##// END OF EJS Templates
reachableroots: handle error of PyList_Append()
Yuya Nishihara -
r26058:e7fe0a12 default
parent child Browse files
Show More
@@ -1,2814 +1,2818 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 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 cur = str + pos;
497 497 /* unpack header */
498 498 state = *cur;
499 499 mode = getbe32(cur + 1);
500 500 size = getbe32(cur + 5);
501 501 mtime = getbe32(cur + 9);
502 502 flen = getbe32(cur + 13);
503 503 pos += 17;
504 504 cur += 17;
505 505 if (flen > len - pos) {
506 506 PyErr_SetString(PyExc_ValueError, "overflow in dirstate");
507 507 goto quit;
508 508 }
509 509
510 510 entry = (PyObject *)make_dirstate_tuple(state, mode, size,
511 511 mtime);
512 512 cpos = memchr(cur, 0, flen);
513 513 if (cpos) {
514 514 fname = PyBytes_FromStringAndSize(cur, cpos - cur);
515 515 cname = PyBytes_FromStringAndSize(cpos + 1,
516 516 flen - (cpos - cur) - 1);
517 517 if (!fname || !cname ||
518 518 PyDict_SetItem(cmap, fname, cname) == -1 ||
519 519 PyDict_SetItem(dmap, fname, entry) == -1)
520 520 goto quit;
521 521 Py_DECREF(cname);
522 522 } else {
523 523 fname = PyBytes_FromStringAndSize(cur, flen);
524 524 if (!fname ||
525 525 PyDict_SetItem(dmap, fname, entry) == -1)
526 526 goto quit;
527 527 }
528 528 Py_DECREF(fname);
529 529 Py_DECREF(entry);
530 530 fname = cname = entry = NULL;
531 531 pos += flen;
532 532 }
533 533
534 534 ret = parents;
535 535 Py_INCREF(ret);
536 536 quit:
537 537 Py_XDECREF(fname);
538 538 Py_XDECREF(cname);
539 539 Py_XDECREF(entry);
540 540 Py_XDECREF(parents);
541 541 return ret;
542 542 }
543 543
544 544 /*
545 545 * Efficiently pack a dirstate object into its on-disk format.
546 546 */
547 547 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
548 548 {
549 549 PyObject *packobj = NULL;
550 550 PyObject *map, *copymap, *pl, *mtime_unset = NULL;
551 551 Py_ssize_t nbytes, pos, l;
552 552 PyObject *k, *v = NULL, *pn;
553 553 char *p, *s;
554 554 double now;
555 555
556 556 if (!PyArg_ParseTuple(args, "O!O!Od:pack_dirstate",
557 557 &PyDict_Type, &map, &PyDict_Type, &copymap,
558 558 &pl, &now))
559 559 return NULL;
560 560
561 561 if (!PySequence_Check(pl) || PySequence_Size(pl) != 2) {
562 562 PyErr_SetString(PyExc_TypeError, "expected 2-element sequence");
563 563 return NULL;
564 564 }
565 565
566 566 /* Figure out how much we need to allocate. */
567 567 for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
568 568 PyObject *c;
569 569 if (!PyString_Check(k)) {
570 570 PyErr_SetString(PyExc_TypeError, "expected string key");
571 571 goto bail;
572 572 }
573 573 nbytes += PyString_GET_SIZE(k) + 17;
574 574 c = PyDict_GetItem(copymap, k);
575 575 if (c) {
576 576 if (!PyString_Check(c)) {
577 577 PyErr_SetString(PyExc_TypeError,
578 578 "expected string key");
579 579 goto bail;
580 580 }
581 581 nbytes += PyString_GET_SIZE(c) + 1;
582 582 }
583 583 }
584 584
585 585 packobj = PyString_FromStringAndSize(NULL, nbytes);
586 586 if (packobj == NULL)
587 587 goto bail;
588 588
589 589 p = PyString_AS_STRING(packobj);
590 590
591 591 pn = PySequence_ITEM(pl, 0);
592 592 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
593 593 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
594 594 goto bail;
595 595 }
596 596 memcpy(p, s, l);
597 597 p += 20;
598 598 pn = PySequence_ITEM(pl, 1);
599 599 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
600 600 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
601 601 goto bail;
602 602 }
603 603 memcpy(p, s, l);
604 604 p += 20;
605 605
606 606 for (pos = 0; PyDict_Next(map, &pos, &k, &v); ) {
607 607 dirstateTupleObject *tuple;
608 608 char state;
609 609 uint32_t mode, size, mtime;
610 610 Py_ssize_t len, l;
611 611 PyObject *o;
612 612 char *t;
613 613
614 614 if (!dirstate_tuple_check(v)) {
615 615 PyErr_SetString(PyExc_TypeError,
616 616 "expected a dirstate tuple");
617 617 goto bail;
618 618 }
619 619 tuple = (dirstateTupleObject *)v;
620 620
621 621 state = tuple->state;
622 622 mode = tuple->mode;
623 623 size = tuple->size;
624 624 mtime = tuple->mtime;
625 625 if (state == 'n' && mtime == (uint32_t)now) {
626 626 /* See pure/parsers.py:pack_dirstate for why we do
627 627 * this. */
628 628 mtime = -1;
629 629 mtime_unset = (PyObject *)make_dirstate_tuple(
630 630 state, mode, size, mtime);
631 631 if (!mtime_unset)
632 632 goto bail;
633 633 if (PyDict_SetItem(map, k, mtime_unset) == -1)
634 634 goto bail;
635 635 Py_DECREF(mtime_unset);
636 636 mtime_unset = NULL;
637 637 }
638 638 *p++ = state;
639 639 putbe32(mode, p);
640 640 putbe32(size, p + 4);
641 641 putbe32(mtime, p + 8);
642 642 t = p + 12;
643 643 p += 16;
644 644 len = PyString_GET_SIZE(k);
645 645 memcpy(p, PyString_AS_STRING(k), len);
646 646 p += len;
647 647 o = PyDict_GetItem(copymap, k);
648 648 if (o) {
649 649 *p++ = '\0';
650 650 l = PyString_GET_SIZE(o);
651 651 memcpy(p, PyString_AS_STRING(o), l);
652 652 p += l;
653 653 len += l + 1;
654 654 }
655 655 putbe32((uint32_t)len, t);
656 656 }
657 657
658 658 pos = p - PyString_AS_STRING(packobj);
659 659 if (pos != nbytes) {
660 660 PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
661 661 (long)pos, (long)nbytes);
662 662 goto bail;
663 663 }
664 664
665 665 return packobj;
666 666 bail:
667 667 Py_XDECREF(mtime_unset);
668 668 Py_XDECREF(packobj);
669 669 Py_XDECREF(v);
670 670 return NULL;
671 671 }
672 672
673 673 /*
674 674 * A base-16 trie for fast node->rev mapping.
675 675 *
676 676 * Positive value is index of the next node in the trie
677 677 * Negative value is a leaf: -(rev + 1)
678 678 * Zero is empty
679 679 */
680 680 typedef struct {
681 681 int children[16];
682 682 } nodetree;
683 683
684 684 /*
685 685 * This class has two behaviours.
686 686 *
687 687 * When used in a list-like way (with integer keys), we decode an
688 688 * entry in a RevlogNG index file on demand. Our last entry is a
689 689 * sentinel, always a nullid. We have limited support for
690 690 * integer-keyed insert and delete, only at elements right before the
691 691 * sentinel.
692 692 *
693 693 * With string keys, we lazily perform a reverse mapping from node to
694 694 * rev, using a base-16 trie.
695 695 */
696 696 typedef struct {
697 697 PyObject_HEAD
698 698 /* Type-specific fields go here. */
699 699 PyObject *data; /* raw bytes of index */
700 700 PyObject **cache; /* cached tuples */
701 701 const char **offsets; /* populated on demand */
702 702 Py_ssize_t raw_length; /* original number of elements */
703 703 Py_ssize_t length; /* current number of elements */
704 704 PyObject *added; /* populated on demand */
705 705 PyObject *headrevs; /* cache, invalidated on changes */
706 706 PyObject *filteredrevs;/* filtered revs set */
707 707 nodetree *nt; /* base-16 trie */
708 708 int ntlength; /* # nodes in use */
709 709 int ntcapacity; /* # nodes allocated */
710 710 int ntdepth; /* maximum depth of tree */
711 711 int ntsplits; /* # splits performed */
712 712 int ntrev; /* last rev scanned */
713 713 int ntlookups; /* # lookups */
714 714 int ntmisses; /* # lookups that miss the cache */
715 715 int inlined;
716 716 } indexObject;
717 717
718 718 static Py_ssize_t index_length(const indexObject *self)
719 719 {
720 720 if (self->added == NULL)
721 721 return self->length;
722 722 return self->length + PyList_GET_SIZE(self->added);
723 723 }
724 724
725 725 static PyObject *nullentry;
726 726 static const char nullid[20];
727 727
728 728 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
729 729
730 730 #if LONG_MAX == 0x7fffffffL
731 731 static char *tuple_format = "Kiiiiiis#";
732 732 #else
733 733 static char *tuple_format = "kiiiiiis#";
734 734 #endif
735 735
736 736 /* A RevlogNG v1 index entry is 64 bytes long. */
737 737 static const long v1_hdrsize = 64;
738 738
739 739 /*
740 740 * Return a pointer to the beginning of a RevlogNG record.
