##// END OF EJS Templates
parsers: check results of PyInt_FromLong (issue4771)
Bryan O'Sullivan -
r27365:ec04370b default
parent child Browse files
Show More
@@ -1,2848 +1,2855
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 Py_DECREF(normed);
258 258 goto quit;
259 259 }
260 260 Py_DECREF(normed);
261 261 }
262 262 }
263 263 return file_foldmap;
264 264 quit:
265 265 Py_XDECREF(file_foldmap);
266 266 return NULL;
267 267 }
268 268
269 269 /*
270 270 * This code assumes that a manifest is stitched together with newline
271 271 * ('\n') characters.
272 272 */
273 273 static PyObject *parse_manifest(PyObject *self, PyObject *args)
274 274 {
275 275 PyObject *mfdict, *fdict;
276 276 char *str, *start, *end;
277 277 int len;
278 278
279 279 if (!PyArg_ParseTuple(args, "O!O!s#:parse_manifest",
280 280 &PyDict_Type, &mfdict,
281 281 &PyDict_Type, &fdict,
282 282 &str, &len))
283 283 goto quit;
284 284
285 285 start = str;
286 286 end = str + len;
287 287 while (start < end) {
288 288 PyObject *file = NULL, *node = NULL;
289 289 PyObject *flags = NULL;
290 290 char *zero = NULL, *newline = NULL;
291 291 ptrdiff_t nlen;
292 292
293 293 zero = memchr(start, '\0', end - start);
294 294 if (!zero) {
295 295 PyErr_SetString(PyExc_ValueError,
296 296 "manifest entry has no separator");
297 297 goto quit;
298 298 }
299 299
300 300 newline = memchr(zero + 1, '\n', end - (zero + 1));
301 301 if (!newline) {
302 302 PyErr_SetString(PyExc_ValueError,
303 303 "manifest contains trailing garbage");
304 304 goto quit;
305 305 }
306 306
307 307 file = PyBytes_FromStringAndSize(start, zero - start);
308 308
309 309 if (!file)
310 310 goto bail;
311 311
312 312 nlen = newline - zero - 1;
313 313
314 314 node = unhexlify(zero + 1, nlen > 40 ? 40 : (int)nlen);
315 315 if (!node)
316 316 goto bail;
317 317
318 318 if (nlen > 40) {
319 319 flags = PyBytes_FromStringAndSize(zero + 41,
320 320 nlen - 40);
321 321 if (!flags)
322 322 goto bail;
323 323
324 324 if (PyDict_SetItem(fdict, file, flags) == -1)
325 325 goto bail;
326 326 }
327 327
328 328 if (PyDict_SetItem(mfdict, file, node) == -1)
329 329 goto bail;
330 330
331 331 start = newline + 1;
332 332
333 333 Py_XDECREF(flags);
334 334 Py_XDECREF(node);
335 335 Py_XDECREF(file);
336 336 continue;
337 337 bail:
338 338 Py_XDECREF(flags);
339 339 Py_XDECREF(node);
340 340 Py_XDECREF(file);
341 341 goto quit;
342 342 }
343 343
344 344 Py_INCREF(Py_None);
345 345 return Py_None;
346 346 quit:
347 347 return NULL;
348 348 }
349 349
350 350 static inline dirstateTupleObject *make_dirstate_tuple(char state, int mode,
351 351 int size, int mtime)
352 352 {
353 353 dirstateTupleObject *t = PyObject_New(dirstateTupleObject,
354 354 &dirstateTupleType);
355 355 if (!t)
356 356 return NULL;
357 357 t->state = state;
358 358 t->mode = mode;
359 359 t->size = size;
360 360 t->mtime = mtime;
361 361 return t;
362 362 }
363 363
364 364 static PyObject *dirstate_tuple_new(PyTypeObject *subtype, PyObject *args,
365 365 PyObject *kwds)
366 366 {
367 367 /* We do all the initialization here and not a tp_init function because
368 368 * dirstate_tuple is immutable. */
369 369 dirstateTupleObject *t;
370 370 char state;
371 371 int size, mode, mtime;
372 372 if (!PyArg_ParseTuple(args, "ciii", &state, &mode, &size, &mtime))
373 373 return NULL;
374 374
375 375 t = (dirstateTupleObject *)subtype->tp_alloc(subtype, 1);
376 376 if (!t)
377 377 return NULL;
378 378 t->state = state;
379 379 t->mode = mode;
380 380 t->size = size;
381 381 t->mtime = mtime;
382 382
383 383 return (PyObject *)t;
384 384 }
385 385
386 386 static void dirstate_tuple_dealloc(PyObject *o)
387 387 {
388 388 PyObject_Del(o);
389 389 }
390 390
391 391 static Py_ssize_t dirstate_tuple_length(PyObject *o)
392 392 {
393 393 return 4;
394 394 }
395 395
396 396 static PyObject *dirstate_tuple_item(PyObject *o, Py_ssize_t i)
397 397 {
398 398 dirstateTupleObject *t = (dirstateTupleObject *)o;
399 399 switch (i) {
400 400 case 0:
401 401 return PyBytes_FromStringAndSize(&t->state, 1);
402 402 case 1:
403 403 return PyInt_FromLong(t->mode);
404 404 case 2:
405 405 return PyInt_FromLong(t->size);
406 406 case 3:
407 407 return PyInt_FromLong(t->mtime);
408 408 default:
409 409 PyErr_SetString(PyExc_IndexError, "index out of range");
410 410 return NULL;
411 411 }
412 412 }
413 413
414 414 static PySequenceMethods dirstate_tuple_sq = {
415 415 dirstate_tuple_length, /* sq_length */
416 416 0, /* sq_concat */
417 417 0, /* sq_repeat */
418 418 dirstate_tuple_item, /* sq_item */
419 419 0, /* sq_ass_item */
420 420 0, /* sq_contains */
421 421 0, /* sq_inplace_concat */
422 422 0 /* sq_inplace_repeat */
423 423 };
424 424
425 425 PyTypeObject dirstateTupleType = {
426 426 PyVarObject_HEAD_INIT(NULL, 0)
427 427 "dirstate_tuple", /* tp_name */
428 428 sizeof(dirstateTupleObject),/* tp_basicsize */
429 429 0, /* tp_itemsize */
430 430 (destructor)dirstate_tuple_dealloc, /* tp_dealloc */
431 431 0, /* tp_print */
432 432 0, /* tp_getattr */
433 433 0, /* tp_setattr */
434 434 0, /* tp_compare */
435 435 0, /* tp_repr */
436 436 0, /* tp_as_number */
437 437 &dirstate_tuple_sq, /* tp_as_sequence */
438 438 0, /* tp_as_mapping */
439 439 0, /* tp_hash */
440 440 0, /* tp_call */
441 441 0, /* tp_str */
442 442 0, /* tp_getattro */
443 443 0, /* tp_setattro */
444 444 0, /* tp_as_buffer */
445 445 Py_TPFLAGS_DEFAULT, /* tp_flags */
446 446 "dirstate tuple", /* tp_doc */
447 447 0, /* tp_traverse */
448 448 0, /* tp_clear */
449 449 0, /* tp_richcompare */
450 450 0, /* tp_weaklistoffset */
451 451 0, /* tp_iter */
452 452 0, /* tp_iternext */
453 453 0, /* tp_methods */
454 454 0, /* tp_members */
455 455 0, /* tp_getset */
456 456 0, /* tp_base */
457 457 0, /* tp_dict */
458 458 0, /* tp_descr_get */
459 459 0, /* tp_descr_set */
460 460 0, /* tp_dictoffset */
461 461 0, /* tp_init */
462 462 0, /* tp_alloc */
463 463 dirstate_tuple_new, /* tp_new */
464 464 };
465 465
466 466 static PyObject *parse_dirstate(PyObject *self, PyObject *args)
467 467 {
468 468 PyObject *dmap, *cmap, *parents = NULL, *ret = NULL;
469 469 PyObject *fname = NULL, *cname = NULL, *entry = NULL;
470 470 char state, *cur, *str, *cpos;
471 471 int mode, size, mtime;
472 472 unsigned int flen, len, pos = 40;
473 473 int readlen;
474 474
475 475 if (!PyArg_ParseTuple(args, "O!O!s#:parse_dirstate",
476 476 &PyDict_Type, &dmap,
477 477 &PyDict_Type, &cmap,
478 478 &str, &readlen))
479 479 goto quit;
480 480
481 481 len = readlen;
482 482
483 483 /* read parents */
484 484 if (len < 40) {
485 485 PyErr_SetString(
486 486 PyExc_ValueError, "too little data for parents");
487 487 goto quit;
488 488 }
489 489
490 490 parents = Py_BuildValue("s#s#", str, 20, str + 20, 20);
491 491 if (!parents)
492 492 goto quit;
493 493
494 494 /* read filenames */
495 495 while (pos >= 40 && pos < len) {
496 496 if (pos + 17 > len) {
497 497 PyErr_SetString(PyExc_ValueError,
498 498 "overflow in dirstate");
499 499 goto quit;
500 500 }
501 501 cur = str + pos;
502 502 /* unpack header */
503 503 state = *cur;
504 504 mode = getbe32(cur + 1);
505 505 size = getbe32(cur + 5);
506 506 mtime = getbe32(cur + 9);
507 507 flen = getbe32(cur + 13);
508 508 pos += 17;
509 509 cur += 17;
510 510 if (flen > len - pos) {
511 511 PyErr_SetString(PyExc_ValueError, "overflow in dirstate");
512 512 goto quit;
513 513 }
514 514
515 515 entry = (PyObject *)make_dirstate_tuple(state, mode, size,
516 516 mtime);
517 517 cpos = memchr(cur, 0, flen);
518 518 if (cpos) {
519 519 fname = PyBytes_FromStringAndSize(cur, cpos - cur);
520 520 cname = PyBytes_FromStringAndSize(cpos + 1,
521 521 flen - (cpos - cur) - 1);
522 522 if (!fname || !cname ||
523 523 PyDict_SetItem(cmap, fname, cname) == -1 ||
524 524 PyDict_SetItem(dmap, fname, entry) == -1)
525 525 goto quit;
526 526 Py_DECREF(cname);
527 527 } else {
528 528 fname = PyBytes_FromStringAndSize(cur, flen);
529 529 if (!fname ||
530 530 PyDict_SetItem(dmap, fname, entry) == -1)
531 531 goto quit;
532 532 }
533 533 Py_DECREF(fname);
534 534 Py_DECREF(entry);
535 535 fname = cname = entry = NULL;
536 536 pos += flen;
537 537 }
538 538
539 539 ret = parents;
540 540 Py_INCREF(ret);
541 541 quit:
542 542 Py_XDECREF(fname);
543 543 Py_XDECREF(cname);
544 544 Py_XDECREF(entry);
545 545 Py_XDECREF(parents);
546 546 return ret;
547 547 }
548 548
549 549 /*
550 550 * Efficiently pack a dirstate object into its on-disk format.
