (!tmp) {
2632 2633 goto bail;
2633 2634 }
2634 2635 PyTuple_SetItem(metadata, i, tmp);
2635 2636 }
2636 2637 ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags,
2637 2638 metadata, mtime, (int)tz * 60, parents);
2638 2639 bail:
2639 2640 Py_XDECREF(prec);
2640 2641 Py_XDECREF(succs);
2641 2642 Py_XDECREF(metadata);
2642 2643 if (parents != Py_None)
2643 2644 Py_XDECREF(parents);
2644 2645 return ret;
2645 2646 }
2646 2647
2647 2648
2648 2649 static PyObject *fm1readmarkers(PyObject *self, PyObject *args) {
2649 2650 const char *data;
2650 2651 Py_ssize_t datalen;
2651 2652 /* only unsigned long because python 2.4, should be Py_ssize_t */
2652 2653 unsigned long offset, stop;
2653 2654 PyObject *markers = NULL;
2654 2655
2655 2656 /* replace kk with nn when we drop Python 2.4 */
2656 2657 if (!PyArg_ParseTuple(args, "s#kk", &data, &datalen, &offset, &stop)) {
2657 2658 return NULL;
2658 2659 }
2659 2660 data += offset;
2660 2661 markers = PyList_New(0);
2661 2662 if (!markers) {
2662 2663 return NULL;
2663 2664 }
2664 2665 while (offset < stop) {
2665 2666 uint32_t msize;
2666 2667 int error;
2667 2668 PyObject *record = fm1readmarker(data, &msize);
2668 2669 if (!record) {
2669 2670 goto bail;
2670 2671 }
2671 2672 error = PyList_Append(markers, record);
2672 2673 Py_DECREF(record);
2673 2674 if (error) {
2674 2675 goto bail;
2675 2676 }
2676 2677 data += msize;
2677 2678 offset += msize;
2678 2679 }
2679 2680 return markers;
2680 2681 bail:
2681 2682 Py_DECREF(markers);
2682 2683 return NULL;
2683 2684 }
2684 2685
2685 2686 static char parsers_doc[] = "Efficient content parsing.";
2686 2687
2687 2688 PyObject *encodedir(PyObject *self, PyObject *args);
2688 2689 PyObject *pathencode(PyObject *self, PyObject *args);
2689 2690 PyObject *lowerencode(PyObject *self, PyObject *args);
2690 2691
2691 2692 static PyMethodDef methods[] = {
2692 2693 {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
2693 2694 {"parse_manifest", parse_manifest, METH_VARARGS, "parse a manifest\n"},
2694 2695 {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
2695 2696 {"parse_index2", parse_index2, METH_VARARGS, "parse a revlog index\n"},
2696 2697 {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
2697 2698 {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
2698 2699 {"dict_new_presized", dict_new_presized, METH_VARARGS,
2699 2700 "construct a dict with an expected size\n"},
2700 2701 {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
2701 2702 "make file foldmap\n"},
2702 2703 {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
2703 2704 {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
2704 2705 {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
2705 2706 {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
2706 2707 "parse v1 obsolete markers\n"},
2707 2708 {NULL, NULL}
2708 2709 };
2709 2710
2710 2711 void dirs_module_init(PyObject *mod);
2711 2712 void manifest_module_init(PyObject *mod);
2712 2713
2713 2714 static void module_init(PyObject *mod)
2714 2715 {
2715 2716 /* This module constant has two purposes. First, it lets us unit test
2716 2717 * the ImportError raised without hard-coding any error text. This
2717 2718 * means we can change the text in the future without breaking tests,
2718 2719 * even across changesets without a recompile. Second, its presence
2719 2720 * can be used to determine whether the version-checking logic is
2720 2721 * present, which also helps in testing across changesets without a
2721 2722 * recompile. Note that this means the pure-Python version of parsers
2722 2723 * should not have this module constant. */
2723 2724 PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
2724 2725
2725 2726 dirs_module_init(mod);
2726 2727 manifest_module_init(mod);
2727 2728
2728 2729 indexType.tp_new = PyType_GenericNew;
2729 2730 if (PyType_Ready(&indexType) < 0 ||
2730 2731 PyType_Ready(&dirstateTupleType) < 0)
2731 2732 return;
2732 2733 Py_INCREF(&indexType);
2733 2734 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2734 2735 Py_INCREF(&dirstateTupleType);
2735 2736 PyModule_AddObject(mod, "dirstatetuple",
2736 2737 (PyObject *)&dirstateTupleType);
2737 2738
2738 2739 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2739 2740 -1, -1, -1, -1, nullid, 20);
2740 2741 if (nullentry)
2741 2742 PyObject_GC_UnTrack(nullentry);
2742 2743 }
2743 2744
2744 2745 static int check_python_version(void)
2745 2746 {
2746 2747 PyObject *sys = PyImport_ImportModule("sys"), *ver;
2747 2748 long hexversion;
2748 2749 if (!sys)
2749 2750 return -1;
2750 2751 ver = PyObject_GetAttrString(sys, "hexversion");
2751 2752 Py_DECREF(sys);
2752 2753 if (!ver)
2753 2754 return -1;
2754 2755 hexversion = PyInt_AsLong(ver);
2755 2756 Py_DECREF(ver);
2756 2757 /* sys.hexversion is a 32-bit number by default, so the -1 case
2757 2758 * should only occur in unusual circumstances (e.g. if sys.hexversion
2758 2759 * is manually set to an invalid value). */
2759 2760 if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
2760 2761 PyErr_Format(PyExc_ImportError, "%s: The Mercurial extension "
2761 2762 "modules were compiled with Python " PY_VERSION ", but "
2762 2763 "Mercurial is currently using Python with sys.hexversion=%ld: "
2763 2764 "Python %s\n at: %s", versionerrortext, hexversion,
2764 2765 Py_GetVersion(), Py_GetProgramFullPath());
2765 2766 return -1;
2766 2767 }
2767 2768 return 0;
2768 2769 }
2769 2770
2770 2771 #ifdef IS_PY3K
2771 2772 static struct PyModuleDef parsers_module = {
2772 2773 PyModuleDef_HEAD_INIT,
2773 2774 "parsers",
2774 2775 parsers_doc,
2775 2776 -1,
2776 2777 methods
2777 2778 };
2778 2779
2779 2780 PyMODINIT_FUNC PyInit_parsers(void)
2780 2781 {
2781 2782 PyObject *mod;
2782 2783
2783 2784 if (check_python_version() == -1)
2784 2785 return;
2785 2786 mod = PyModule_Create(&parsers_module);
2786 2787 module_init(mod);
2787 2788 return mod;
2788 2789 }
2789 2790 #else
2790 2791 PyMODINIT_FUNC initparsers(void)
2791 2792 {
2792 2793 PyObject *mod;
2793 2794
2794 2795 if (check_python_version() == -1)
2795 2796 return;
2796 2797 mod = Py_InitModule3("parsers", methods, parsers_doc);
2797 2798 module_init(mod);
2798 2799 }
2799 2800 #endif
General Comments 0
You need to be logged in to leave comments. Login now