741 741 */
742 742 static const char *index_deref(indexObject *self, Py_ssize_t pos)
743 743 {
744 744 if (self->inlined && pos > 0) {
745 745 if (self->offsets == NULL) {
746 746 self->offsets = malloc(self->raw_length *
747 747 sizeof(*self->offsets));
748 748 if (self->offsets == NULL)
749 749 return (const char *)PyErr_NoMemory();
750 750 inline_scan(self, self->offsets);
751 751 }
752 752 return self->offsets[pos];
753 753 }
754 754
755 755 return PyString_AS_STRING(self->data) + pos * v1_hdrsize;
756 756 }
757 757
758 758 static inline int index_get_parents(indexObject *self, Py_ssize_t rev,
759 759 int *ps, int maxrev)
760 760 {
761 761 if (rev >= self->length - 1) {
762 762 PyObject *tuple = PyList_GET_ITEM(self->added,
763 763 rev - self->length + 1);
764 764 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
765 765 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
766 766 } else {
767 767 const char *data = index_deref(self, rev);
768 768 ps[0] = getbe32(data + 24);
769 769 ps[1] = getbe32(data + 28);
770 770 }
771 771 /* If index file is corrupted, ps[] may point to invalid revisions. So
772 772 * there is a risk of buffer overflow to trust them unconditionally. */
773 773 if (ps[0] > maxrev || ps[1] > maxrev) {
774 774 PyErr_SetString(PyExc_ValueError, "parent out of range");
775 775 return -1;
776 776 }
777 777 return 0;
778 778 }
779 779
780 780
781 781 /*
782 782 * RevlogNG format (all in big endian, data may be inlined):
783 783 * 6 bytes: offset
784 784 * 2 bytes: flags
785 785 * 4 bytes: compressed length
786 786 * 4 bytes: uncompressed length
787 787 * 4 bytes: base revision
788 788 * 4 bytes: link revision
789 789 * 4 bytes: parent 1 revision
790 790 * 4 bytes: parent 2 revision
791 791 * 32 bytes: nodeid (only 20 bytes used)
792 792 */
793 793 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
794 794 {
795 795 uint64_t offset_flags;
796 796 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
797 797 const char *c_node_id;
798 798 const char *data;
799 799 Py_ssize_t length = index_length(self);
800 800 PyObject *entry;
801 801
802 802 if (pos < 0)
803 803 pos += length;
804 804
805 805 if (pos < 0 || pos >= length) {
806 806 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
807 807 return NULL;
808 808 }
809 809
810 810 if (pos == length - 1) {
811 811 Py_INCREF(nullentry);
812 812 return nullentry;
813 813 }
814 814
815 815 if (pos >= self->length - 1) {
816 816 PyObject *obj;
817 817 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
818 818 Py_INCREF(obj);
819 819 return obj;
820 820 }
821 821
822 822 if (self->cache) {
823 823 if (self->cache[pos]) {
824 824 Py_INCREF(self->cache[pos]);
825 825 return self->cache[pos];
826 826 }
827 827 } else {
828 828 self->cache = calloc(self->raw_length, sizeof(PyObject *));
829 829 if (self->cache == NULL)
830 830 return PyErr_NoMemory();
831 831 }
832 832
833 833 data = index_deref(self, pos);
834 834 if (data == NULL)
835 835 return NULL;
836 836
837 837 offset_flags = getbe32(data + 4);
838 838 if (pos == 0) /* mask out version number for the first entry */
839 839 offset_flags &= 0xFFFF;
840 840 else {
841 841 uint32_t offset_high = getbe32(data);
842 842 offset_flags |= ((uint64_t)offset_high) << 32;
843 843 }
844 844
845 845 comp_len = getbe32(data + 8);
846 846 uncomp_len = getbe32(data + 12);
847 847 base_rev = getbe32(data + 16);
848 848 link_rev = getbe32(data + 20);
849 849 parent_1 = getbe32(data + 24);
850 850 parent_2 = getbe32(data + 28);
851 851 c_node_id = data + 32;
852 852
853 853 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
854 854 uncomp_len, base_rev, link_rev,
855 855 parent_1, parent_2, c_node_id, 20);
856 856
857 857 if (entry) {
858 858 PyObject_GC_UnTrack(entry);
859 859 Py_INCREF(entry);
860 860 }
861 861
862 862 self->cache[pos] = entry;
863 863
864 864 return entry;
865 865 }
866 866
867 867 /*
868 868 * Return the 20-byte SHA of the node corresponding to the given rev.
869 869 */
870 870 static const char *index_node(indexObject *self, Py_ssize_t pos)
871 871 {
872 872 Py_ssize_t length = index_length(self);
873 873 const char *data;
874 874
875 875 if (pos == length - 1 || pos == INT_MAX)
876 876 return nullid;
877 877
878 878 if (pos >= length)
879 879 return NULL;
880 880
881 881 if (pos >= self->length - 1) {
882 882 PyObject *tuple, *str;
883 883 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
884 884 str = PyTuple_GetItem(tuple, 7);
885 885 return str ? PyString_AS_STRING(str) : NULL;
886 886 }
887 887
888 888 data = index_deref(self, pos);
889 889 return data ? data + 32 : NULL;
890 890 }
891 891
892 892 static int nt_insert(indexObject *self, const char *node, int rev);
893 893
894 894 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
895 895 {
896 896 if (PyString_AsStringAndSize(obj, node, nodelen) == -1)
897 897 return -1;
898 898 if (*nodelen == 20)
899 899 return 0;
900 900 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
901 901 return -1;
902 902 }
903 903
904 904 static PyObject *index_insert(indexObject *self, PyObject *args)
905 905 {
906 906 PyObject *obj;
907 907 char *node;
908 908 int index;
909 909 Py_ssize_t len, nodelen;
910 910
911 911 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
912 912 return NULL;
913 913
914 914 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
915 915 PyErr_SetString(PyExc_TypeError, "8-tuple required");
916 916 return NULL;
917 917 }
918 918
919 919 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
920 920 return NULL;
921 921
922 922 len = index_length(self);
923 923
924 924 if (index < 0)
925 925 index += len;
926 926
927 927 if (index != len - 1) {
928 928 PyErr_SetString(PyExc_IndexError,
929 929 "insert only supported at index -1");
930 930 return NULL;
931 931 }
932 932
933 933 if (self->added == NULL) {
934 934 self->added = PyList_New(0);
935 935 if (self->added == NULL)
936 936 return NULL;
937 937 }
938 938
939 939 if (PyList_Append(self->added, obj) == -1)
940 940 return NULL;
941 941
942 942 if (self->nt)
943 943 nt_insert(self, node, index);
944 944
945 945 Py_CLEAR(self->headrevs);
946 946 Py_RETURN_NONE;
947 947 }
948 948
949 949 static void _index_clearcaches(indexObject *self)
950 950 {
951 951 if (self->cache) {
952 952 Py_ssize_t i;
953 953
954 954 for (i = 0; i < self->raw_length; i++)
955 955 Py_CLEAR(self->cache[i]);
956 956 free(self->cache);
957 957 self->cache = NULL;
958 958 }
959 959 if (self->offsets) {
960 960 free(self->offsets);
961 961 self->offsets = NULL;
962 962 }
963 963 if (self->nt) {
964 964 free(self->nt);
965 965 self->nt = NULL;
966 966 }
967 967 Py_CLEAR(self->headrevs);
968 968 }
969 969
970 970 static PyObject *index_clearcaches(indexObject *self)
971 971 {
972 972 _index_clearcaches(self);
973 973 self->ntlength = self->ntcapacity = 0;
974 974 self->ntdepth = self->ntsplits = 0;
975 975 self->ntrev = -1;
976 976 self->ntlookups = self->ntmisses = 0;
977 977 Py_RETURN_NONE;
978 978 }
979 979
980 980 static PyObject *index_stats(indexObject *self)
981 981 {
982 982 PyObject *obj = PyDict_New();
983 983 PyObject *t = NULL;
984 984
985 985 if (obj == NULL)
986 986 return NULL;
987 987
988 988 #define istat(__n, __d) \
989 989 t = PyInt_FromSsize_t(self->__n); \
990 990 if (!t) \
991 991 goto bail; \
992 992 if (PyDict_SetItemString(obj, __d, t) == -1) \
993 993 goto bail; \
994 994 Py_DECREF(t);
995 995
996 996 if (self->added) {
997 997 Py_ssize_t len = PyList_GET_SIZE(self->added);
998 998 t = PyInt_FromSsize_t(len);
999 999 if (!t)
1000 1000 goto bail;
1001 1001 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
1002 1002 goto bail;
1003 1003 Py_DECREF(t);
1004 1004 }
1005 1005
1006 1006 if (self->raw_length != self->length - 1)
1007 1007 istat(raw_length, "revs on disk");
1008 1008 istat(length, "revs in memory");
1009 1009 istat(ntcapacity, "node trie capacity");
1010 1010 istat(ntdepth, "node trie depth");
1011 1011 istat(ntlength, "node trie count");
1012 1012 istat(ntlookups, "node trie lookups");
1013 1013 istat(ntmisses, "node trie misses");
1014 1014 istat(ntrev, "node trie last rev scanned");
1015 1015 istat(ntsplits, "node trie splits");
1016 1016
1017 1017 #undef istat
1018 1018
1019 1019 return obj;
1020 1020
1021 1021 bail:
1022 1022 Py_XDECREF(obj);
1023 1023 Py_XDECREF(t);
1024 1024 return NULL;
1025 1025 }
1026 1026
1027 1027 /*
1028 1028 * When we cache a list, we want to be sure the caller can't mutate
1029 1029 * the cached copy.