551 551 */
552 552 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
553 553 {
554 554 PyObject *packobj = NULL;
555 555 PyObject *map, *copymap, *pl, *mtime_unset = NULL;
556 556 Py_ssize_t nbytes, pos, l;
557 557 PyObject *k, *v = NULL, *pn;
558 558 char *p, *s;
559 559 int now;
560 560
561 561 if (!PyArg_ParseTuple(args, "O!O!Oi:pack_dirstate",
562 562 &PyDict_Type, &map, &PyDict_Type, &copymap,
563 563 &pl, &now))
564 564 return NULL;
565 565
566 566 if (!PySequence_Check(pl) || PySequence_Size(pl) != 2) {
567 567 PyErr_SetString(PyExc_TypeError, "expected 2-element sequence");
568 568 return NULL;
569 569 }
570 570
571 571 /* Figure out how much we need to allocate. */
572 572 for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
573 573 PyObject *c;
574 574 if (!PyString_Check(k)) {
575 575 PyErr_SetString(PyExc_TypeError, "expected string key");
576 576 goto bail;
577 577 }
578 578 nbytes += PyString_GET_SIZE(k) + 17;
579 579 c = PyDict_GetItem(copymap, k);
580 580 if (c) {
581 581 if (!PyString_Check(c)) {
582 582 PyErr_SetString(PyExc_TypeError,
583 583 "expected string key");
584 584 goto bail;
585 585 }
586 586 nbytes += PyString_GET_SIZE(c) + 1;
587 587 }
588 588 }
589 589
590 590 packobj = PyString_FromStringAndSize(NULL, nbytes);
591 591 if (packobj == NULL)
592 592 goto bail;
593 593
594 594 p = PyString_AS_STRING(packobj);
595 595
596 596 pn = PySequence_ITEM(pl, 0);
597 597 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
598 598 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
599 599 goto bail;
600 600 }
601 601 memcpy(p, s, l);
602 602 p += 20;
603 603 pn = PySequence_ITEM(pl, 1);
604 604 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
605 605 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
606 606 goto bail;
607 607 }
608 608 memcpy(p, s, l);
609 609 p += 20;
610 610
611 611 for (pos = 0; PyDict_Next(map, &pos, &k, &v); ) {
612 612 dirstateTupleObject *tuple;
613 613 char state;
614 614 int mode, size, mtime;
615 615 Py_ssize_t len, l;
616 616 PyObject *o;
617 617 char *t;
618 618
619 619 if (!dirstate_tuple_check(v)) {
620 620 PyErr_SetString(PyExc_TypeError,
621 621 "expected a dirstate tuple");
622 622 goto bail;
623 623 }
624 624 tuple = (dirstateTupleObject *)v;
625 625
626 626 state = tuple->state;
627 627 mode = tuple->mode;
628 628 size = tuple->size;
629 629 mtime = tuple->mtime;
630 630 if (state == 'n' && mtime == now) {
631 631 /* See pure/parsers.py:pack_dirstate for why we do
632 632 * this. */
633 633 mtime = -1;
634 634 mtime_unset = (PyObject *)make_dirstate_tuple(
635 635 state, mode, size, mtime);
636 636 if (!mtime_unset)
637 637 goto bail;
638 638 if (PyDict_SetItem(map, k, mtime_unset) == -1)
639 639 goto bail;
640 640 Py_DECREF(mtime_unset);
641 641 mtime_unset = NULL;
642 642 }
643 643 *p++ = state;
644 644 putbe32((uint32_t)mode, p);
645 645 putbe32((uint32_t)size, p + 4);
646 646 putbe32((uint32_t)mtime, p + 8);
647 647 t = p + 12;
648 648 p += 16;
649 649 len = PyString_GET_SIZE(k);
650 650 memcpy(p, PyString_AS_STRING(k), len);
651 651 p += len;
652 652 o = PyDict_GetItem(copymap, k);
653 653 if (o) {
654 654 *p++ = '\0';
655 655 l = PyString_GET_SIZE(o);
656 656 memcpy(p, PyString_AS_STRING(o), l);
657 657 p += l;
658 658 len += l + 1;
659 659 }
660 660 putbe32((uint32_t)len, t);
661 661 }
662 662
663 663 pos = p - PyString_AS_STRING(packobj);
664 664 if (pos != nbytes) {
665 665 PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
666 666 (long)pos, (long)nbytes);
667 667 goto bail;
668 668 }
669 669
670 670 return packobj;
671 671 bail:
672 672 Py_XDECREF(mtime_unset);
673 673 Py_XDECREF(packobj);
674 674 Py_XDECREF(v);
675 675 return NULL;
676 676 }
677 677
678 678 /*
679 679 * A base-16 trie for fast node->rev mapping.
680 680 *
681 681 * Positive value is index of the next node in the trie
682 682 * Negative value is a leaf: -(rev + 1)
683 683 * Zero is empty
684 684 */
685 685 typedef struct {
686 686 int children[16];
687 687 } nodetree;
688 688
689 689 /*
690 690 * This class has two behaviors.
691 691 *
692 692 * When used in a list-like way (with integer keys), we decode an
693 693 * entry in a RevlogNG index file on demand. Our last entry is a
694 694 * sentinel, always a nullid. We have limited support for
695 695 * integer-keyed insert and delete, only at elements right before the
696 696 * sentinel.
697 697 *
698 698 * With string keys, we lazily perform a reverse mapping from node to
699 699 * rev, using a base-16 trie.
700 700 */
701 701 typedef struct {
702 702 PyObject_HEAD
703 703 /* Type-specific fields go here. */
704 704 PyObject *data; /* raw bytes of index */
705 705 PyObject **cache; /* cached tuples */
706 706 const char **offsets; /* populated on demand */
707 707 Py_ssize_t raw_length; /* original number of elements */
708 708 Py_ssize_t length; /* current number of elements */
709 709 PyObject *added; /* populated on demand */
710 710 PyObject *headrevs; /* cache, invalidated on changes */
711 711 PyObject *filteredrevs;/* filtered revs set */
712 712 nodetree *nt; /* base-16 trie */
713 713 unsigned ntlength; /* # nodes in use */
714 714 unsigned ntcapacity; /* # nodes allocated */
715 715 int ntdepth; /* maximum depth of tree */
716 716 int ntsplits; /* # splits performed */
717 717 int ntrev; /* last rev scanned */
718 718 int ntlookups; /* # lookups */
719 719 int ntmisses; /* # lookups that miss the cache */
720 720 int inlined;
721 721 } indexObject;
722 722
723 723 static Py_ssize_t index_length(const indexObject *self)
724 724 {
725 725 if (self->added == NULL)
726 726 return self->length;
727 727 return self->length + PyList_GET_SIZE(self->added);
728 728 }
729 729
730 730 static PyObject *nullentry;
731 731 static const char nullid[20];
732 732
733 733 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
734 734
735 735 #if LONG_MAX == 0x7fffffffL
736 736 static char *tuple_format = "Kiiiiiis#";
737 737 #else
738 738 static char *tuple_format = "kiiiiiis#";
739 739 #endif
740 740
741 741 /* A RevlogNG v1 index entry is 64 bytes long. */
742 742 static const long v1_hdrsize = 64;
743 743
744 744 /*
745 745 * Return a pointer to the beginning of a RevlogNG record.
746 746 */
747 747 static const char *index_deref(indexObject *self, Py_ssize_t pos)
748 748 {
749 749 if (self->inlined && pos > 0) {
750 750 if (self->offsets == NULL) {
751 751 self->offsets = malloc(self->raw_length *
752 752 sizeof(*self->offsets));
753 753 if (self->offsets == NULL)
754 754 return (const char *)PyErr_NoMemory();
755 755 inline_scan(self, self->offsets);
756 756 }
757 757 return self->offsets[pos];
758 758 }
759 759
760 760 return PyString_AS_STRING(self->data) + pos * v1_hdrsize;
761 761 }
762 762
763 763 static inline int index_get_parents(indexObject *self, Py_ssize_t rev,
764 764 int *ps, int maxrev)
765 765 {
766 766 if (rev >= self->length - 1) {
767 767 PyObject *tuple = PyList_GET_ITEM(self->added,
768 768 rev - self->length + 1);
769 769 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
770 770 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
771 771 } else {
772 772 const char *data = index_deref(self, rev);
773 773 ps[0] = getbe32(data + 24);
774 774 ps[1] = getbe32(data + 28);
775 775 }
776 776 /* If index file is corrupted, ps[] may point to invalid revisions. So
777 777 * there is a risk of buffer overflow to trust them unconditionally. */
778 778 if (ps[0] > maxrev || ps[1] > maxrev) {
779 779 PyErr_SetString(PyExc_ValueError, "parent out of range");
780 780 return -1;
781 781 }
782 782 return 0;
783 783 }
784 784
785 785
786 786 /*
787 787 * RevlogNG format (all in big endian, data may be inlined):
788 788 * 6 bytes: offset
789 789 * 2 bytes: flags
790 790 * 4 bytes: compressed length
791 791 * 4 bytes: uncompressed length
792 792 * 4 bytes: base revision
793 793 * 4 bytes: link revision
794 794 * 4 bytes: parent 1 revision
795 795 * 4 bytes: parent 2 revision
796 796 * 32 bytes: nodeid (only 20 bytes used)
797 797 */
798 798 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
799 799 {
800 800 uint64_t offset_flags;
801 801 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
802 802 const char *c_node_id;
803 803 const char *data;
804 804 Py_ssize_t length = index_length(self);
805 805 PyObject *entry;
806 806
807 807 if (pos < 0)
808 808 pos += length;
809 809
810 810 if (pos < 0 || pos >= length) {
811 811 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
812 812 return NULL;
813 813 }
814 814
815 815 if (pos == length - 1) {
816 816 Py_INCREF(nullentry);
817 817 return nullentry;
818 818 }
819 819
820 820 if (pos >= self->length - 1) {
821 821 PyObject *obj;
822 822 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
823 823 Py_INCREF(obj);
824 824 return obj;
825 825 }
826 826
827 827 if (self->cache) {
828 828 if (self->cache[pos]) {
829 829 Py_INCREF(self->cache[pos]);
830 830 return self->cache[pos];
831 831 }
832 832 } else {
833 833 self->cache = calloc(self->raw_length, sizeof(PyObject *));
834 834 if (self->cache == NULL)
835 835 return PyErr_NoMemory();
836 836 }
837 837
838 838 data = index_deref(self, pos);
839 839 if (data == NULL)
840 840 return NULL;
841 841
842 842 offset_flags = getbe32(data + 4);
843 843 if (pos == 0) /* mask out version number for the first entry */
844 844 offset_flags &= 0xFFFF;
845 845 else {
846 846 uint32_t offset_high = getbe32(data);
847 847 offset_flags |= ((uint64_t)offset_high) << 32;
848 848 }
849 849
850 850 comp_len = getbe32(data + 8);
851 851 uncomp_len = getbe32(data + 12);
852 852 base_rev = getbe32(data + 16);
853 853 link_rev = getbe32(data + 20);
854 854 parent_1 = getbe32(data + 24);
855 855 parent_2 = getbe32(data + 28);
856 856 c_node_id = data + 32;
857 857
858 858 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
859 859 uncomp_len, base_rev, link_rev,
860 860 parent_1, parent_2, c_node_id, 20);
861 861
862 862 if (entry) {
863 863 PyObject_GC_UnTrack(entry);
864 864 Py_INCREF(entry);
865 865 }
866 866
867 867 self->cache[pos] = entry;
868 868
869 869 return entry;
870 870 }
871 871
872 872 /*
873 873 * Return the 20-byte SHA of the node corresponding to the given rev.