1030 1030 */
1031 1031 static PyObject *list_copy(PyObject *list)
1032 1032 {
1033 1033 Py_ssize_t len = PyList_GET_SIZE(list);
1034 1034 PyObject *newlist = PyList_New(len);
1035 1035 Py_ssize_t i;
1036 1036
1037 1037 if (newlist == NULL)
1038 1038 return NULL;
1039 1039
1040 1040 for (i = 0; i < len; i++) {
1041 1041 PyObject *obj = PyList_GET_ITEM(list, i);
1042 1042 Py_INCREF(obj);
1043 1043 PyList_SET_ITEM(newlist, i, obj);
1044 1044 }
1045 1045
1046 1046 return newlist;
1047 1047 }
1048 1048
1049 1049 /* arg should be Py_ssize_t but Python 2.4 do not support the n format */
1050 1050 static int check_filter(PyObject *filter, unsigned long arg) {
1051 1051 if (filter) {
1052 1052 PyObject *arglist, *result;
1053 1053 int isfiltered;
1054 1054
1055 1055 arglist = Py_BuildValue("(k)", arg);
1056 1056 if (!arglist) {
1057 1057 return -1;
1058 1058 }
1059 1059
1060 1060 result = PyEval_CallObject(filter, arglist);
1061 1061 Py_DECREF(arglist);
1062 1062 if (!result) {
1063 1063 return -1;
1064 1064 }
1065 1065
1066 1066 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
1067 1067 * same as this function, so we can just return it directly.*/
1068 1068 isfiltered = PyObject_IsTrue(result);
1069 1069 Py_DECREF(result);
1070 1070 return isfiltered;
1071 1071 } else {
1072 1072 return 0;
1073 1073 }
1074 1074 }
1075 1075
1076 1076 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
1077 1077 Py_ssize_t marker, char *phases)
1078 1078 {
1079 1079 PyObject *iter = NULL;
1080 1080 PyObject *iter_item = NULL;
1081 1081 Py_ssize_t min_idx = index_length(self) + 1;
1082 1082 long iter_item_long;
1083 1083
1084 1084 if (PyList_GET_SIZE(list) != 0) {
1085 1085 iter = PyObject_GetIter(list);
1086 1086 if (iter == NULL)
1087 1087 return -2;
1088 1088 while ((iter_item = PyIter_Next(iter)))
1089 1089 {
1090 1090 iter_item_long = PyInt_AS_LONG(iter_item);
1091 1091 Py_DECREF(iter_item);
1092 1092 if (iter_item_long < min_idx)
1093 1093 min_idx = iter_item_long;
1094 1094 phases[iter_item_long] = marker;
1095 1095 }
1096 1096 Py_DECREF(iter);
1097 1097 }
1098 1098
1099 1099 return min_idx;
1100 1100 }
1101 1101
1102 1102 static inline void set_phase_from_parents(char *phases, int parent_1,
1103 1103 int parent_2, Py_ssize_t i)
1104 1104 {
1105 1105 if (parent_1 >= 0 && phases[parent_1] > phases[i])
1106 1106 phases[i] = phases[parent_1];
1107 1107 if (parent_2 >= 0 && phases[parent_2] > phases[i])
1108 1108 phases[i] = phases[parent_2];
1109 1109 }
1110 1110
1111 1111 static PyObject *reachableroots2(indexObject *self, PyObject *args)
1112 1112 {
1113 1113
1114 1114 /* Input */
1115 1115 long minroot;
1116 1116 PyObject *includepatharg = NULL;
1117 1117 int includepath = 0;
1118 1118 /* heads and roots are lists */
1119 1119 PyObject *heads = NULL;
1120 1120 PyObject *roots = NULL;
1121 1121 PyObject *reachable = NULL;
1122 1122
1123 1123 PyObject *val;
1124 1124 Py_ssize_t len = index_length(self) - 1;
1125 1125 long revnum;
1126 1126 Py_ssize_t k;
1127 1127 Py_ssize_t i;
1128 1128 Py_ssize_t l;
1129 1129 int r;
1130 1130 int minidx;
1131 1131 int parents[2];
1132 1132
1133 1133 /* Internal data structure:
1134 1134 * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit
1135 1135 * revstates: array of length len+1 (all revs + nullrev) */
1136 1136 int *tovisit = NULL;
1137 1137 long lentovisit = 0;
1138 1138 enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 };
1139 1139 char *revstates = NULL;
1140 1140
1141 1141 /* Get arguments */
1142 1142 if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
1143 1143 &PyList_Type, &roots,
1144 1144 &PyBool_Type, &includepatharg))
1145 1145 goto bail;
1146 1146
1147 1147 if (includepatharg == Py_True)
1148 1148 includepath = 1;
1149 1149
1150 1150 /* Initialize return set */
1151 1151 reachable = PyList_New(0);
1152 1152 if (reachable == NULL)
1153 1153 goto bail;
1154 1154
1155 1155 /* Initialize internal datastructures */
1156 1156 tovisit = (int *)malloc((len + 1) * sizeof(int));
1157 1157 if (tovisit == NULL) {
1158 1158 PyErr_NoMemory();
1159 1159 goto bail;
1160 1160 }
1161 1161
1162 1162 revstates = (char *)calloc(len + 1, 1);
1163 1163 if (revstates == NULL) {
1164 1164 PyErr_NoMemory();
1165 1165 goto bail;
1166 1166 }
1167 1167
1168 1168 l = PyList_GET_SIZE(roots);
1169 1169 for (i = 0; i < l; i++) {
1170 1170 revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i));
1171 1171 if (revnum == -1 && PyErr_Occurred())
1172 1172 goto bail;
1173 1173 /* If root is out of range, e.g. wdir(), it must be unreachable
1174 1174 * from heads. So we can just ignore it. */
1175 1175 if (revnum + 1 < 0 || revnum + 1 >= len + 1)
1176 1176 continue;
1177 1177 revstates[revnum + 1] |= RS_ROOT;
1178 1178 }
1179 1179
1180 1180 /* Populate tovisit with all the heads */
1181 1181 l = PyList_GET_SIZE(heads);
1182 1182 for (i = 0; i < l; i++) {
1183 1183 revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
1184 1184 if (revnum == -1 && PyErr_Occurred())
1185 1185 goto bail;
1186 1186 if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
1187 1187 PyErr_SetString(PyExc_IndexError, "head out of range");
1188 1188 goto bail;
1189 1189 }
1190 1190 if (!(revstates[revnum + 1] & RS_SEEN)) {
1191 1191 tovisit[lentovisit++] = revnum;
1192 1192 revstates[revnum + 1] |= RS_SEEN;
1193 1193 }
1194 1194 }
1195 1195
1196 1196 /* Visit the tovisit list and find the reachable roots */
1197 1197 k = 0;
1198 1198 while (k < lentovisit) {
1199 1199 /* Add the node to reachable if it is a root*/
1200 1200 revnum = tovisit[k++];
1201 1201 if (revstates[revnum + 1] & RS_ROOT) {
1202 1202 revstates[revnum + 1] |= RS_REACHABLE;
1203 1203 val = PyInt_FromLong(revnum);
1204 1204 if (val == NULL)
1205 1205 goto bail;
1206 PyList_Append(reachable, val);
1206 r = PyList_Append(reachable, val);
1207 1207 Py_DECREF(val);
1208 if (r < 0)
1209 goto bail;
1208 1210 if (includepath == 0)
1209 1211 continue;
1210 1212 }
1211 1213
1212 1214 /* Add its parents to the list of nodes to visit */
1213 1215 if (revnum == -1)
1214 1216 continue;
1215 1217 r = index_get_parents(self, revnum, parents, (int)len - 1);
1216 1218 if (r < 0)
1217 1219 goto bail;
1218 1220 for (i = 0; i < 2; i++) {
1219 1221 if (!(revstates[parents[i] + 1] & RS_SEEN)
1220 1222 && parents[i] >= minroot) {
1221 1223 tovisit[lentovisit++] = parents[i];
1222 1224 revstates[parents[i] + 1] |= RS_SEEN;
1223 1225 }
1224 1226 }
1225 1227 }
1226 1228
1227 1229 /* Find all the nodes in between the roots we found and the heads
1228 1230 * and add them to the reachable set */
1229 1231 if (includepath == 1) {
1230 1232 minidx = minroot;
1231 1233 if (minidx < 0)
1232 1234 minidx = 0;
1233 1235 for (i = minidx; i < len; i++) {
1234 1236 if (!(revstates[i + 1] & RS_SEEN))
1235 1237 continue;
1236 1238 r = index_get_parents(self, i, parents, (int)len - 1);
1237 1239 /* Corrupted index file, error is set from
1238 1240 * index_get_parents */
1239 1241 if (r < 0)
1240 1242 goto bail;
1241 1243 for (k = 0; k < 2; k++) {
1242 1244 if ((revstates[parents[k] + 1] & RS_REACHABLE)
1243 1245 && !(revstates[i + 1] & RS_REACHABLE)) {
1244 1246 revstates[i + 1] |= RS_REACHABLE;
1245 1247 val = PyInt_FromLong(i);
1246 1248 if (val == NULL)
1247 1249 goto bail;
1248 PyList_Append(reachable, val);
1250 r = PyList_Append(reachable, val);
1249 1251 Py_DECREF(val);
1252 if (r < 0)
1253 goto bail;
1250 1254 }
1251 1255 }
1252 1256 }
1253 1257 }
1254 1258
1255 1259 free(revstates);
1256 1260 free(tovisit);
1257 1261 return reachable;
1258 1262 bail:
1259 1263 Py_XDECREF(reachable);
1260 1264 free(revstates);
1261 1265 free(tovisit);
1262 1266 return NULL;
1263 1267 }
1264 1268
1265 1269 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
1266 1270 {
1267 1271 PyObject *roots = Py_None;
1268 1272 PyObject *ret = NULL;
1269 1273 PyObject *phaseslist = NULL;
1270 1274 PyObject *phaseroots = NULL;
1271 1275 PyObject *phaseset = NULL;
1272 1276 PyObject *phasessetlist = NULL;
1273 1277 PyObject *rev = NULL;
1274 1278 Py_ssize_t len = index_length(self) - 1;
1275 1279 Py_ssize_t numphase = 0;
1276 1280 Py_ssize_t minrevallphases = 0;
1277 1281 Py_ssize_t minrevphase = 0;
1278 1282 Py_ssize_t i = 0;
1279 1283 char *phases = NULL;
1280 1284 long phase;
1281 1285
1282 1286 if (!PyArg_ParseTuple(args, "O", &roots))
1283 1287 goto release_none;
1284 1288 if (roots == NULL || !PyList_Check(roots))
1285 1289 goto release_none;
1286 1290
1287 1291 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
1288 1292 if (phases == NULL)
1289 1293 goto release_none;
1290 1294 /* Put the phase information of all the roots in phases */
1291 1295 numphase = PyList_GET_SIZE(roots)+1;
1292 1296 minrevallphases = len + 1;
1293 1297 phasessetlist = PyList_New(numphase);
1294 1298 if (phasessetlist == NULL)
1295 1299 goto release_none;
1296 1300
1297 1301 PyList_SET_ITEM(phasessetlist, 0, Py_None);
1298 1302 Py_INCREF(Py_None);
1299 1303
1300 1304 for (i = 0; i < numphase-1; i++) {
1301 1305 phaseroots = PyList_GET_ITEM(roots, i);
1302 1306 phaseset = PySet_New(NULL);
1303 1307 if (phaseset == NULL)
1304 1308 goto release_phasesetlist;
1305 1309 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
1306 1310 if (!PyList_Check(phaseroots))
1307 1311 goto release_phasesetlist;
1308 1312 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
1309 1313 if (minrevphase == -2) /* Error from add_roots_get_min */
1310 1314 goto release_phasesetlist;
1311 1315 minrevallphases = MIN(minrevallphases, minrevphase);
1312 1316 }
1313 1317 /* Propagate the phase information from the roots to the revs */
1314 1318 if (minrevallphases != -1) {
1315 1319 int parents[2];
1316 1320 for (i = minrevallphases; i < len; i++) {
1317 1321 if (index_get_parents(self, i, parents,
1318 1322 (int)len - 1) < 0)
1319 1323 goto release_phasesetlist;
1320 1324 set_phase_from_parents(phases, parents[0], parents[1], i);
1321 1325 }
1322 1326 }
1323 1327 /* Transform phase list to a python list */
1324 1328 phaseslist = PyList_New(len);
1325 1329 if (phaseslist == NULL)
1326 1330 goto release_phasesetlist;
1327 1331 for (i = 0; i < len; i++) {
1328 1332 phase = phases[i];
1329 1333 /* We only store the sets of phase for non public phase, the public phase
1330 1334 * is computed as a difference */
1331 1335 if (phase != 0) {
1332 1336 phaseset = PyList_GET_ITEM(phasessetlist, phase);
1333 1337 rev = PyInt_FromLong(i);
1334 1338 PySet_Add(phaseset, rev);
1335 1339 Py_XDECREF(rev);
1336 1340 }
1337 1341 PyList_SET_ITEM(phaseslist, i, PyInt_FromLong(phase));
1338 1342 }
1339 1343 ret = PyList_New(2);
1340 1344 if (ret == NULL)
1341 1345 goto release_phaseslist;
1342 1346
1343 1347 PyList_SET_ITEM(ret, 0, phaseslist);
1344 1348 PyList_SET_ITEM(ret, 1, phasessetlist);
1345 1349 /* We don't release phaseslist and phasessetlist as we return them to
1346 1350 * python */
1347 1351 goto release_phases;
1348 1352
1349 1353 release_phaseslist:
1350 1354 Py_XDECREF(phaseslist);
1351 1355 release_phasesetlist:
1352 1356 Py_XDECREF(phasessetlist);
1353 1357 release_phases:
1354 1358 free(phases);
1355 1359 release_none:
1356 1360 return ret;
1357 1361 }
1358 1362
1359 1363 static PyObject *index_headrevs(indexObject *self, PyObject *args)
1360 1364 {
1361 1365 Py_ssize_t i, j, len;
1362 1366 char *nothead = NULL;
1363 1367 PyObject *heads = NULL;
1364 1368 PyObject *filter = NULL;
1365 1369 PyObject *filteredrevs = Py_None;
1366 1370
1367 1371 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
1368 1372 return NULL;
1369 1373 }
1370 1374
1371 1375 if (self->headrevs && filteredrevs == self->filteredrevs)
1372 1376 return list_copy(self->headrevs);
1373 1377
1374 1378 Py_DECREF(self->filteredrevs);
1375 1379 self->filteredrevs = filteredrevs;
1376 1380 Py_INCREF(filteredrevs);
1377 1381
1378 1382 if (filteredrevs != Py_None) {
1379 1383 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
1380 1384 if (!filter) {
1381 1385 PyErr_SetString(PyExc_TypeError,
1382 1386 "filteredrevs has no attribute __contains__");
1383 1387 goto bail;
1384 1388 }
1385 1389 }
1386 1390
1387 1391 len = index_length(self) - 1;
1388 1392 heads = PyList_New(0);
1389 1393 if (heads == NULL)
1390 1394 goto bail;
1391 1395 if (len == 0) {
1392 1396 PyObject *nullid = PyInt_FromLong(-1);
1393 1397 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
1394 1398 Py_XDECREF(nullid);
1395 1399 goto bail;
1396 1400 }
1397 1401 goto done;
1398 1402 }
1399 1403
1400 1404 nothead = calloc(len, 1);
1401 1405 if (nothead == NULL)
1402 1406 goto bail;
1403 1407
1404 1408 for (i = 0; i < len; i++) {
1405 1409 int isfiltered;
1406 1410 int parents[2];
1407 1411
1408 1412 isfiltered = check_filter(filter, i);
1409 1413 if (isfiltered == -1) {
1410 1414 PyErr_SetString(PyExc_TypeError,
1411 1415 "unable to check filter");
1412 1416 goto bail;
1413 1417 }
1414 1418
1415 1419 if (isfiltered) {
1416 1420 nothead[i] = 1;
1417 1421 continue;
1418 1422 }
1419 1423
1420 1424 if (index_get_parents(self, i, parents, (int)len - 1) < 0)
1421 1425 goto bail;
1422 1426 for (j = 0; j < 2; j++) {
1423 1427 if (parents[j] >= 0)
1424 1428 nothead[parents[j]] = 1;
1425 1429 }
1426 1430 }
1427 1431
1428 1432 for (i = 0; i < len; i++) {
1429 1433 PyObject *head;
1430 1434
1431 1435 if (nothead[i])
1432 1436 continue;
1433 1437 head = PyInt_FromSsize_t(i);
1434 1438 if (head == NULL || PyList_Append(heads, head) == -1) {
1435 1439 Py_XDECREF(head);
1436 1440 goto bail;
1437 1441 }
1438 1442 }
1439 1443
1440 1444 done:
1441 1445 self->headrevs = heads;
1442 1446 Py_XDECREF(filter);
1443 1447 free(nothead);
1444 1448 return list_copy(self->headrevs);
1445 1449 bail:
1446 1450 Py_XDECREF(filter);
1447 1451 Py_XDECREF(heads);
1448 1452 free(nothead);
1449 1453 return NULL;
1450 1454 }
1451 1455
1452 1456 static inline int nt_level(const char *node, Py_ssize_t level)
1453 1457 {
1454 1458 int v = node[level>>1];
1455 1459 if (!(level & 1))
1456 1460 v >>= 4;
1457 1461 return v & 0xf;
1458 1462 }
1459 1463
1460 1464 /*
1461 1465 * Return values:
1462 1466 *