874 874 */
875 875 static const char *index_node(indexObject *self, Py_ssize_t pos)
876 876 {
877 877 Py_ssize_t length = index_length(self);
878 878 const char *data;
879 879
880 880 if (pos == length - 1 || pos == INT_MAX)
881 881 return nullid;
882 882
883 883 if (pos >= length)
884 884 return NULL;
885 885
886 886 if (pos >= self->length - 1) {
887 887 PyObject *tuple, *str;
888 888 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
889 889 str = PyTuple_GetItem(tuple, 7);
890 890 return str ? PyString_AS_STRING(str) : NULL;
891 891 }
892 892
893 893 data = index_deref(self, pos);
894 894 return data ? data + 32 : NULL;
895 895 }
896 896
897 897 static int nt_insert(indexObject *self, const char *node, int rev);
898 898
899 899 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
900 900 {
901 901 if (PyString_AsStringAndSize(obj, node, nodelen) == -1)
902 902 return -1;
903 903 if (*nodelen == 20)
904 904 return 0;
905 905 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
906 906 return -1;
907 907 }
908 908
909 909 static PyObject *index_insert(indexObject *self, PyObject *args)
910 910 {
911 911 PyObject *obj;
912 912 char *node;
913 913 int index;
914 914 Py_ssize_t len, nodelen;
915 915
916 916 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
917 917 return NULL;
918 918
919 919 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
920 920 PyErr_SetString(PyExc_TypeError, "8-tuple required");
921 921 return NULL;
922 922 }
923 923
924 924 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
925 925 return NULL;
926 926
927 927 len = index_length(self);
928 928
929 929 if (index < 0)
930 930 index += len;
931 931
932 932 if (index != len - 1) {
933 933 PyErr_SetString(PyExc_IndexError,
934 934 "insert only supported at index -1");
935 935 return NULL;
936 936 }
937 937
938 938 if (self->added == NULL) {
939 939 self->added = PyList_New(0);
940 940 if (self->added == NULL)
941 941 return NULL;
942 942 }
943 943
944 944 if (PyList_Append(self->added, obj) == -1)
945 945 return NULL;
946 946
947 947 if (self->nt)
948 948 nt_insert(self, node, index);
949 949
950 950 Py_CLEAR(self->headrevs);
951 951 Py_RETURN_NONE;
952 952 }
953 953
954 954 static void _index_clearcaches(indexObject *self)
955 955 {
956 956 if (self->cache) {
957 957 Py_ssize_t i;
958 958
959 959 for (i = 0; i < self->raw_length; i++)
960 960 Py_CLEAR(self->cache[i]);
961 961 free(self->cache);
962 962 self->cache = NULL;
963 963 }
964 964 if (self->offsets) {
965 965 free(self->offsets);
966 966 self->offsets = NULL;
967 967 }
968 968 if (self->nt) {
969 969 free(self->nt);
970 970 self->nt = NULL;
971 971 }
972 972 Py_CLEAR(self->headrevs);
973 973 }
974 974
975 975 static PyObject *index_clearcaches(indexObject *self)
976 976 {
977 977 _index_clearcaches(self);
978 978 self->ntlength = self->ntcapacity = 0;
979 979 self->ntdepth = self->ntsplits = 0;
980 980 self->ntrev = -1;
981 981 self->ntlookups = self->ntmisses = 0;
982 982 Py_RETURN_NONE;
983 983 }
984 984
985 985 static PyObject *index_stats(indexObject *self)
986 986 {
987 987 PyObject *obj = PyDict_New();
988 988 PyObject *t = NULL;
989 989
990 990 if (obj == NULL)
991 991 return NULL;
992 992
993 993 #define istat(__n, __d) \
994 994 t = PyInt_FromSsize_t(self->__n); \
995 995 if (!t) \
996 996 goto bail; \
997 997 if (PyDict_SetItemString(obj, __d, t) == -1) \
998 998 goto bail; \
999 999 Py_DECREF(t);
1000 1000
1001 1001 if (self->added) {
1002 1002 Py_ssize_t len = PyList_GET_SIZE(self->added);
1003 1003 t = PyInt_FromSsize_t(len);
1004 1004 if (!t)
1005 1005 goto bail;
1006 1006 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
1007 1007 goto bail;
1008 1008 Py_DECREF(t);
1009 1009 }
1010 1010
1011 1011 if (self->raw_length != self->length - 1)
1012 1012 istat(raw_length, "revs on disk");
1013 1013 istat(length, "revs in memory");
1014 1014 istat(ntcapacity, "node trie capacity");
1015 1015 istat(ntdepth, "node trie depth");
1016 1016 istat(ntlength, "node trie count");
1017 1017 istat(ntlookups, "node trie lookups");
1018 1018 istat(ntmisses, "node trie misses");
1019 1019 istat(ntrev, "node trie last rev scanned");
1020 1020 istat(ntsplits, "node trie splits");
1021 1021
1022 1022 #undef istat
1023 1023
1024 1024 return obj;
1025 1025
1026 1026 bail:
1027 1027 Py_XDECREF(obj);
1028 1028 Py_XDECREF(t);
1029 1029 return NULL;
1030 1030 }
1031 1031
1032 1032 /*
1033 1033 * When we cache a list, we want to be sure the caller can't mutate
1034 1034 * the cached copy.
1035 1035 */
1036 1036 static PyObject *list_copy(PyObject *list)
1037 1037 {
1038 1038 Py_ssize_t len = PyList_GET_SIZE(list);
1039 1039 PyObject *newlist = PyList_New(len);
1040 1040 Py_ssize_t i;
1041 1041
1042 1042 if (newlist == NULL)
1043 1043 return NULL;
1044 1044
1045 1045 for (i = 0; i < len; i++) {
1046 1046 PyObject *obj = PyList_GET_ITEM(list, i);
1047 1047 Py_INCREF(obj);
1048 1048 PyList_SET_ITEM(newlist, i, obj);
1049 1049 }
1050 1050
1051 1051 return newlist;
1052 1052 }
1053 1053
1054 1054 static int check_filter(PyObject *filter, Py_ssize_t arg) {
1055 1055 if (filter) {
1056 1056 PyObject *arglist, *result;
1057 1057 int isfiltered;
1058 1058
1059 1059 arglist = Py_BuildValue("(n)", arg);
1060 1060 if (!arglist) {
1061 1061 return -1;
1062 1062 }
1063 1063
1064 1064 result = PyEval_CallObject(filter, arglist);
1065 1065 Py_DECREF(arglist);
1066 1066 if (!result) {
1067 1067 return -1;
1068 1068 }
1069 1069
1070 1070 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
1071 1071 * same as this function, so we can just return it directly.*/
1072 1072 isfiltered = PyObject_IsTrue(result);
1073 1073 Py_DECREF(result);
1074 1074 return isfiltered;
1075 1075 } else {
1076 1076 return 0;
1077 1077 }
1078 1078 }
1079 1079
1080 1080 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
1081 1081 Py_ssize_t marker, char *phases)
1082 1082 {
1083 1083 PyObject *iter = NULL;
1084 1084 PyObject *iter_item = NULL;
1085 1085 Py_ssize_t min_idx = index_length(self) + 1;
1086 1086 long iter_item_long;
1087 1087
1088 1088 if (PyList_GET_SIZE(list) != 0) {
1089 1089 iter = PyObject_GetIter(list);
1090 1090 if (iter == NULL)
1091 1091 return -2;
1092 1092 while ((iter_item = PyIter_Next(iter)))
1093 1093 {
1094 1094 iter_item_long = PyInt_AS_LONG(iter_item);
1095 1095 Py_DECREF(iter_item);
1096 1096 if (iter_item_long < min_idx)
1097 1097 min_idx = iter_item_long;
1098 1098 phases[iter_item_long] = marker;
1099 1099 }
1100 1100 Py_DECREF(iter);
1101 1101 }
1102 1102
1103 1103 return min_idx;
1104 1104 }
1105 1105
1106 1106 static inline void set_phase_from_parents(char *phases, int parent_1,
1107 1107 int parent_2, Py_ssize_t i)
1108 1108 {
1109 1109 if (parent_1 >= 0 && phases[parent_1] > phases[i])
1110 1110 phases[i] = phases[parent_1];
1111 1111 if (parent_2 >= 0 && phases[parent_2] > phases[i])
1112 1112 phases[i] = phases[parent_2];
1113 1113 }
1114 1114
1115 1115 static PyObject *reachableroots2(indexObject *self, PyObject *args)
1116 1116 {
1117 1117
1118 1118 /* Input */
1119 1119 long minroot;
1120 1120 PyObject *includepatharg = NULL;
1121 1121 int includepath = 0;
1122 1122 /* heads and roots are lists */
1123 1123 PyObject *heads = NULL;
1124 1124 PyObject *roots = NULL;
1125 1125 PyObject *reachable = NULL;
1126 1126
1127 1127 PyObject *val;
1128 1128 Py_ssize_t len = index_length(self) - 1;
1129 1129 long revnum;
1130 1130 Py_ssize_t k;
1131 1131 Py_ssize_t i;
1132 1132 Py_ssize_t l;
1133 1133 int r;
1134 1134 int parents[2];
1135 1135
1136 1136 /* Internal data structure:
1137 1137 * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit
1138 1138 * revstates: array of length len+1 (all revs + nullrev) */
1139 1139 int *tovisit = NULL;
1140 1140 long lentovisit = 0;
1141 1141 enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 };
1142 1142 char *revstates = NULL;
1143 1143
1144 1144 /* Get arguments */
1145 1145 if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
1146 1146 &PyList_Type, &roots,
1147 1147 &PyBool_Type, &includepatharg))
1148 1148 goto bail;
1149 1149
1150 1150 if (includepatharg == Py_True)
1151 1151 includepath = 1;
1152 1152
1153 1153 /* Initialize return set */
1154 1154 reachable = PyList_New(0);
1155 1155 if (reachable == NULL)
1156 1156 goto bail;
1157 1157
1158 1158 /* Initialize internal datastructures */
1159 1159 tovisit = (int *)malloc((len + 1) * sizeof(int));
1160 1160 if (tovisit == NULL) {
1161 1161 PyErr_NoMemory();
1162 1162 goto bail;
1163 1163 }
1164 1164
1165 1165 revstates = (char *)calloc(len + 1, 1);
1166 1166 if (revstates == NULL) {
1167 1167 PyErr_NoMemory();
1168 1168 goto bail;
1169 1169 }
1170 1170
1171 1171 l = PyList_GET_SIZE(roots);
1172 1172 for (i = 0; i < l; i++) {
1173 1173 revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i));
1174 1174 if (revnum == -1 && PyErr_Occurred())
1175 1175 goto bail;
1176 1176 /* If root is out of range, e.g. wdir(), it must be unreachable
1177 1177 * from heads. So we can just ignore it. */
1178 1178 if (revnum + 1 < 0 || revnum + 1 >= len + 1)
1179 1179 continue;
1180 1180 revstates[revnum + 1] |= RS_ROOT;
1181 1181 }
1182 1182
1183 1183 /* Populate tovisit with all the heads */
1184 1184 l = PyList_GET_SIZE(heads);
1185 1185 for (i = 0; i < l; i++) {
1186 1186 revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
1187 1187 if (revnum == -1 && PyErr_Occurred())
1188 1188 goto bail;
1189 1189 if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
1190 1190 PyErr_SetString(PyExc_IndexError, "head out of range");
1191 1191 goto bail;
1192 1192 }
1193 1193 if (!(revstates[revnum + 1] & RS_SEEN)) {
1194 1194 tovisit[lentovisit++] = (int)revnum;
1195 1195 revstates[revnum + 1] |= RS_SEEN;
1196 1196 }
1197 1197 }
1198 1198
1199 1199 /* Visit the tovisit list and find the reachable roots */
1200 1200 k = 0;
1201 1201 while (k < lentovisit) {
1202 1202 /* Add the node to reachable if it is a root*/
1203 1203 revnum = tovisit[k++];
1204 1204 if (revstates[revnum + 1] & RS_ROOT) {
1205 1205 revstates[revnum + 1] |= RS_REACHABLE;
1206 1206 val = PyInt_FromLong(revnum);
1207 1207 if (val == NULL)
1208 1208 goto bail;
1209 1209 r = PyList_Append(reachable, val);
1210 1210 Py_DECREF(val);
1211 1211 if (r < 0)
1212 1212 goto bail;
1213 1213 if (includepath == 0)
1214 1214 continue;
1215 1215 }
1216 1216
1217 1217 /* Add its parents to the list of nodes to visit */
1218 1218 if (revnum == -1)
1219 1219 continue;
1220 1220 r = index_get_parents(self, revnum, parents, (int)len - 1);
1221 1221 if (r < 0)
1222 1222 goto bail;
1223 1223 for (i = 0; i < 2; i++) {
1224 1224 if (!(revstates[parents[i] + 1] & RS_SEEN)
1225 1225 && parents[i] >= minroot) {
1226 1226 tovisit[lentovisit++] = parents[i];
1227 1227 revstates[parents[i] + 1] |= RS_SEEN;
1228 1228 }
1229 1229 }
1230 1230 }
1231 1231
1232 1232 /* Find all the nodes in between the roots we found and the heads
1233 1233 * and add them to the reachable set */
1234 1234 if (includepath == 1) {
1235 1235 long minidx = minroot;
1236 1236 if (minidx < 0)
1237 1237 minidx = 0;
1238 1238 for (i = minidx; i < len; i++) {
1239 1239 if (!(revstates[i + 1] & RS_SEEN))
1240 1240 continue;
1241 1241 r = index_get_parents(self, i, parents, (int)len - 1);
1242 1242 /* Corrupted index file, error is set from
1243 1243 * index_get_parents */
1244 1244 if (r < 0)
1245 1245 goto bail;
1246 1246 if (((revstates[parents[0] + 1] |
1247 1247 revstates[parents[1] + 1]) & RS_REACHABLE)
1248 1248 && !(revstates[i + 1] & RS_REACHABLE)) {
1249 1249 revstates[i + 1] |= RS_REACHABLE;
1250 1250 val = PyInt_FromLong(i);
1251 1251 if (val == NULL)