1463 1467 * -4: match is ambiguous (multiple candidates)
1464 1468 * -2: not found
1465 1469 * rest: valid rev
1466 1470 */
1467 1471 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
1468 1472 int hex)
1469 1473 {
1470 1474 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
1471 1475 int level, maxlevel, off;
1472 1476
1473 1477 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
1474 1478 return -1;
1475 1479
1476 1480 if (self->nt == NULL)
1477 1481 return -2;
1478 1482
1479 1483 if (hex)
1480 1484 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
1481 1485 else
1482 1486 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
1483 1487
1484 1488 for (level = off = 0; level < maxlevel; level++) {
1485 1489 int k = getnybble(node, level);
1486 1490 nodetree *n = &self->nt[off];
1487 1491 int v = n->children[k];
1488 1492
1489 1493 if (v < 0) {
1490 1494 const char *n;
1491 1495 Py_ssize_t i;
1492 1496
1493 1497 v = -(v + 1);
1494 1498 n = index_node(self, v);
1495 1499 if (n == NULL)
1496 1500 return -2;
1497 1501 for (i = level; i < maxlevel; i++)
1498 1502 if (getnybble(node, i) != nt_level(n, i))
1499 1503 return -2;
1500 1504 return v;
1501 1505 }
1502 1506 if (v == 0)
1503 1507 return -2;
1504 1508 off = v;
1505 1509 }
1506 1510 /* multiple matches against an ambiguous prefix */
1507 1511 return -4;
1508 1512 }
1509 1513
1510 1514 static int nt_new(indexObject *self)
1511 1515 {
1512 1516 if (self->ntlength == self->ntcapacity) {
1513 1517 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1514 1518 PyErr_SetString(PyExc_MemoryError,
1515 1519 "overflow in nt_new");
1516 1520 return -1;
1517 1521 }
1518 1522 self->ntcapacity *= 2;
1519 1523 self->nt = realloc(self->nt,
1520 1524 self->ntcapacity * sizeof(nodetree));
1521 1525 if (self->nt == NULL) {
1522 1526 PyErr_SetString(PyExc_MemoryError, "out of memory");
1523 1527 return -1;
1524 1528 }
1525 1529 memset(&self->nt[self->ntlength], 0,
1526 1530 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1527 1531 }
1528 1532 return self->ntlength++;
1529 1533 }
1530 1534
1531 1535 static int nt_insert(indexObject *self, const char *node, int rev)
1532 1536 {
1533 1537 int level = 0;
1534 1538 int off = 0;
1535 1539
1536 1540 while (level < 40) {
1537 1541 int k = nt_level(node, level);
1538 1542 nodetree *n;
1539 1543 int v;
1540 1544
1541 1545 n = &self->nt[off];
1542 1546 v = n->children[k];
1543 1547
1544 1548 if (v == 0) {
1545 1549 n->children[k] = -rev - 1;
1546 1550 return 0;
1547 1551 }
1548 1552 if (v < 0) {
1549 1553 const char *oldnode = index_node(self, -(v + 1));
1550 1554 int noff;
1551 1555
1552 1556 if (!oldnode || !memcmp(oldnode, node, 20)) {
1553 1557 n->children[k] = -rev - 1;
1554 1558 return 0;
1555 1559 }
1556 1560 noff = nt_new(self);
1557 1561 if (noff == -1)
1558 1562 return -1;
1559 1563 /* self->nt may have been changed by realloc */
1560 1564 self->nt[off].children[k] = noff;
1561 1565 off = noff;
1562 1566 n = &self->nt[off];
1563 1567 n->children[nt_level(oldnode, ++level)] = v;
1564 1568 if (level > self->ntdepth)
1565 1569 self->ntdepth = level;
1566 1570 self->ntsplits += 1;
1567 1571 } else {
1568 1572 level += 1;
1569 1573 off = v;
1570 1574 }
1571 1575 }
1572 1576
1573 1577 return -1;
1574 1578 }
1575 1579
1576 1580 static int nt_init(indexObject *self)
1577 1581 {
1578 1582 if (self->nt == NULL) {
1579 1583 if (self->raw_length > INT_MAX / sizeof(nodetree)) {
1580 1584 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1581 1585 return -1;
1582 1586 }
1583 1587 self->ntcapacity = self->raw_length < 4
1584 1588 ? 4 : (int)self->raw_length / 2;
1585 1589
1586 1590 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1587 1591 if (self->nt == NULL) {
1588 1592 PyErr_NoMemory();
1589 1593 return -1;
1590 1594 }
1591 1595 self->ntlength = 1;
1592 1596 self->ntrev = (int)index_length(self) - 1;
1593 1597 self->ntlookups = 1;
1594 1598 self->ntmisses = 0;
1595 1599 if (nt_insert(self, nullid, INT_MAX) == -1)
1596 1600 return -1;
1597 1601 }
1598 1602 return 0;
1599 1603 }
1600 1604
1601 1605 /*
1602 1606 * Return values:
1603 1607 *
1604 1608 * -3: error (exception set)
1605 1609 * -2: not found (no exception set)
1606 1610 * rest: valid rev
1607 1611 */
1608 1612 static int index_find_node(indexObject *self,
1609 1613 const char *node, Py_ssize_t nodelen)
1610 1614 {
1611 1615 int rev;
1612 1616
1613 1617 self->ntlookups++;
1614 1618 rev = nt_find(self, node, nodelen, 0);
1615 1619 if (rev >= -1)
1616 1620 return rev;
1617 1621
1618 1622 if (nt_init(self) == -1)
1619 1623 return -3;
1620 1624
1621 1625 /*
1622 1626 * For the first handful of lookups, we scan the entire index,
1623 1627 * and cache only the matching nodes. This optimizes for cases
1624 1628 * like "hg tip", where only a few nodes are accessed.
1625 1629 *
1626 1630 * After that, we cache every node we visit, using a single
1627 1631 * scan amortized over multiple lookups. This gives the best
1628 1632 * bulk performance, e.g. for "hg log".
1629 1633 */
1630 1634 if (self->ntmisses++ < 4) {
1631 1635 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1632 1636 const char *n = index_node(self, rev);
1633 1637 if (n == NULL)
1634 1638 return -2;
1635 1639 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1636 1640 if (nt_insert(self, n, rev) == -1)
1637 1641 return -3;
1638 1642 break;
1639 1643 }
1640 1644 }
1641 1645 } else {
1642 1646 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1643 1647 const char *n = index_node(self, rev);
1644 1648 if (n == NULL) {
1645 1649 self->ntrev = rev + 1;
1646 1650 return -2;
1647 1651 }
1648 1652 if (nt_insert(self, n, rev) == -1) {
1649 1653 self->ntrev = rev + 1;
1650 1654 return -3;
1651 1655 }
1652 1656 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1653 1657 break;
1654 1658 }
1655 1659 }
1656 1660 self->ntrev = rev;
1657 1661 }
1658 1662
1659 1663 if (rev >= 0)
1660 1664 return rev;
1661 1665 return -2;
1662 1666 }
1663 1667
1664 1668 static void raise_revlog_error(void)
1665 1669 {
1666 1670 PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
1667 1671
1668 1672 mod = PyImport_ImportModule("mercurial.error");
1669 1673 if (mod == NULL) {
1670 1674 goto cleanup;
1671 1675 }
1672 1676
1673 1677 dict = PyModule_GetDict(mod);
1674 1678 if (dict == NULL) {
1675 1679 goto cleanup;
1676 1680 }
1677 1681 Py_INCREF(dict);
1678 1682
1679 1683 errclass = PyDict_GetItemString(dict, "RevlogError");
1680 1684 if (errclass == NULL) {
1681 1685 PyErr_SetString(PyExc_SystemError,
1682 1686 "could not find RevlogError");
1683 1687 goto cleanup;
1684 1688 }
1685 1689
1686 1690 /* value of exception is ignored by callers */
1687 1691 PyErr_SetString(errclass, "RevlogError");
1688 1692
1689 1693 cleanup:
1690 1694 Py_XDECREF(dict);
1691 1695 Py_XDECREF(mod);
1692 1696 }
1693 1697
1694 1698 static PyObject *index_getitem(indexObject *self, PyObject *value)
1695 1699 {
1696 1700 char *node;
1697 1701 Py_ssize_t nodelen;
1698 1702 int rev;
1699 1703
1700 1704 if (PyInt_Check(value))