1252 1252 goto bail;
1253 1253 r = PyList_Append(reachable, val);
1254 1254 Py_DECREF(val);
1255 1255 if (r < 0)
1256 1256 goto bail;
1257 1257 }
1258 1258 }
1259 1259 }
1260 1260
1261 1261 free(revstates);
1262 1262 free(tovisit);
1263 1263 return reachable;
1264 1264 bail:
1265 1265 Py_XDECREF(reachable);
1266 1266 free(revstates);
1267 1267 free(tovisit);
1268 1268 return NULL;
1269 1269 }
1270 1270
1271 1271 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
1272 1272 {
1273 1273 PyObject *roots = Py_None;
1274 1274 PyObject *ret = NULL;
1275 1275 PyObject *phaseslist = NULL;
1276 1276 PyObject *phaseroots = NULL;
1277 1277 PyObject *phaseset = NULL;
1278 1278 PyObject *phasessetlist = NULL;
1279 1279 PyObject *rev = NULL;
1280 1280 Py_ssize_t len = index_length(self) - 1;
1281 1281 Py_ssize_t numphase = 0;
1282 1282 Py_ssize_t minrevallphases = 0;
1283 1283 Py_ssize_t minrevphase = 0;
1284 1284 Py_ssize_t i = 0;
1285 1285 char *phases = NULL;
1286 1286 long phase;
1287 1287
1288 1288 if (!PyArg_ParseTuple(args, "O", &roots))
1289 1289 goto done;
1290 1290 if (roots == NULL || !PyList_Check(roots))
1291 1291 goto done;
1292 1292
1293 1293 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
1294 1294 if (phases == NULL) {
1295 1295 PyErr_NoMemory();
1296 1296 goto done;
1297 1297 }
1298 1298 /* Put the phase information of all the roots in phases */
1299 1299 numphase = PyList_GET_SIZE(roots)+1;
1300 1300 minrevallphases = len + 1;
1301 1301 phasessetlist = PyList_New(numphase);
1302 1302 if (phasessetlist == NULL)
1303 1303 goto done;
1304 1304
1305 1305 PyList_SET_ITEM(phasessetlist, 0, Py_None);
1306 1306 Py_INCREF(Py_None);
1307 1307
1308 1308 for (i = 0; i < numphase-1; i++) {
1309 1309 phaseroots = PyList_GET_ITEM(roots, i);
1310 1310 phaseset = PySet_New(NULL);
1311 1311 if (phaseset == NULL)
1312 1312 goto release;
1313 1313 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
1314 1314 if (!PyList_Check(phaseroots))
1315 1315 goto release;
1316 1316 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
1317 1317 if (minrevphase == -2) /* Error from add_roots_get_min */
1318 1318 goto release;
1319 1319 minrevallphases = MIN(minrevallphases, minrevphase);
1320 1320 }
1321 1321 /* Propagate the phase information from the roots to the revs */
1322 1322 if (minrevallphases != -1) {
1323 1323 int parents[2];
1324 1324 for (i = minrevallphases; i < len; i++) {
1325 1325 if (index_get_parents(self, i, parents,
1326 1326 (int)len - 1) < 0)
1327 1327 goto release;
1328 1328 set_phase_from_parents(phases, parents[0], parents[1], i);
1329 1329 }
1330 1330 }
1331 1331 /* Transform phase list to a python list */
1332 1332 phaseslist = PyList_New(len);
1333 1333 if (phaseslist == NULL)
1334 1334 goto release;
1335 1335 for (i = 0; i < len; i++) {
1336 PyObject *phaseval;
1337
1336 1338 phase = phases[i];
1337 1339 /* We only store the sets of phase for non public phase, the public phase
1338 1340 * is computed as a difference */
1339 1341 if (phase != 0) {
1340 1342 phaseset = PyList_GET_ITEM(phasessetlist, phase);
1341 1343 rev = PyInt_FromLong(i);
1344 if (rev == NULL)
1345 goto release;
1342 1346 PySet_Add(phaseset, rev);
1343 1347 Py_XDECREF(rev);
1344 1348 }
1345 PyList_SET_ITEM(phaseslist, i, PyInt_FromLong(phase));
1349 phaseval = PyInt_FromLong(phase);
1350 if (phaseval == NULL)
1351 goto release;
1352 PyList_SET_ITEM(phaseslist, i, phaseval);
1346 1353 }
1347 1354 ret = PyList_New(2);
1348 1355 if (ret == NULL)
1349 1356 goto release;
1350 1357
1351 1358 PyList_SET_ITEM(ret, 0, phaseslist);
1352 1359 PyList_SET_ITEM(ret, 1, phasessetlist);
1353 1360 /* We don't release phaseslist and phasessetlist as we return them to
1354 1361 * python */
1355 1362 goto done;
1356 1363
1357 1364 release:
1358 1365 Py_XDECREF(phaseslist);
1359 1366 Py_XDECREF(phasessetlist);
1360 1367 done:
1361 1368 free(phases);
1362 1369 return ret;
1363 1370 }
1364 1371
1365 1372 static PyObject *index_headrevs(indexObject *self, PyObject *args)
1366 1373 {
1367 1374 Py_ssize_t i, j, len;
1368 1375 char *nothead = NULL;
1369 1376 PyObject *heads = NULL;
1370 1377 PyObject *filter = NULL;
1371 1378 PyObject *filteredrevs = Py_None;
1372 1379
1373 1380 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
1374 1381 return NULL;
1375 1382 }
1376 1383
1377 1384 if (self->headrevs && filteredrevs == self->filteredrevs)
1378 1385 return list_copy(self->headrevs);
1379 1386
1380 1387 Py_DECREF(self->filteredrevs);
1381 1388 self->filteredrevs = filteredrevs;
1382 1389 Py_INCREF(filteredrevs);
1383 1390
1384 1391 if (filteredrevs != Py_None) {
1385 1392 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
1386 1393 if (!filter) {
1387 1394 PyErr_SetString(PyExc_TypeError,
1388 1395 "filteredrevs has no attribute __contains__");
1389 1396 goto bail;
1390 1397 }
1391 1398 }
1392 1399
1393 1400 len = index_length(self) - 1;
1394 1401 heads = PyList_New(0);
1395 1402 if (heads == NULL)
1396 1403 goto bail;
1397 1404 if (len == 0) {
1398 1405 PyObject *nullid = PyInt_FromLong(-1);
1399 1406 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
1400 1407 Py_XDECREF(nullid);
1401 1408 goto bail;
1402 1409 }
1403 1410 goto done;
1404 1411 }
1405 1412
1406 1413 nothead = calloc(len, 1);
1407 1414 if (nothead == NULL)
1408 1415 goto bail;
1409 1416
1410 1417 for (i = 0; i < len; i++) {
1411 1418 int isfiltered;
1412 1419 int parents[2];
1413 1420
1414 1421 isfiltered = check_filter(filter, i);
1415 1422 if (isfiltered == -1) {
1416 1423 PyErr_SetString(PyExc_TypeError,
1417 1424 "unable to check filter");
1418 1425 goto bail;
1419 1426 }
1420 1427
1421 1428 if (isfiltered) {
1422 1429 nothead[i] = 1;
1423 1430 continue;
1424 1431 }
1425 1432
1426 1433 if (index_get_parents(self, i, parents, (int)len - 1) < 0)
1427 1434 goto bail;
1428 1435 for (j = 0; j < 2; j++) {
1429 1436 if (parents[j] >= 0)
1430 1437 nothead[parents[j]] = 1;
1431 1438 }
1432 1439 }
1433 1440
1434 1441 for (i = 0; i < len; i++) {
1435 1442 PyObject *head;
1436 1443
1437 1444 if (nothead[i])
1438 1445 continue;
1439 1446 head = PyInt_FromSsize_t(i);
1440 1447 if (head == NULL || PyList_Append(heads, head) == -1) {
1441 1448 Py_XDECREF(head);
1442 1449 goto bail;
1443 1450 }
1444 1451 }
1445 1452
1446 1453 done:
1447 1454 self->headrevs = heads;
1448 1455 Py_XDECREF(filter);
1449 1456 free(nothead);
1450 1457 return list_copy(self->headrevs);
1451 1458 bail:
1452 1459 Py_XDECREF(filter);
1453 1460 Py_XDECREF(heads);
1454 1461 free(nothead);
1455 1462 return NULL;
1456 1463 }
1457 1464
1458 1465 static inline int nt_level(const char *node, Py_ssize_t level)
1459 1466 {
1460 1467 int v = node[level>>1];
1461 1468 if (!(level & 1))
1462 1469 v >>= 4;
1463 1470 return v & 0xf;
1464 1471 }
1465 1472
1466 1473 /*
1467 1474 * Return values:
1468 1475 *
1469 1476 * -4: match is ambiguous (multiple candidates)
1470 1477 * -2: not found
1471 1478 * rest: valid rev
1472 1479 */
1473 1480 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
1474 1481 int hex)
1475 1482 {
1476 1483 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
1477 1484 int level, maxlevel, off;
1478 1485
1479 1486 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
1480 1487 return -1;
1481 1488
1482 1489 if (self->nt == NULL)
1483 1490 return -2;
1484 1491
1485 1492 if (hex)
1486 1493 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
1487 1494 else
1488 1495 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
1489 1496
1490 1497 for (level = off = 0; level < maxlevel; level++) {
1491 1498 int k = getnybble(node, level);
1492 1499 nodetree *n = &self->nt[off];
1493 1500 int v = n->children[k];
1494 1501
1495 1502 if (v < 0) {
1496 1503 const char *n;
1497 1504 Py_ssize_t i;
1498 1505
1499 1506 v = -(v + 1);
1500 1507 n = index_node(self, v);
1501 1508 if (n == NULL)
1502 1509 return -2;
1503 1510 for (i = level; i < maxlevel; i++)
1504 1511 if (getnybble(node, i) != nt_level(n, i))
1505 1512 return -2;
1506 1513 return v;
1507 1514 }
1508 1515 if (v == 0)
1509 1516 return -2;
1510 1517 off = v;
1511 1518 }
1512 1519 /* multiple matches against an ambiguous prefix */
1513 1520 return -4;
1514 1521 }
1515 1522
1516 1523 static int nt_new(indexObject *self)
1517 1524 {
1518 1525 if (self->ntlength == self->ntcapacity) {
1519 1526 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1520 1527 PyErr_SetString(PyExc_MemoryError,
1521 1528 "overflow in nt_new");
1522 1529 return -1;
1523 1530 }
1524 1531 self->ntcapacity *= 2;
1525 1532 self->nt = realloc(self->nt,
1526 1533 self->ntcapacity * sizeof(nodetree));
1527 1534 if (self->nt == NULL) {
1528 1535 PyErr_SetString(PyExc_MemoryError, "out of memory");
1529 1536 return -1;
1530 1537 }
1531 1538 memset(&self->nt[self->ntlength], 0,
1532 1539 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1533 1540 }
1534 1541 return self->ntlength++;
1535 1542 }
1536 1543
1537 1544 static int nt_insert(indexObject *self, const char *node, int rev)
1538 1545 {
1539 1546 int level = 0;
1540 1547 int off = 0;
1541 1548
1542 1549 while (level < 40) {
1543 1550 int k = nt_level(node, level);
1544 1551 nodetree *n;
1545 1552 int v;
1546 1553
1547 1554 n = &self->nt[off];
1548 1555 v = n->children[k];
1549 1556
1550 1557 if (v == 0) {
1551 1558 n->children[k] = -rev - 1;
1552 1559 return 0;
1553 1560 }
1554 1561 if (v < 0) {
1555 1562 const char *oldnode = index_node(self, -(v + 1));
1556 1563 int noff;
1557 1564
1558 1565 if (!oldnode || !memcmp(oldnode, node, 20)) {
1559 1566 n->children[k] = -rev - 1;
1560 1567 return 0;
1561 1568 }
1562 1569 noff = nt_new(self);
1563 1570 if (noff == -1)
1564 1571 return -1;
1565 1572 /* self->nt may have been changed by realloc */
1566 1573 self->nt[off].children[k] = noff;
1567 1574 off = noff;
1568 1575 n = &self->nt[off];
1569 1576 n->children[nt_level(oldnode, ++level)] = v;
1570 1577 if (level > self->ntdepth)
1571 1578 self->ntdepth = level;
1572 1579 self->ntsplits += 1;
1573 1580 } else {
1574 1581 level += 1;
1575 1582 off = v;
1576 1583 }
1577 1584 }
1578 1585
1579 1586 return -1;
1580 1587 }
1581 1588
1582 1589 static int nt_init(indexObject *self)
1583 1590 {
1584 1591 if (self->nt == NULL) {
1585 1592 if ((size_t)self->raw_length > INT_MAX / sizeof(nodetree)) {
1586 1593 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1587 1594 return -1;
1588 1595 }
1589 1596 self->ntcapacity = self->raw_length < 4
1590 1597 ? 4 : (int)self->raw_length / 2;
1591 1598
1592 1599 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1593 1600 if (self->nt == NULL) {
1594 1601 PyErr_NoMemory();
1595 1602 return -1;
1596 1603 }
1597 1604 self->ntlength = 1;
1598 1605 self->ntrev = (int)index_length(self) - 1;
1599 1606 self->ntlookups = 1;
1600 1607 self->ntmisses = 0;
1601 1608 if (nt_insert(self, nullid, INT_MAX) == -1)
1602 1609 return -1;
1603 1610 }
1604 1611 return 0;
1605 1612 }
1606 1613
1607 1614 /*
1608 1615 * Return values:
1609 1616 *
1610 1617 * -3: error (exception set)
1611 1618 * -2: not found (no exception set)
1612 1619 * rest: valid rev
1613 1620 */
1614 1621 static int index_find_node(indexObject *self,
1615 1622 const char *node, Py_ssize_t nodelen)
1616 1623 {
1617 1624 int rev;
1618 1625
1619 1626 self->ntlookups++;
1620 1627 rev = nt_find(self, node, nodelen, 0);
1621 1628 if (rev >= -1)
1622 1629 return rev;
1623 1630
1624 1631 if (nt_init(self) == -1)
1625 1632 return -3;
1626 1633
1627 1634 /*
1628 1635 * For the first handful of lookups, we scan the entire index,
1629 1636 * and cache only the matching nodes. This optimizes for cases
1630 1637 * like "hg tip", where only a few nodes are accessed.
1631 1638 *
1632 1639 * After that, we cache every node we visit, using a single
1633 1640 * scan amortized over multiple lookups. This gives the best
1634 1641 * bulk performance, e.g. for "hg log".
1635 1642 */
1636 1643 if (self->ntmisses++ < 4) {
1637 1644 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1638 1645 const char *n = index_node(self, rev);
1639 1646 if (n == NULL)
1640 1647 return -2;
1641 1648 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1642 1649 if (nt_insert(self, n, rev) == -1)
1643 1650 return -3;
1644 1651 break;
1645 1652 }
1646 1653 }
1647 1654 } else {
1648 1655 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1649 1656 const char *n = index_node(self, rev);
1650 1657 if (n == NULL) {
1651 1658 self->ntrev = rev + 1;
1652 1659 return -2;
1653 1660 }
1654 1661 if (nt_insert(self, n, rev) == -1) {
1655 1662 self->ntrev = rev + 1;
1656 1663 return -3;
1657 1664 }
1658 1665 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1659 1666 break;
1660 1667 }
1661 1668 }
1662 1669 self->ntrev = rev;
1663 1670 }
1664 1671
1665 1672 if (rev >= 0)
1666 1673 return rev;
1667 1674 return -2;
1668 1675 }
1669 1676
1670 1677 static void raise_revlog_error(void)
1671 1678 {
1672 1679 PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
1673 1680
1674 1681 mod = PyImport_ImportModule("mercurial.error");
1675 1682 if (mod == NULL) {
1676 1683 goto cleanup;
1677 1684 }
1678 1685
1679 1686 dict = PyModule_GetDict(mod);
1680 1687 if (dict == NULL) {
1681 1688 goto cleanup;
1682 1689 }
1683 1690 Py_INCREF(dict);
1684 1691
1685 1692 errclass = PyDict_GetItemString(dict, "RevlogError");
1686 1693 if (errclass == NULL) {
1687 1694 PyErr_SetString(PyExc_SystemError,
1688 1695 "could not find RevlogError");
1689 1696 goto cleanup;
1690 1697 }
1691 1698
1692 1699 /* value of exception is ignored by callers */
1693 1700 PyErr_SetString(errclass, "RevlogError");
1694 1701
1695 1702 cleanup:
1696 1703 Py_XDECREF(dict);
1697 1704 Py_XDECREF(mod);
1698 1705 }
1699 1706
1700 1707 static PyObject *index_getitem(indexObject *self, PyObject *value)
1701 1708 {
1702 1709 char *node;
1703 1710 Py_ssize_t nodelen;
1704 1711 int rev;
1705 1712
1706 1713 if (PyInt_Check(value))
1707 1714 return index_get(self, PyInt_AS_LONG(value));
1708 1715
1709 1716 if (node_check(value, &node, &nodelen) == -1)
1710 1717 return NULL;
1711 1718 rev = index_find_node(self, node, nodelen);
1712 1719 if (rev >= -1)
1713 1720 return PyInt_FromLong(rev);
1714 1721 if (rev == -2)
1715 1722 raise_revlog_error();
1716 1723 return NULL;
1717 1724 }
1718 1725
1719 1726 static int nt_partialmatch(indexObject *self, const char *node,
1720 1727 Py_ssize_t nodelen)
1721 1728 {
1722 1729 int rev;
1723 1730
1724 1731 if (nt_init(self) == -1)
1725 1732 return -3;
1726 1733
1727 1734 if (self->ntrev > 0) {
1728 1735 /* ensure that the radix tree is fully populated */
1729 1736 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1730 1737 const char *n = index_node(self, rev);
1731 1738 if (n == NULL)
1732 1739 return -2;
1733 1740 if (nt_insert(self, n, rev) == -1)
1734 1741 return -3;
1735 1742 }
1736 1743 self->ntrev = rev;
1737 1744 }
1738 1745
1739 1746 return nt_find(self, node, nodelen, 1);
1740 1747 }
1741 1748
1742 1749 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1743 1750 {
1744 1751 const char *fullnode;
1745 1752 int nodelen;
1746 1753 char *node;
1747 1754 int rev, i;
1748 1755
1749 1756 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1750 1757 return NULL;
1751 1758
1752 1759 if (nodelen < 4) {
1753 1760 PyErr_SetString(PyExc_ValueError, "key too short");
1754 1761 return NULL;
1755 1762 }
1756 1763
1757 1764 if (nodelen > 40) {
1758 1765 PyErr_SetString(PyExc_ValueError, "key too long");
1759 1766 return NULL;
1760 1767 }
1761 1768
1762 1769 for (i = 0; i < nodelen; i++)
1763 1770 hexdigit(node, i);
1764 1771 if (PyErr_Occurred()) {
1765 1772 /* input contains non-hex characters */
1766 1773 PyErr_Clear();
1767 1774 Py_RETURN_NONE;
1768 1775 }
1769 1776
1770 1777 rev = nt_partialmatch(self, node, nodelen);
1771 1778
1772 1779 switch (rev) {
1773 1780 case -4:
1774 1781 raise_revlog_error();
1775 1782 case -3:
1776 1783 return NULL;
1777 1784 case -2:
1778 1785 Py_RETURN_NONE;
1779 1786 case -1:
1780 1787 return PyString_FromStringAndSize(nullid, 20);
1781 1788 }
1782 1789
1783 1790 fullnode = index_node(self, rev);
1784 1791 if (fullnode == NULL) {
1785 1792 PyErr_Format(PyExc_IndexError,
1786 1793 "could not access rev %d", rev);
1787 1794 return NULL;
1788 1795 }
1789 1796 return PyString_FromStringAndSize(fullnode, 20);
1790 1797 }
1791 1798
1792 1799 static PyObject *index_m_get(indexObject *self, PyObject *args)
1793 1800 {
1794 1801 Py_ssize_t nodelen;
1795 1802 PyObject *val;
1796 1803 char *node;
1797 1804 int rev;
1798 1805
1799 1806 if (!PyArg_ParseTuple(args, "O", &val))
1800 1807 return NULL;
1801 1808 if (node_check(val, &node, &nodelen) == -1)
1802 1809 return NULL;
1803 1810 rev = index_find_node(self, node, nodelen);
1804 1811 if (rev == -3)
1805 1812 return NULL;
1806 1813 if (rev == -2)
1807 1814 Py_RETURN_NONE;
1808 1815 return PyInt_FromLong(rev);
1809 1816 }
1810 1817
1811 1818 static int index_contains(indexObject *self, PyObject *value)
1812 1819 {
1813 1820 char *node;
1814 1821 Py_ssize_t nodelen;
1815 1822
1816 1823 if (PyInt_Check(value)) {
1817 1824 long rev = PyInt_AS_LONG(value);
1818 1825 return rev >= -1 && rev < index_length(self);
1819 1826 }
1820 1827
1821 1828 if (node_check(value, &node, &nodelen) == -1)
1822 1829 return -1;
1823 1830
1824 1831 switch (index_find_node(self, node, nodelen)) {
1825 1832 case -3:
1826 1833 return -1;
1827 1834 case -2:
1828 1835 return 0;
1829 1836 default:
1830 1837 return 1;
1831 1838 }
1832 1839 }
1833 1840
1834 1841 typedef uint64_t bitmask;
1835 1842
1836 1843 /*
1837 1844 * Given a disjoint set of revs, return all candidates for the
1838 1845 * greatest common ancestor. In revset notation, this is the set
1839 1846 * "heads(::a and ::b and ...)"
1840 1847 */
1841 1848 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1842 1849 int revcount)
1843 1850 {
1844 1851 const bitmask allseen = (1ull << revcount) - 1;
1845 1852 const bitmask poison = 1ull << revcount;
1846 1853 PyObject *gca = PyList_New(0);
1847 1854 int i, v, interesting;
1848 1855 int maxrev = -1;
1849 1856 bitmask sp;
1850 1857 bitmask *seen;
1851 1858
1852 1859 if (gca == NULL)
1853 1860 return PyErr_NoMemory();
1854 1861
1855 1862 for (i = 0; i < revcount; i++) {
1856 1863 if (revs[i] > maxrev)
1857 1864 maxrev = revs[i];
1858 1865 }
1859 1866
1860 1867 seen = calloc(sizeof(*seen), maxrev + 1);
1861 1868 if (seen == NULL) {
1862 1869 Py_DECREF(gca);
1863 1870 return PyErr_NoMemory();
1864 1871 }
1865 1872
1866 1873 for (i = 0; i < revcount; i++)
1867 1874 seen[revs[i]] = 1ull << i;
1868 1875
1869 1876 interesting = revcount;
1870 1877
1871 1878 for (v = maxrev; v >= 0 && interesting; v--) {
1872 1879 bitmask sv = seen[v];
1873 1880 int parents[2];
1874 1881
1875 1882 if (!sv)
1876 1883 continue;
1877 1884
1878 1885 if (sv < poison) {
1879 1886 interesting -= 1;
1880 1887 if (sv == allseen) {
1881 1888 PyObject *obj = PyInt_FromLong(v);
1882 1889 if (obj == NULL)
1883 1890 goto bail;
1884 1891 if (PyList_Append(gca, obj) == -1) {
1885 1892 Py_DECREF(obj);
1886 1893 goto bail;
1887 1894 }
1888 1895 sv |= poison;
1889 1896 for (i = 0; i < revcount; i++) {
1890 1897 if (revs[i] == v)
1891 1898 goto done;
1892 1899 }
1893 1900 }
1894 1901 }
1895 1902 if (index_get_parents(self, v, parents, maxrev) < 0)
1896 1903 goto bail;
1897 1904
1898 1905 for (i = 0; i < 2; i++) {
1899 1906 int p = parents[i];
1900 1907 if (p == -1)
1901 1908 continue;
1902 1909 sp = seen[p];
1903 1910 if (sv < poison) {
1904 1911 if (sp == 0) {
1905 1912 seen[p] = sv;
1906 1913 interesting++;
1907 1914 }
1908 1915 else if (sp != sv)
1909 1916 seen[p] |= sv;
1910 1917 } else {
1911 1918 if (sp && sp < poison)
1912 1919 interesting--;
1913 1920 seen[p] = sv;
1914 1921 }
1915 1922 }
1916 1923 }
1917 1924
1918 1925 done:
1919 1926 free(seen);
1920 1927 return gca;
1921 1928 bail:
1922 1929 free(seen);
1923 1930 Py_XDECREF(gca);
1924 1931 return NULL;
1925 1932 }
1926 1933
1927 1934 /*
1928 1935 * Given a disjoint set of revs, return the subset with the longest
1929 1936 * path to the root.