1701 1705 return index_get(self, PyInt_AS_LONG(value));
1702 1706
1703 1707 if (node_check(value, &node, &nodelen) == -1)
1704 1708 return NULL;
1705 1709 rev = index_find_node(self, node, nodelen);
1706 1710 if (rev >= -1)
1707 1711 return PyInt_FromLong(rev);
1708 1712 if (rev == -2)
1709 1713 raise_revlog_error();
1710 1714 return NULL;
1711 1715 }
1712 1716
1713 1717 static int nt_partialmatch(indexObject *self, const char *node,
1714 1718 Py_ssize_t nodelen)
1715 1719 {
1716 1720 int rev;
1717 1721
1718 1722 if (nt_init(self) == -1)
1719 1723 return -3;
1720 1724
1721 1725 if (self->ntrev > 0) {
1722 1726 /* ensure that the radix tree is fully populated */
1723 1727 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1724 1728 const char *n = index_node(self, rev);
1725 1729 if (n == NULL)
1726 1730 return -2;
1727 1731 if (nt_insert(self, n, rev) == -1)
1728 1732 return -3;
1729 1733 }
1730 1734 self->ntrev = rev;
1731 1735 }
1732 1736
1733 1737 return nt_find(self, node, nodelen, 1);
1734 1738 }
1735 1739
1736 1740 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1737 1741 {
1738 1742 const char *fullnode;
1739 1743 int nodelen;
1740 1744 char *node;
1741 1745 int rev, i;
1742 1746
1743 1747 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1744 1748 return NULL;
1745 1749
1746 1750 if (nodelen < 4) {
1747 1751 PyErr_SetString(PyExc_ValueError, "key too short");
1748 1752 return NULL;
1749 1753 }
1750 1754
1751 1755 if (nodelen > 40) {
1752 1756 PyErr_SetString(PyExc_ValueError, "key too long");
1753 1757 return NULL;
1754 1758 }
1755 1759
1756 1760 for (i = 0; i < nodelen; i++)
1757 1761 hexdigit(node, i);
1758 1762 if (PyErr_Occurred()) {
1759 1763 /* input contains non-hex characters */
1760 1764 PyErr_Clear();
1761 1765 Py_RETURN_NONE;
1762 1766 }
1763 1767
1764 1768 rev = nt_partialmatch(self, node, nodelen);
1765 1769
1766 1770 switch (rev) {
1767 1771 case -4:
1768 1772 raise_revlog_error();
1769 1773 case -3:
1770 1774 return NULL;
1771 1775 case -2:
1772 1776 Py_RETURN_NONE;
1773 1777 case -1:
1774 1778 return PyString_FromStringAndSize(nullid, 20);
1775 1779 }
1776 1780
1777 1781 fullnode = index_node(self, rev);
1778 1782 if (fullnode == NULL) {
1779 1783 PyErr_Format(PyExc_IndexError,
1780 1784 "could not access rev %d", rev);
1781 1785 return NULL;
1782 1786 }
1783 1787 return PyString_FromStringAndSize(fullnode, 20);
1784 1788 }
1785 1789
1786 1790 static PyObject *index_m_get(indexObject *self, PyObject *args)
1787 1791 {
1788 1792 Py_ssize_t nodelen;
1789 1793 PyObject *val;
1790 1794 char *node;
1791 1795 int rev;
1792 1796
1793 1797 if (!PyArg_ParseTuple(args, "O", &val))
1794 1798 return NULL;
1795 1799 if (node_check(val, &node, &nodelen) == -1)
1796 1800 return NULL;
1797 1801 rev = index_find_node(self, node, nodelen);
1798 1802 if (rev == -3)
1799 1803 return NULL;
1800 1804 if (rev == -2)
1801 1805 Py_RETURN_NONE;
1802 1806 return PyInt_FromLong(rev);
1803 1807 }
1804 1808
1805 1809 static int index_contains(indexObject *self, PyObject *value)
1806 1810 {
1807 1811 char *node;
1808 1812 Py_ssize_t nodelen;
1809 1813
1810 1814 if (PyInt_Check(value)) {
1811 1815 long rev = PyInt_AS_LONG(value);
1812 1816 return rev >= -1 && rev < index_length(self);
1813 1817 }
1814 1818
1815 1819 if (node_check(value, &node, &nodelen) == -1)
1816 1820 return -1;
1817 1821
1818 1822 switch (index_find_node(self, node, nodelen)) {
1819 1823 case -3:
1820 1824 return -1;
1821 1825 case -2:
1822 1826 return 0;
1823 1827 default:
1824 1828 return 1;
1825 1829 }
1826 1830 }
1827 1831
1828 1832 typedef uint64_t bitmask;
1829 1833
1830 1834 /*
1831 1835 * Given a disjoint set of revs, return all candidates for the
1832 1836 * greatest common ancestor. In revset notation, this is the set
1833 1837 * "heads(::a and ::b and ...)"
1834 1838 */
1835 1839 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1836 1840 int revcount)
1837 1841 {
1838 1842 const bitmask allseen = (1ull << revcount) - 1;
1839 1843 const bitmask poison = 1ull << revcount;
1840 1844 PyObject *gca = PyList_New(0);
1841 1845 int i, v, interesting;
1842 1846 int maxrev = -1;
1843 1847 bitmask sp;
1844 1848 bitmask *seen;
1845 1849
1846 1850 if (gca == NULL)
1847 1851 return PyErr_NoMemory();
1848 1852
1849 1853 for (i = 0; i < revcount; i++) {
1850 1854 if (revs[i] > maxrev)
1851 1855 maxrev = revs[i];
1852 1856 }
1853 1857
1854 1858 seen = calloc(sizeof(*seen), maxrev + 1);
1855 1859 if (seen == NULL) {
1856 1860 Py_DECREF(gca);
1857 1861 return PyErr_NoMemory();
1858 1862 }
1859 1863
1860 1864 for (i = 0; i < revcount; i++)
1861 1865 seen[revs[i]] = 1ull << i;
1862 1866
1863 1867 interesting = revcount;
1864 1868
1865 1869 for (v = maxrev; v >= 0 && interesting; v--) {
1866 1870 bitmask sv = seen[v];
1867 1871 int parents[2];
1868 1872
1869 1873 if (!sv)
1870 1874 continue;
1871 1875
1872 1876 if (sv < poison) {
1873 1877 interesting -= 1;
1874 1878 if (sv == allseen) {
1875 1879 PyObject *obj = PyInt_FromLong(v);
1876 1880 if (obj == NULL)
1877 1881 goto bail;
1878 1882 if (PyList_Append(gca, obj) == -1) {
1879 1883 Py_DECREF(obj);
1880 1884 goto bail;
1881 1885 }
1882 1886 sv |= poison;
1883 1887 for (i = 0; i < revcount; i++) {
1884 1888 if (revs[i] == v)
1885 1889 goto done;
1886 1890 }
1887 1891 }
1888 1892 }
1889 1893 if (index_get_parents(self, v, parents, maxrev) < 0)
1890 1894 goto bail;
1891 1895
1892 1896 for (i = 0; i < 2; i++) {
1893 1897 int p = parents[i];
1894 1898 if (p == -1)
1895 1899 continue;
1896 1900 sp = seen[p];
1897 1901 if (sv < poison) {
1898 1902 if (sp == 0) {
1899 1903 seen[p] = sv;
1900 1904 interesting++;
1901 1905 }
1902 1906 else if (sp != sv)
1903 1907 seen[p] |= sv;
1904 1908 } else {
1905 1909 if (sp && sp < poison)
1906 1910 interesting--;
1907 1911 seen[p] = sv;
1908 1912 }
1909 1913 }
1910 1914 }
1911 1915
1912 1916 done:
1913 1917 free(seen);
1914 1918 return gca;
1915 1919 bail:
1916 1920 free(seen);
1917 1921 Py_XDECREF(gca);
1918 1922 return NULL;
1919 1923 }
1920 1924
1921 1925 /*
1922 1926 * Given a disjoint set of revs, return the subset with the longest
1923 1927 * path to the root.
1924 1928 */
1925 1929 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1926 1930 {
1927 1931 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1928 1932 static const Py_ssize_t capacity = 24;
1929 1933 int *depth, *interesting = NULL;
1930 1934 int i, j, v, ninteresting;
1931 1935 PyObject *dict = NULL, *keys = NULL;
1932 1936 long *seen = NULL;
1933 1937 int maxrev = -1;
1934 1938 long final;
1935 1939
1936 1940 if (revcount > capacity) {
1937 1941 PyErr_Format(PyExc_OverflowError,
1938 1942 "bitset size (%ld) > capacity (%ld)",
1939 1943 (long)revcount, (long)capacity);
1940 1944 return NULL;
1941 1945 }
1942 1946
1943 1947 for (i = 0; i < revcount; i++) {
1944 1948 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1945 1949 if (n > maxrev)
1946 1950 maxrev = n;
1947 1951 }
1948 1952
1949 1953 depth = calloc(sizeof(*depth), maxrev + 1);
1950 1954 if (depth == NULL)