1930 1937 */
1931 1938 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1932 1939 {
1933 1940 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1934 1941 static const Py_ssize_t capacity = 24;
1935 1942 int *depth, *interesting = NULL;
1936 1943 int i, j, v, ninteresting;
1937 1944 PyObject *dict = NULL, *keys = NULL;
1938 1945 long *seen = NULL;
1939 1946 int maxrev = -1;
1940 1947 long final;
1941 1948
1942 1949 if (revcount > capacity) {
1943 1950 PyErr_Format(PyExc_OverflowError,
1944 1951 "bitset size (%ld) > capacity (%ld)",
1945 1952 (long)revcount, (long)capacity);
1946 1953 return NULL;
1947 1954 }
1948 1955
1949 1956 for (i = 0; i < revcount; i++) {
1950 1957 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1951 1958 if (n > maxrev)
1952 1959 maxrev = n;
1953 1960 }
1954 1961
1955 1962 depth = calloc(sizeof(*depth), maxrev + 1);
1956 1963 if (depth == NULL)
1957 1964 return PyErr_NoMemory();
1958 1965
1959 1966 seen = calloc(sizeof(*seen), maxrev + 1);
1960 1967 if (seen == NULL) {
1961 1968 PyErr_NoMemory();
1962 1969 goto bail;
1963 1970 }
1964 1971
1965 1972 interesting = calloc(sizeof(*interesting), 2 << revcount);
1966 1973 if (interesting == NULL) {
1967 1974 PyErr_NoMemory();
1968 1975 goto bail;
1969 1976 }
1970 1977
1971 1978 if (PyList_Sort(revs) == -1)
1972 1979 goto bail;
1973 1980
1974 1981 for (i = 0; i < revcount; i++) {
1975 1982 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1976 1983 long b = 1l << i;
1977 1984 depth[n] = 1;
1978 1985 seen[n] = b;
1979 1986 interesting[b] = 1;
1980 1987 }
1981 1988
1982 1989 ninteresting = (int)revcount;
1983 1990
1984 1991 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
1985 1992 int dv = depth[v];
1986 1993 int parents[2];
1987 1994 long sv;
1988 1995
1989 1996 if (dv == 0)
1990 1997 continue;
1991 1998
1992 1999 sv = seen[v];
1993 2000 if (index_get_parents(self, v, parents, maxrev) < 0)
1994 2001 goto bail;
1995 2002
1996 2003 for (i = 0; i < 2; i++) {
1997 2004 int p = parents[i];
1998 2005 long sp;
1999 2006 int dp;
2000 2007
2001 2008 if (p == -1)
2002 2009 continue;
2003 2010
2004 2011 dp = depth[p];
2005 2012 sp = seen[p];
2006 2013 if (dp <= dv) {
2007 2014 depth[p] = dv + 1;
2008 2015 if (sp != sv) {
2009 2016 interesting[sv] += 1;
2010 2017 seen[p] = sv;
2011 2018 if (sp) {
2012 2019 interesting[sp] -= 1;
2013 2020 if (interesting[sp] == 0)
2014 2021 ninteresting -= 1;
2015 2022 }
2016 2023 }
2017 2024 }
2018 2025 else if (dv == dp - 1) {
2019 2026 long nsp = sp | sv;
2020 2027 if (nsp == sp)
2021 2028 continue;
2022 2029 seen[p] = nsp;
2023 2030 interesting[sp] -= 1;
2024 2031 if (interesting[sp] == 0 && interesting[nsp] > 0)
2025 2032 ninteresting -= 1;
2026 2033 interesting[nsp] += 1;
2027 2034 }
2028 2035 }
2029 2036 interesting[sv] -= 1;
2030 2037 if (interesting[sv] == 0)
2031 2038 ninteresting -= 1;
2032 2039 }
2033 2040
2034 2041 final = 0;
2035 2042 j = ninteresting;
2036 2043 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
2037 2044 if (interesting[i] == 0)
2038 2045 continue;
2039 2046 final |= i;
2040 2047 j -= 1;
2041 2048 }
2042 2049 if (final == 0) {
2043 2050 keys = PyList_New(0);
2044 2051 goto bail;
2045 2052 }
2046 2053
2047 2054 dict = PyDict_New();
2048 2055 if (dict == NULL)
2049 2056 goto bail;
2050 2057
2051 2058 for (i = 0; i < revcount; i++) {
2052 2059 PyObject *key;
2053 2060
2054 2061 if ((final & (1 << i)) == 0)
2055 2062 continue;
2056 2063
2057 2064 key = PyList_GET_ITEM(revs, i);
2058 2065 Py_INCREF(key);
2059 2066 Py_INCREF(Py_None);
2060 2067 if (PyDict_SetItem(dict, key, Py_None) == -1) {
2061 2068 Py_DECREF(key);
2062 2069 Py_DECREF(Py_None);
2063 2070 goto bail;
2064 2071 }
2065 2072 }
2066 2073
2067 2074 keys = PyDict_Keys(dict);
2068 2075
2069 2076 bail:
2070 2077 free(depth);
2071 2078 free(seen);
2072 2079 free(interesting);
2073 2080 Py_XDECREF(dict);
2074 2081
2075 2082 return keys;
2076 2083 }
2077 2084
2078 2085 /*
2079 2086 * Given a (possibly overlapping) set of revs, return all the
2080 2087 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
2081 2088 */
2082 2089 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
2083 2090 {
2084 2091 PyObject *ret = NULL;
2085 2092 Py_ssize_t argcount, i, len;
2086 2093 bitmask repeat = 0;
2087 2094 int revcount = 0;
2088 2095 int *revs;
2089 2096
2090 2097 argcount = PySequence_Length(args);
2091 2098 revs = malloc(argcount * sizeof(*revs));
2092 2099 if (argcount > 0 && revs == NULL)
2093 2100 return PyErr_NoMemory();
2094 2101 len = index_length(self) - 1;
2095 2102
2096 2103 for (i = 0; i < argcount; i++) {
2097 2104 static const int capacity = 24;
2098 2105 PyObject *obj = PySequence_GetItem(args, i);
2099 2106 bitmask x;
2100 2107 long val;
2101 2108
2102 2109 if (!PyInt_Check(obj)) {
2103 2110 PyErr_SetString(PyExc_TypeError,
2104 2111 "arguments must all be ints");
2105 2112 Py_DECREF(obj);
2106 2113 goto bail;
2107 2114 }
2108 2115 val = PyInt_AsLong(obj);
2109 2116 Py_DECREF(obj);
2110 2117 if (val == -1) {
2111 2118 ret = PyList_New(0);
2112 2119 goto done;
2113 2120 }
2114 2121 if (val < 0 || val >= len) {
2115 2122 PyErr_SetString(PyExc_IndexError,
2116 2123 "index out of range");
2117 2124 goto bail;
2118 2125 }
2119 2126 /* this cheesy bloom filter lets us avoid some more
2120 2127 * expensive duplicate checks in the common set-is-disjoint
2121 2128 * case */
2122 2129 x = 1ull << (val & 0x3f);
2123 2130 if (repeat & x) {
2124 2131 int k;
2125 2132 for (k = 0; k < revcount; k++) {
2126 2133 if (val == revs[k])
2127 2134 goto duplicate;
2128 2135 }
2129 2136 }
2130 2137 else repeat |= x;
2131 2138 if (revcount >= capacity) {
2132 2139 PyErr_Format(PyExc_OverflowError,
2133 2140 "bitset size (%d) > capacity (%d)",
2134 2141 revcount, capacity);
2135 2142 goto bail;
2136 2143 }
2137 2144 revs[revcount++] = (int)val;
2138 2145 duplicate:;
2139 2146 }
2140 2147
2141 2148 if (revcount == 0) {
2142 2149 ret = PyList_New(0);
2143 2150 goto done;
2144 2151 }
2145 2152 if (revcount == 1) {
2146 2153 PyObject *obj;
2147 2154 ret = PyList_New(1);
2148 2155 if (ret == NULL)
2149 2156 goto bail;
2150 2157 obj = PyInt_FromLong(revs[0]);
2151 2158 if (obj == NULL)
2152 2159 goto bail;
2153 2160 PyList_SET_ITEM(ret, 0, obj);
2154 2161 goto done;
2155 2162 }
2156 2163
2157 2164 ret = find_gca_candidates(self, revs, revcount);
2158 2165 if (ret == NULL)
2159 2166 goto bail;
2160 2167
2161 2168 done:
2162 2169 free(revs);
2163 2170 return ret;
2164 2171
2165 2172 bail:
2166 2173 free(revs);
2167 2174 Py_XDECREF(ret);
2168 2175 return NULL;
2169 2176 }
2170 2177
2171 2178 /*
2172 2179 * Given a (possibly overlapping) set of revs, return the greatest
2173 2180 * common ancestors: those with the longest path to the root.
2174 2181 */
2175 2182 static PyObject *index_ancestors(indexObject *self, PyObject *args)
2176 2183 {
2177 2184 PyObject *ret;
2178 2185 PyObject *gca = index_commonancestorsheads(self, args);
2179 2186 if (gca == NULL)
2180 2187 return NULL;
2181 2188
2182 2189 if (PyList_GET_SIZE(gca) <= 1) {
2183 2190 return gca;
2184 2191 }
2185 2192
2186 2193 ret = find_deepest(self, gca);
2187 2194 Py_DECREF(gca);
2188 2195 return ret;
2189 2196 }
2190 2197
2191 2198 /*
2192 2199 * Invalidate any trie entries introduced by added revs.
2193 2200 */
2194 2201 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
2195 2202 {
2196 2203 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
2197 2204
2198 2205 for (i = start; i < len; i++) {
2199 2206 PyObject *tuple = PyList_GET_ITEM(self->added, i);
2200 2207 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
2201 2208
2202 2209 nt_insert(self, PyString_AS_STRING(node), -1);
2203 2210 }
2204 2211
2205 2212 if (start == 0)
2206 2213 Py_CLEAR(self->added);
2207 2214 }
2208 2215
2209 2216 /*
2210 2217 * Delete a numeric range of revs, which must be at the end of the
2211 2218 * range, but exclude the sentinel nullid entry.
2212 2219 */
2213 2220 static int index_slice_del(indexObject *self, PyObject *item)
2214 2221 {
2215 2222 Py_ssize_t start, stop, step, slicelength;
2216 2223 Py_ssize_t length = index_length(self);
2217 2224 int ret = 0;
2218 2225
2219 2226 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
2220 2227 &start, &stop, &step, &slicelength) < 0)
2221 2228 return -1;
2222 2229
2223 2230 if (slicelength <= 0)
2224 2231 return 0;
2225 2232
2226 2233 if ((step < 0 && start < stop) || (step > 0 && start > stop))
2227 2234 stop = start;
2228 2235
2229 2236 if (step < 0) {
2230 2237 stop = start + 1;
2231 2238 start = stop + step*(slicelength - 1) - 1;
2232 2239 step = -step;
2233 2240 }
2234 2241
2235 2242 if (step != 1) {
2236 2243 PyErr_SetString(PyExc_ValueError,
2237 2244 "revlog index delete requires step size of 1");
2238 2245 return -1;
2239 2246 }
2240 2247
2241 2248 if (stop != length - 1) {
2242 2249 PyErr_SetString(PyExc_IndexError,
2243 2250 "revlog index deletion indices are invalid");
2244 2251 return -1;
2245 2252 }
2246 2253
2247 2254 if (start < self->length - 1) {
2248 2255 if (self->nt) {
2249 2256 Py_ssize_t i;
2250 2257
2251 2258 for (i = start + 1; i < self->length - 1; i++) {
2252 2259 const char *node = index_node(self, i);
2253 2260
2254 2261 if (node)
2255 2262 nt_insert(self, node, -1);
2256 2263 }
2257 2264 if (self->added)
2258 2265 nt_invalidate_added(self, 0);
2259 2266 if (self->ntrev > start)
2260 2267 self->ntrev = (int)start;
2261 2268 }
2262 2269 self->length = start + 1;
2263 2270 if (start < self->raw_length) {
2264 2271 if (self->cache) {
2265 2272 Py_ssize_t i;
2266 2273 for (i = start; i < self->raw_length; i++)
2267 2274 Py_CLEAR(self->cache[i]);
2268 2275 }
2269 2276 self->raw_length = start;
2270 2277 }
2271 2278 goto done;
2272 2279 }
2273 2280
2274 2281 if (self->nt) {
2275 2282 nt_invalidate_added(self, start - self->length + 1);
2276 2283 if (self->ntrev > start)
2277 2284 self->ntrev = (int)start;
2278 2285 }
2279 2286 if (self->added)
2280 2287 ret = PyList_SetSlice(self->added, start - self->length + 1,
2281 2288 PyList_GET_SIZE(self->added), NULL);
2282 2289 done:
2283 2290 Py_CLEAR(self->headrevs);
2284 2291 return ret;
2285 2292 }
2286 2293
2287 2294 /*
2288 2295 * Supported ops:
2289 2296 *
2290 2297 * slice deletion
2291 2298 * string assignment (extend node->rev mapping)
2292 2299 * string deletion (shrink node->rev mapping)
2293 2300 */
2294 2301 static int index_assign_subscript(indexObject *self, PyObject *item,
2295 2302 PyObject *value)
2296 2303 {
2297 2304 char *node;
2298 2305 Py_ssize_t nodelen;
2299 2306 long rev;
2300 2307
2301 2308 if (PySlice_Check(item) && value == NULL)
2302 2309 return index_slice_del(self, item);
2303 2310
2304 2311 if (node_check(item, &node, &nodelen) == -1)
2305 2312 return -1;
2306 2313
2307 2314 if (value == NULL)
2308 2315 return self->nt ? nt_insert(self, node, -1) : 0;
2309 2316 rev = PyInt_AsLong(value);
2310 2317 if (rev > INT_MAX || rev < 0) {
2311 2318 if (!PyErr_Occurred())
2312 2319 PyErr_SetString(PyExc_ValueError, "rev out of range");
2313 2320 return -1;
2314 2321 }
2315 2322
2316 2323 if (nt_init(self) == -1)
2317 2324 return -1;
2318 2325 return nt_insert(self, node, (int)rev);
2319 2326 }
2320 2327
2321 2328 /*
2322 2329 * Find all RevlogNG entries in an index that has inline data. Update
2323 2330 * the optional "offsets" table with those entries.
2324 2331 */
2325 2332 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
2326 2333 {
2327 2334 const char *data = PyString_AS_STRING(self->data);
2328 2335 Py_ssize_t pos = 0;
2329 2336 Py_ssize_t end = PyString_GET_SIZE(self->data);
2330 2337 long incr = v1_hdrsize;
2331 2338 Py_ssize_t len = 0;
2332 2339
2333 2340 while (pos + v1_hdrsize <= end && pos >= 0) {
2334 2341 uint32_t comp_len;
2335 2342 /* 3rd element of header is length of compressed inline data */
2336 2343 comp_len = getbe32(data + pos + 8);
2337 2344 incr = v1_hdrsize + comp_len;
2338 2345 if (offsets)
2339 2346 offsets[len] = data + pos;
2340 2347 len++;
2341 2348 pos += incr;
2342 2349 }
2343 2350
2344 2351 if (pos != end) {
2345 2352 if (!PyErr_Occurred())
2346 2353 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2347 2354 return -1;
2348 2355 }
2349 2356
2350 2357 return len;
2351 2358 }
2352 2359
2353 2360 static int index_init(indexObject *self, PyObject *args)
2354 2361 {
2355 2362 PyObject *data_obj, *inlined_obj;
2356 2363 Py_ssize_t size;
2357 2364
2358 2365 /* Initialize before argument-checking to avoid index_dealloc() crash. */
2359 2366 self->raw_length = 0;
2360 2367 self->added = NULL;
2361 2368 self->cache = NULL;
2362 2369 self->data = NULL;
2363 2370 self->headrevs = NULL;
2364 2371 self->filteredrevs = Py_None;
2365 2372 Py_INCREF(Py_None);
2366 2373 self->nt = NULL;
2367 2374 self->offsets = NULL;
2368 2375
2369 2376 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
2370 2377 return -1;
2371 2378 if (!PyString_Check(data_obj)) {
2372 2379 PyErr_SetString(PyExc_TypeError, "data is not a string");
2373 2380 return -1;
2374 2381 }
2375 2382 size = PyString_GET_SIZE(data_obj);
2376 2383
2377 2384 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
2378 2385 self->data = data_obj;
2379 2386
2380 2387 self->ntlength = self->ntcapacity = 0;
2381 2388 self->ntdepth = self->ntsplits = 0;
2382 2389 self->ntlookups = self->ntmisses = 0;
2383 2390 self->ntrev = -1;
2384 2391 Py_INCREF(self->data);
2385 2392
2386 2393 if (self->inlined) {
2387 2394 Py_ssize_t len = inline_scan(self, NULL);
2388 2395 if (len == -1)
2389 2396 goto bail;
2390 2397 self->raw_length = len;
2391 2398 self->length = len + 1;
2392 2399 } else {
2393 2400 if (size % v1_hdrsize) {
2394 2401 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2395 2402 goto bail;
2396 2403 }
2397 2404 self->raw_length = size / v1_hdrsize;
2398 2405 self->length = self->raw_length + 1;
2399 2406 }
2400 2407
2401 2408 return 0;
2402 2409 bail:
2403 2410 return -1;
2404 2411 }
2405 2412
2406 2413 static PyObject *index_nodemap(indexObject *self)
2407 2414 {
2408 2415 Py_INCREF(self);
2409 2416 return (PyObject *)self;
2410 2417 }
2411 2418
2412 2419 static void index_dealloc(indexObject *self)
2413 2420 {
2414 2421 _index_clearcaches(self);
2415 2422 Py_XDECREF(self->filteredrevs);
2416 2423 Py_XDECREF(self->data);
2417 2424 Py_XDECREF(self->added);
2418 2425 PyObject_Del(self);
2419 2426 }
2420 2427
2421 2428 static PySequenceMethods index_sequence_methods = {
2422 2429 (lenfunc)index_length, /* sq_length */
2423 2430 0, /* sq_concat */
2424 2431 0, /* sq_repeat */
2425 2432 (ssizeargfunc)index_get, /* sq_item */
2426 2433 0, /* sq_slice */
2427 2434 0, /* sq_ass_item */
2428 2435 0, /* sq_ass_slice */
2429 2436 (objobjproc)index_contains, /* sq_contains */
2430 2437 };
2431 2438
2432 2439 static PyMappingMethods index_mapping_methods = {
2433 2440 (lenfunc)index_length, /* mp_length */
2434 2441 (binaryfunc)index_getitem, /* mp_subscript */
2435 2442 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
2436 2443 };
2437 2444
2438 2445 static PyMethodDef index_methods[] = {
2439 2446 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
2440 2447 "return the gca set of the given revs"},
2441 2448 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
2442 2449 METH_VARARGS,
2443 2450 "return the heads of the common ancestors of the given revs"},
2444 2451 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
2445 2452 "clear the index caches"},
2446 2453 {"get", (PyCFunction)index_m_get, METH_VARARGS,
2447 2454 "get an index entry"},
2448 2455 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
2449 2456 METH_VARARGS, "compute phases"},
2450 2457 {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS,
2451 2458 "reachableroots"},
2452 2459 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
2453 2460 "get head revisions"}, /* Can do filtering since 3.2 */
2454 2461 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
2455 2462 "get filtered head revisions"}, /* Can always do filtering */
2456 2463 {"insert", (PyCFunction)index_insert, METH_VARARGS,
2457 2464 "insert an index entry"},
2458 2465 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
2459 2466 "match a potentially ambiguous node ID"},
2460 2467 {"stats", (PyCFunction)index_stats, METH_NOARGS,
2461 2468 "stats for the index"},
2462 2469 {NULL} /* Sentinel */
2463 2470 };
2464 2471
2465 2472 static PyGetSetDef index_getset[] = {
2466 2473 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
2467 2474 {NULL} /* Sentinel */
2468 2475 };
2469 2476
2470 2477 static PyTypeObject indexType = {
2471 2478 PyObject_HEAD_INIT(NULL)
2472 2479 0, /* ob_size */
2473 2480 "parsers.index", /* tp_name */
2474 2481 sizeof(indexObject), /* tp_basicsize */
2475 2482 0, /* tp_itemsize */
2476 2483 (destructor)index_dealloc, /* tp_dealloc */
2477 2484 0, /* tp_print */
2478 2485 0, /* tp_getattr */
2479 2486 0, /* tp_setattr */
2480 2487 0, /* tp_compare */
2481 2488 0, /* tp_repr */
2482 2489 0, /* tp_as_number */
2483 2490 &index_sequence_methods, /* tp_as_sequence */
2484 2491 &index_mapping_methods, /* tp_as_mapping */
2485 2492 0, /* tp_hash */
2486 2493 0, /* tp_call */
2487 2494 0, /* tp_str */
2488 2495 0, /* tp_getattro */
2489 2496 0, /* tp_setattro */
2490 2497 0, /* tp_as_buffer */
2491 2498 Py_TPFLAGS_DEFAULT, /* tp_flags */
2492 2499 "revlog index", /* tp_doc */
2493 2500 0, /* tp_traverse */
2494 2501 0, /* tp_clear */
2495 2502 0, /* tp_richcompare */
2496 2503 0, /* tp_weaklistoffset */
2497 2504 0, /* tp_iter */
2498 2505 0, /* tp_iternext */
2499 2506 index_methods, /* tp_methods */
2500 2507 0, /* tp_members */
2501 2508 index_getset, /* tp_getset */
2502 2509 0, /* tp_base */
2503 2510 0, /* tp_dict */
2504 2511 0, /* tp_descr_get */
2505 2512 0, /* tp_descr_set */
2506 2513 0, /* tp_dictoffset */
2507 2514 (initproc)index_init, /* tp_init */
2508 2515 0, /* tp_alloc */
2509 2516 };
2510 2517
2511 2518 /*
2512 2519 * returns a tuple of the form (index, index, cache) with elements as
2513 2520 * follows:
2514 2521 *
2515 2522 * index: an index object that lazily parses RevlogNG records
2516 2523 * cache: if data is inlined, a tuple (index_file_content, 0), else None
2517 2524 *
2518 2525 * added complications are for backwards compatibility
2519 2526 */
2520 2527 static PyObject *parse_index2(PyObject *self, PyObject *args)
2521 2528 {
2522 2529 PyObject *tuple = NULL, *cache = NULL;
2523 2530 indexObject *idx;
2524 2531 int ret;
2525 2532
2526 2533 idx = PyObject_New(indexObject, &indexType);
2527 2534 if (idx == NULL)
2528 2535 goto bail;
2529 2536
2530 2537 ret = index_init(idx, args);
2531 2538 if (ret == -1)