1951 1955 return PyErr_NoMemory();
1952 1956
1953 1957 seen = calloc(sizeof(*seen), maxrev + 1);
1954 1958 if (seen == NULL) {
1955 1959 PyErr_NoMemory();
1956 1960 goto bail;
1957 1961 }
1958 1962
1959 1963 interesting = calloc(sizeof(*interesting), 2 << revcount);
1960 1964 if (interesting == NULL) {
1961 1965 PyErr_NoMemory();
1962 1966 goto bail;
1963 1967 }
1964 1968
1965 1969 if (PyList_Sort(revs) == -1)
1966 1970 goto bail;
1967 1971
1968 1972 for (i = 0; i < revcount; i++) {
1969 1973 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1970 1974 long b = 1l << i;
1971 1975 depth[n] = 1;
1972 1976 seen[n] = b;
1973 1977 interesting[b] = 1;
1974 1978 }
1975 1979
1976 1980 ninteresting = (int)revcount;
1977 1981
1978 1982 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
1979 1983 int dv = depth[v];
1980 1984 int parents[2];
1981 1985 long sv;
1982 1986
1983 1987 if (dv == 0)
1984 1988 continue;
1985 1989
1986 1990 sv = seen[v];
1987 1991 if (index_get_parents(self, v, parents, maxrev) < 0)
1988 1992 goto bail;
1989 1993
1990 1994 for (i = 0; i < 2; i++) {
1991 1995 int p = parents[i];
1992 1996 long nsp, sp;
1993 1997 int dp;
1994 1998
1995 1999 if (p == -1)
1996 2000 continue;
1997 2001
1998 2002 dp = depth[p];
1999 2003 nsp = sp = seen[p];
2000 2004 if (dp <= dv) {
2001 2005 depth[p] = dv + 1;
2002 2006 if (sp != sv) {
2003 2007 interesting[sv] += 1;
2004 2008 nsp = seen[p] = sv;
2005 2009 if (sp) {
2006 2010 interesting[sp] -= 1;
2007 2011 if (interesting[sp] == 0)
2008 2012 ninteresting -= 1;
2009 2013 }
2010 2014 }
2011 2015 }
2012 2016 else if (dv == dp - 1) {
2013 2017 nsp = sp | sv;
2014 2018 if (nsp == sp)
2015 2019 continue;
2016 2020 seen[p] = nsp;
2017 2021 interesting[sp] -= 1;
2018 2022 if (interesting[sp] == 0 && interesting[nsp] > 0)
2019 2023 ninteresting -= 1;
2020 2024 interesting[nsp] += 1;
2021 2025 }
2022 2026 }
2023 2027 interesting[sv] -= 1;
2024 2028 if (interesting[sv] == 0)
2025 2029 ninteresting -= 1;
2026 2030 }
2027 2031
2028 2032 final = 0;
2029 2033 j = ninteresting;
2030 2034 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
2031 2035 if (interesting[i] == 0)
2032 2036 continue;
2033 2037 final |= i;
2034 2038 j -= 1;
2035 2039 }
2036 2040 if (final == 0) {
2037 2041 keys = PyList_New(0);
2038 2042 goto bail;
2039 2043 }
2040 2044
2041 2045 dict = PyDict_New();
2042 2046 if (dict == NULL)
2043 2047 goto bail;
2044 2048
2045 2049 for (i = 0; i < revcount; i++) {
2046 2050 PyObject *key;
2047 2051
2048 2052 if ((final & (1 << i)) == 0)
2049 2053 continue;
2050 2054
2051 2055 key = PyList_GET_ITEM(revs, i);
2052 2056 Py_INCREF(key);
2053 2057 Py_INCREF(Py_None);
2054 2058 if (PyDict_SetItem(dict, key, Py_None) == -1) {
2055 2059 Py_DECREF(key);
2056 2060 Py_DECREF(Py_None);
2057 2061 goto bail;
2058 2062 }
2059 2063 }
2060 2064
2061 2065 keys = PyDict_Keys(dict);
2062 2066
2063 2067 bail:
2064 2068 free(depth);
2065 2069 free(seen);
2066 2070 free(interesting);
2067 2071 Py_XDECREF(dict);
2068 2072
2069 2073 return keys;
2070 2074 }
2071 2075
2072 2076 /*
2073 2077 * Given a (possibly overlapping) set of revs, return all the
2074 2078 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
2075 2079 */
2076 2080 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
2077 2081 {
2078 2082 PyObject *ret = NULL;
2079 2083 Py_ssize_t argcount, i, len;
2080 2084 bitmask repeat = 0;
2081 2085 int revcount = 0;
2082 2086 int *revs;
2083 2087
2084 2088 argcount = PySequence_Length(args);
2085 2089 revs = malloc(argcount * sizeof(*revs));
2086 2090 if (argcount > 0 && revs == NULL)
2087 2091 return PyErr_NoMemory();
2088 2092 len = index_length(self) - 1;
2089 2093
2090 2094 for (i = 0; i < argcount; i++) {
2091 2095 static const int capacity = 24;
2092 2096 PyObject *obj = PySequence_GetItem(args, i);
2093 2097 bitmask x;
2094 2098 long val;
2095 2099
2096 2100 if (!PyInt_Check(obj)) {
2097 2101 PyErr_SetString(PyExc_TypeError,
2098 2102 "arguments must all be ints");
2099 2103 Py_DECREF(obj);
2100 2104 goto bail;
2101 2105 }
2102 2106 val = PyInt_AsLong(obj);
2103 2107 Py_DECREF(obj);
2104 2108 if (val == -1) {
2105 2109 ret = PyList_New(0);
2106 2110 goto done;
2107 2111 }
2108 2112 if (val < 0 || val >= len) {
2109 2113 PyErr_SetString(PyExc_IndexError,
2110 2114 "index out of range");
2111 2115 goto bail;
2112 2116 }
2113 2117 /* this cheesy bloom filter lets us avoid some more
2114 2118 * expensive duplicate checks in the common set-is-disjoint
2115 2119 * case */
2116 2120 x = 1ull << (val & 0x3f);
2117 2121 if (repeat & x) {
2118 2122 int k;
2119 2123 for (k = 0; k < revcount; k++) {
2120 2124 if (val == revs[k])
2121 2125 goto duplicate;
2122 2126 }
2123 2127 }
2124 2128 else repeat |= x;
2125 2129 if (revcount >= capacity) {
2126 2130 PyErr_Format(PyExc_OverflowError,
2127 2131 "bitset size (%d) > capacity (%d)",
2128 2132 revcount, capacity);
2129 2133 goto bail;
2130 2134 }
2131 2135 revs[revcount++] = (int)val;
2132 2136 duplicate:;
2133 2137 }
2134 2138
2135 2139 if (revcount == 0) {
2136 2140 ret = PyList_New(0);
2137 2141 goto done;
2138 2142 }
2139 2143 if (revcount == 1) {
2140 2144 PyObject *obj;
2141 2145 ret = PyList_New(1);
2142 2146 if (ret == NULL)
2143 2147 goto bail;
2144 2148 obj = PyInt_FromLong(revs[0]);
2145 2149 if (obj == NULL)
2146 2150 goto bail;
2147 2151 PyList_SET_ITEM(ret, 0, obj);
2148 2152 goto done;
2149 2153 }
2150 2154
2151 2155 ret = find_gca_candidates(self, revs, revcount);
2152 2156 if (ret == NULL)
2153 2157 goto bail;
2154 2158
2155 2159 done:
2156 2160 free(revs);
2157 2161 return ret;
2158 2162
2159 2163 bail:
2160 2164 free(revs);
2161 2165 Py_XDECREF(ret);
2162 2166 return NULL;
2163 2167 }
2164 2168
2165 2169 /*
2166 2170 * Given a (possibly overlapping) set of revs, return the greatest
2167 2171 * common ancestors: those with the longest path to the root.
2168 2172 */
2169 2173 static PyObject *index_ancestors(indexObject *self, PyObject *args)
2170 2174 {
2171 2175 PyObject *ret;
2172 2176 PyObject *gca = index_commonancestorsheads(self, args);
2173 2177 if (gca == NULL)
2174 2178 return NULL;
2175 2179
2176 2180 if (PyList_GET_SIZE(gca) <= 1) {
2177 2181 return gca;
2178 2182 }
2179 2183
2180 2184 ret = find_deepest(self, gca);
2181 2185 Py_DECREF(gca);
2182 2186 return ret;
2183 2187 }
2184 2188
2185 2189 /*
2186 2190 * Invalidate any trie entries introduced by added revs.
2187 2191 */
2188 2192 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
2189 2193 {
2190 2194 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
2191 2195
2192 2196 for (i = start; i < len; i++) {
2193 2197 PyObject *tuple = PyList_GET_ITEM(self->added, i);
2194 2198 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
2195 2199
2196 2200 nt_insert(self, PyString_AS_STRING(node), -1);
2197 2201 }
2198 2202
2199 2203 if (start == 0)
2200 2204 Py_CLEAR(self->added);
2201 2205 }
2202 2206
2203 2207 /*
2204 2208 * Delete a numeric range of revs, which must be at the end of the
2205 2209 * range, but exclude the sentinel nullid entry.