2532 2539 goto bail;
2533 2540
2534 2541 if (idx->inlined) {
2535 2542 cache = Py_BuildValue("iO", 0, idx->data);
2536 2543 if (cache == NULL)
2537 2544 goto bail;
2538 2545 } else {
2539 2546 cache = Py_None;
2540 2547 Py_INCREF(cache);
2541 2548 }
2542 2549
2543 2550 tuple = Py_BuildValue("NN", idx, cache);
2544 2551 if (!tuple)
2545 2552 goto bail;
2546 2553 return tuple;
2547 2554
2548 2555 bail:
2549 2556 Py_XDECREF(idx);
2550 2557 Py_XDECREF(cache);
2551 2558 Py_XDECREF(tuple);
2552 2559 return NULL;
2553 2560 }
2554 2561
2555 2562 #define BUMPED_FIX 1
2556 2563 #define USING_SHA_256 2
2557 2564 #define FM1_HEADER_SIZE (4 + 8 + 2 + 2 + 1 + 1 + 1)
2558 2565
2559 2566 static PyObject *readshas(
2560 2567 const char *source, unsigned char num, Py_ssize_t hashwidth)
2561 2568 {
2562 2569 int i;
2563 2570 PyObject *list = PyTuple_New(num);
2564 2571 if (list == NULL) {
2565 2572 return NULL;
2566 2573 }
2567 2574 for (i = 0; i < num; i++) {
2568 2575 PyObject *hash = PyString_FromStringAndSize(source, hashwidth);
2569 2576 if (hash == NULL) {
2570 2577 Py_DECREF(list);
2571 2578 return NULL;
2572 2579 }
2573 2580 PyTuple_SET_ITEM(list, i, hash);
2574 2581 source += hashwidth;
2575 2582 }
2576 2583 return list;
2577 2584 }
2578 2585
2579 2586 static PyObject *fm1readmarker(const char *databegin, const char *dataend,
2580 2587 uint32_t *msize)
2581 2588 {
2582 2589 const char *data = databegin;
2583 2590 const char *meta;
2584 2591
2585 2592 double mtime;
2586 2593 int16_t tz;
2587 2594 uint16_t flags;
2588 2595 unsigned char nsuccs, nparents, nmetadata;
2589 2596 Py_ssize_t hashwidth = 20;
2590 2597
2591 2598 PyObject *prec = NULL, *parents = NULL, *succs = NULL;
2592 2599 PyObject *metadata = NULL, *ret = NULL;
2593 2600 int i;
2594 2601
2595 2602 if (data + FM1_HEADER_SIZE > dataend) {
2596 2603 goto overflow;
2597 2604 }
2598 2605
2599 2606 *msize = getbe32(data);
2600 2607 data += 4;
2601 2608 mtime = getbefloat64(data);
2602 2609 data += 8;
2603 2610 tz = getbeint16(data);
2604 2611 data += 2;
2605 2612 flags = getbeuint16(data);
2606 2613 data += 2;
2607 2614
2608 2615 if (flags & USING_SHA_256) {
2609 2616 hashwidth = 32;
2610 2617 }
2611 2618
2612 2619 nsuccs = (unsigned char)(*data++);
2613 2620 nparents = (unsigned char)(*data++);
2614 2621 nmetadata = (unsigned char)(*data++);
2615 2622
2616 2623 if (databegin + *msize > dataend) {
2617 2624 goto overflow;
2618 2625 }
2619 2626 dataend = databegin + *msize; /* narrow down to marker size */
2620 2627
2621 2628 if (data + hashwidth > dataend) {
2622 2629 goto overflow;
2623 2630 }
2624 2631 prec = PyString_FromStringAndSize(data, hashwidth);
2625 2632 data += hashwidth;
2626 2633 if (prec == NULL) {
2627 2634 goto bail;
2628 2635 }
2629 2636
2630 2637 if (data + nsuccs * hashwidth > dataend) {
2631 2638 goto overflow;
2632 2639 }
2633 2640 succs = readshas(data, nsuccs, hashwidth);
2634 2641 if (succs == NULL) {
2635 2642 goto bail;
2636 2643 }
2637 2644 data += nsuccs * hashwidth;
2638 2645
2639 2646 if (nparents == 1 || nparents == 2) {
2640 2647 if (data + nparents * hashwidth > dataend) {
2641 2648 goto overflow;
2642 2649 }
2643 2650 parents = readshas(data, nparents, hashwidth);
2644 2651 if (parents == NULL) {
2645 2652 goto bail;
2646 2653 }
2647 2654 data += nparents * hashwidth;
2648 2655 } else {
2649 2656 parents = Py_None;
2650 2657 }
2651 2658
2652 2659 if (data + 2 * nmetadata > dataend) {
2653 2660 goto overflow;
2654 2661 }
2655 2662 meta = data + (2 * nmetadata);
2656 2663 metadata = PyTuple_New(nmetadata);
2657 2664 if (metadata == NULL) {
2658 2665 goto bail;
2659 2666 }
2660 2667 for (i = 0; i < nmetadata; i++) {
2661 2668 PyObject *tmp, *left = NULL, *right = NULL;
2662 2669 Py_ssize_t leftsize = (unsigned char)(*data++);
2663 2670 Py_ssize_t rightsize = (unsigned char)(*data++);
2664 2671 if (meta + leftsize + rightsize > dataend) {
2665 2672 goto overflow;
2666 2673 }
2667 2674 left = PyString_FromStringAndSize(meta, leftsize);
2668 2675 meta += leftsize;
2669 2676 right = PyString_FromStringAndSize(meta, rightsize);
2670 2677 meta += rightsize;
2671 2678 tmp = PyTuple_New(2);
2672 2679 if (!left || !right || !tmp) {
2673 2680 Py_XDECREF(left);
2674 2681 Py_XDECREF(right);
2675 2682 Py_XDECREF(tmp);
2676 2683 goto bail;
2677 2684 }
2678 2685 PyTuple_SET_ITEM(tmp, 0, left);
2679 2686 PyTuple_SET_ITEM(tmp, 1, right);
2680 2687 PyTuple_SET_ITEM(metadata, i, tmp);
2681 2688 }
2682 2689 ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags,
2683 2690 metadata, mtime, (int)tz * 60, parents);
2684 2691 goto bail; /* return successfully */
2685 2692
2686 2693 overflow:
2687 2694 PyErr_SetString(PyExc_ValueError, "overflow in obsstore");
2688 2695 bail:
2689 2696 Py_XDECREF(prec);
2690 2697 Py_XDECREF(succs);
2691 2698 Py_XDECREF(metadata);
2692 2699 if (parents != Py_None)
2693 2700 Py_XDECREF(parents);
2694 2701 return ret;
2695 2702 }
2696 2703
2697 2704
2698 2705 static PyObject *fm1readmarkers(PyObject *self, PyObject *args) {
2699 2706 const char *data, *dataend;
2700 2707 int datalen;
2701 2708 Py_ssize_t offset, stop;
2702 2709 PyObject *markers = NULL;
2703 2710
2704 2711 if (!PyArg_ParseTuple(args, "s#nn", &data, &datalen, &offset, &stop)) {
2705 2712 return NULL;
2706 2713 }
2707 2714 dataend = data + datalen;
2708 2715 data += offset;
2709 2716 markers = PyList_New(0);
2710 2717 if (!markers) {
2711 2718 return NULL;
2712 2719 }
2713 2720 while (offset < stop) {
2714 2721 uint32_t msize;
2715 2722 int error;
2716 2723 PyObject *record = fm1readmarker(data, dataend, &msize);
2717 2724 if (!record) {
2718 2725 goto bail;
2719 2726 }
2720 2727 error = PyList_Append(markers, record);
2721 2728 Py_DECREF(record);
2722 2729 if (error) {
2723 2730 goto bail;
2724 2731 }
2725 2732 data += msize;
2726 2733 offset += msize;
2727 2734 }
2728 2735 return markers;
2729 2736 bail:
2730 2737 Py_DECREF(markers);
2731 2738 return NULL;
2732 2739 }
2733 2740
2734 2741 static char parsers_doc[] = "Efficient content parsing.";
2735 2742
2736 2743 PyObject *encodedir(PyObject *self, PyObject *args);
2737 2744 PyObject *pathencode(PyObject *self, PyObject *args);
2738 2745 PyObject *lowerencode(PyObject *self, PyObject *args);
2739 2746
2740 2747 static PyMethodDef methods[] = {
2741 2748 {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
2742 2749 {"parse_manifest", parse_manifest, METH_VARARGS, "parse a manifest\n"},
2743 2750 {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
2744 2751 {"parse_index2", parse_index2, METH_VARARGS, "parse a revlog index\n"},
2745 2752 {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
2746 2753 {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
2747 2754 {"dict_new_presized", dict_new_presized, METH_VARARGS,
2748 2755 "construct a dict with an expected size\n"},
2749 2756 {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
2750 2757 "make file foldmap\n"},
2751 2758 {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
2752 2759 {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
2753 2760 {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
2754 2761 {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
2755 2762 "parse v1 obsolete markers\n"},
2756 2763 {NULL, NULL}
2757 2764 };
2758 2765
2759 2766 void dirs_module_init(PyObject *mod);
2760 2767 void manifest_module_init(PyObject *mod);
2761 2768
2762 2769 static void module_init(PyObject *mod)
2763 2770 {
2764 2771 /* This module constant has two purposes. First, it lets us unit test
2765 2772 * the ImportError raised without hard-coding any error text. This
2766 2773 * means we can change the text in the future without breaking tests,
2767 2774 * even across changesets without a recompile. Second, its presence
2768 2775 * can be used to determine whether the version-checking logic is
2769 2776 * present, which also helps in testing across changesets without a
2770 2777 * recompile. Note that this means the pure-Python version of parsers
2771 2778 * should not have this module constant. */
2772 2779 PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
2773 2780
2774 2781 dirs_module_init(mod);
2775 2782 manifest_module_init(mod);
2776 2783
2777 2784 indexType.tp_new = PyType_GenericNew;
2778 2785 if (PyType_Ready(&indexType) < 0 ||
2779 2786 PyType_Ready(&dirstateTupleType) < 0)
2780 2787 return;
2781 2788 Py_INCREF(&indexType);
2782 2789 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2783 2790 Py_INCREF(&dirstateTupleType);
2784 2791 PyModule_AddObject(mod, "dirstatetuple",
2785 2792 (PyObject *)&dirstateTupleType);
2786 2793
2787 2794 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2788 2795 -1, -1, -1, -1, nullid, 20);
2789 2796 if (nullentry)
2790 2797 PyObject_GC_UnTrack(nullentry);
2791 2798 }
2792 2799
2793 2800 static int check_python_version(void)
2794 2801 {
2795 2802 PyObject *sys = PyImport_ImportModule("sys"), *ver;
2796 2803 long hexversion;
2797 2804 if (!sys)
2798 2805 return -1;
2799 2806 ver = PyObject_GetAttrString(sys, "hexversion");
2800 2807 Py_DECREF(sys);
2801 2808 if (!ver)
2802 2809 return -1;
2803 2810 hexversion = PyInt_AsLong(ver);
2804 2811 Py_DECREF(ver);
2805 2812 /* sys.hexversion is a 32-bit number by default, so the -1 case
2806 2813 * should only occur in unusual circumstances (e.g. if sys.hexversion
2807 2814 * is manually set to an invalid value). */
2808 2815 if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
2809 2816 PyErr_Format(PyExc_ImportError, "%s: The Mercurial extension "
2810 2817 "modules were compiled with Python " PY_VERSION ", but "
2811 2818 "Mercurial is currently using Python with sys.hexversion=%ld: "
2812 2819 "Python %s\n at: %s", versionerrortext, hexversion,
2813 2820 Py_GetVersion(), Py_GetProgramFullPath());
2814 2821 return -1;
2815 2822 }
2816 2823 return 0;
2817 2824 }
2818 2825
2819 2826 #ifdef IS_PY3K
2820 2827 static struct PyModuleDef parsers_module = {
2821 2828 PyModuleDef_HEAD_INIT,
2822 2829 "parsers",
2823 2830 parsers_doc,
2824 2831 -1,
2825 2832 methods
2826 2833 };
2827 2834
2828 2835 PyMODINIT_FUNC PyInit_parsers(void)
2829 2836 {
2830 2837 PyObject *mod;
2831 2838
2832 2839 if (check_python_version() == -1)
2833 2840 return;
2834 2841 mod = PyModule_Create(&parsers_module);
2835 2842 module_init(mod);
2836 2843 return mod;
2837 2844 }
2838 2845 #else
2839 2846 PyMODINIT_FUNC initparsers(void)
2840 2847 {
2841 2848 PyObject *mod;
2842 2849
2843 2850 if (check_python_version() == -1)
2844 2851 return;
2845 2852 mod = Py_InitModule3("parsers", methods, parsers_doc);
2846 2853 module_init(mod);
2847 2854 }
2848 2855 #endif
General Comments 0
You need to be logged in to leave comments. Login now