2206 2210 */
2207 2211 static int index_slice_del(indexObject *self, PyObject *item)
2208 2212 {
2209 2213 Py_ssize_t start, stop, step, slicelength;
2210 2214 Py_ssize_t length = index_length(self);
2211 2215 int ret = 0;
2212 2216
2213 2217 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
2214 2218 &start, &stop, &step, &slicelength) < 0)
2215 2219 return -1;
2216 2220
2217 2221 if (slicelength <= 0)
2218 2222 return 0;
2219 2223
2220 2224 if ((step < 0 && start < stop) || (step > 0 && start > stop))
2221 2225 stop = start;
2222 2226
2223 2227 if (step < 0) {
2224 2228 stop = start + 1;
2225 2229 start = stop + step*(slicelength - 1) - 1;
2226 2230 step = -step;
2227 2231 }
2228 2232
2229 2233 if (step != 1) {
2230 2234 PyErr_SetString(PyExc_ValueError,
2231 2235 "revlog index delete requires step size of 1");
2232 2236 return -1;
2233 2237 }
2234 2238
2235 2239 if (stop != length - 1) {
2236 2240 PyErr_SetString(PyExc_IndexError,
2237 2241 "revlog index deletion indices are invalid");
2238 2242 return -1;
2239 2243 }
2240 2244
2241 2245 if (start < self->length - 1) {
2242 2246 if (self->nt) {
2243 2247 Py_ssize_t i;
2244 2248
2245 2249 for (i = start + 1; i < self->length - 1; i++) {
2246 2250 const char *node = index_node(self, i);
2247 2251
2248 2252 if (node)
2249 2253 nt_insert(self, node, -1);
2250 2254 }
2251 2255 if (self->added)
2252 2256 nt_invalidate_added(self, 0);
2253 2257 if (self->ntrev > start)
2254 2258 self->ntrev = (int)start;
2255 2259 }
2256 2260 self->length = start + 1;
2257 2261 if (start < self->raw_length) {
2258 2262 if (self->cache) {
2259 2263 Py_ssize_t i;
2260 2264 for (i = start; i < self->raw_length; i++)
2261 2265 Py_CLEAR(self->cache[i]);
2262 2266 }
2263 2267 self->raw_length = start;
2264 2268 }
2265 2269 goto done;
2266 2270 }
2267 2271
2268 2272 if (self->nt) {
2269 2273 nt_invalidate_added(self, start - self->length + 1);
2270 2274 if (self->ntrev > start)
2271 2275 self->ntrev = (int)start;
2272 2276 }
2273 2277 if (self->added)
2274 2278 ret = PyList_SetSlice(self->added, start - self->length + 1,
2275 2279 PyList_GET_SIZE(self->added), NULL);
2276 2280 done:
2277 2281 Py_CLEAR(self->headrevs);
2278 2282 return ret;
2279 2283 }
2280 2284
2281 2285 /*
2282 2286 * Supported ops:
2283 2287 *
2284 2288 * slice deletion
2285 2289 * string assignment (extend node->rev mapping)
2286 2290 * string deletion (shrink node->rev mapping)
2287 2291 */
2288 2292 static int index_assign_subscript(indexObject *self, PyObject *item,
2289 2293 PyObject *value)
2290 2294 {
2291 2295 char *node;
2292 2296 Py_ssize_t nodelen;
2293 2297 long rev;
2294 2298
2295 2299 if (PySlice_Check(item) && value == NULL)
2296 2300 return index_slice_del(self, item);
2297 2301
2298 2302 if (node_check(item, &node, &nodelen) == -1)
2299 2303 return -1;
2300 2304
2301 2305 if (value == NULL)
2302 2306 return self->nt ? nt_insert(self, node, -1) : 0;
2303 2307 rev = PyInt_AsLong(value);
2304 2308 if (rev > INT_MAX || rev < 0) {
2305 2309 if (!PyErr_Occurred())
2306 2310 PyErr_SetString(PyExc_ValueError, "rev out of range");
2307 2311 return -1;
2308 2312 }
2309 2313
2310 2314 if (nt_init(self) == -1)
2311 2315 return -1;
2312 2316 return nt_insert(self, node, (int)rev);
2313 2317 }
2314 2318
2315 2319 /*
2316 2320 * Find all RevlogNG entries in an index that has inline data. Update
2317 2321 * the optional "offsets" table with those entries.
2318 2322 */
2319 2323 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
2320 2324 {
2321 2325 const char *data = PyString_AS_STRING(self->data);
2322 2326 Py_ssize_t pos = 0;
2323 2327 Py_ssize_t end = PyString_GET_SIZE(self->data);
2324 2328 long incr = v1_hdrsize;
2325 2329 Py_ssize_t len = 0;
2326 2330
2327 2331 while (pos + v1_hdrsize <= end && pos >= 0) {
2328 2332 uint32_t comp_len;
2329 2333 /* 3rd element of header is length of compressed inline data */
2330 2334 comp_len = getbe32(data + pos + 8);
2331 2335 incr = v1_hdrsize + comp_len;
2332 2336 if (offsets)
2333 2337 offsets[len] = data + pos;
2334 2338 len++;
2335 2339 pos += incr;
2336 2340 }
2337 2341
2338 2342 if (pos != end) {
2339 2343 if (!PyErr_Occurred())
2340 2344 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2341 2345 return -1;
2342 2346 }
2343 2347
2344 2348 return len;
2345 2349 }
2346 2350
2347 2351 static int index_init(indexObject *self, PyObject *args)
2348 2352 {
2349 2353 PyObject *data_obj, *inlined_obj;
2350 2354 Py_ssize_t size;
2351 2355
2352 2356 /* Initialize before argument-checking to avoid index_dealloc() crash. */
2353 2357 self->raw_length = 0;
2354 2358 self->added = NULL;
2355 2359 self->cache = NULL;
2356 2360 self->data = NULL;
2357 2361 self->headrevs = NULL;
2358 2362 self->filteredrevs = Py_None;
2359 2363 Py_INCREF(Py_None);
2360 2364 self->nt = NULL;
2361 2365 self->offsets = NULL;
2362 2366
2363 2367 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
2364 2368 return -1;
2365 2369 if (!PyString_Check(data_obj)) {
2366 2370 PyErr_SetString(PyExc_TypeError, "data is not a string");
2367 2371 return -1;
2368 2372 }
2369 2373 size = PyString_GET_SIZE(data_obj);
2370 2374
2371 2375 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
2372 2376 self->data = data_obj;
2373 2377
2374 2378 self->ntlength = self->ntcapacity = 0;
2375 2379 self->ntdepth = self->ntsplits = 0;
2376 2380 self->ntlookups = self->ntmisses = 0;
2377 2381 self->ntrev = -1;
2378 2382 Py_INCREF(self->data);
2379 2383
2380 2384 if (self->inlined) {
2381 2385 Py_ssize_t len = inline_scan(self, NULL);
2382 2386 if (len == -1)
2383 2387 goto bail;
2384 2388 self->raw_length = len;
2385 2389 self->length = len + 1;
2386 2390 } else {
2387 2391 if (size % v1_hdrsize) {
2388 2392 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2389 2393 goto bail;
2390 2394 }
2391 2395 self->raw_length = size / v1_hdrsize;
2392 2396 self->length = self->raw_length + 1;
2393 2397 }
2394 2398
2395 2399 return 0;
2396 2400 bail:
2397 2401 return -1;
2398 2402 }
2399 2403
2400 2404 static PyObject *index_nodemap(indexObject *self)
2401 2405 {
2402 2406 Py_INCREF(self);
2403 2407 return (PyObject *)self;
2404 2408 }
2405 2409
2406 2410 static void index_dealloc(indexObject *self)
2407 2411 {
2408 2412 _index_clearcaches(self);
2409 2413 Py_XDECREF(self->filteredrevs);
2410 2414 Py_XDECREF(self->data);
2411 2415 Py_XDECREF(self->added);
2412 2416 PyObject_Del(self);
2413 2417 }
2414 2418
2415 2419 static PySequenceMethods index_sequence_methods = {
2416 2420 (lenfunc)index_length, /* sq_length */
2417 2421 0, /* sq_concat */
2418 2422 0, /* sq_repeat */
2419 2423 (ssizeargfunc)index_get, /* sq_item */
2420 2424 0, /* sq_slice */
2421 2425 0, /* sq_ass_item */
2422 2426 0, /* sq_ass_slice */
2423 2427 (objobjproc)index_contains, /* sq_contains */
2424 2428 };
2425 2429
2426 2430 static PyMappingMethods index_mapping_methods = {
2427 2431 (lenfunc)index_length, /* mp_length */
2428 2432 (binaryfunc)index_getitem, /* mp_subscript */
2429 2433 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
2430 2434 };
2431 2435
2432 2436 static PyMethodDef index_methods[] = {
2433 2437 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
2434 2438 "return the gca set of the given revs"},
2435 2439 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
2436 2440 METH_VARARGS,
2437 2441 "return the heads of the common ancestors of the given revs"},
2438 2442 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
2439 2443 "clear the index caches"},
2440 2444 {"get", (PyCFunction)index_m_get, METH_VARARGS,
2441 2445 "get an index entry"},
2442 2446 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
2443 2447 METH_VARARGS, "compute phases"},
2444 2448 {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS,
2445 2449 "reachableroots"},
2446 2450 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
2447 2451 "get head revisions"}, /* Can do filtering since 3.2 */
2448 2452 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
2449 2453 "get filtered head revisions"}, /* Can always do filtering */
2450 2454 {"insert", (PyCFunction)index_insert, METH_VARARGS,
2451 2455 "insert an index entry"},
2452 2456 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
2453 2457 "match a potentially ambiguous node ID"},
2454 2458 {"stats", (PyCFunction)index_stats, METH_NOARGS,
2455 2459 "stats for the index"},
2456 2460 {NULL} /* Sentinel */
2457 2461 };
2458 2462
2459 2463 static PyGetSetDef index_getset[] = {
2460 2464 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
2461 2465 {NULL} /* Sentinel */
2462 2466 };
2463 2467
2464 2468 static PyTypeObject indexType = {
2465 2469 PyObject_HEAD_INIT(NULL)
2466 2470 0, /* ob_size */
2467 2471 "parsers.index", /* tp_name */
2468 2472 sizeof(indexObject), /* tp_basicsize */
2469 2473 0, /* tp_itemsize */
2470 2474 (destructor)index_dealloc, /* tp_dealloc */
2471 2475 0, /* tp_print */
2472 2476 0, /* tp_getattr */
2473 2477 0, /* tp_setattr */
2474 2478 0, /* tp_compare */
2475 2479 0, /* tp_repr */
2476 2480 0, /* tp_as_number */
2477 2481 &index_sequence_methods, /* tp_as_sequence */
2478 2482 &index_mapping_methods, /* tp_as_mapping */
2479 2483 0, /* tp_hash */
2480 2484 0, /* tp_call */
2481 2485 0, /* tp_str */
2482 2486 0, /* tp_getattro */
2483 2487 0, /* tp_setattro */
2484 2488 0, /* tp_as_buffer */
2485 2489 Py_TPFLAGS_DEFAULT, /* tp_flags */
2486 2490 "revlog index", /* tp_doc */
2487 2491 0, /* tp_traverse */
2488 2492 0, /* tp_clear */
2489 2493 0, /* tp_richcompare */
2490 2494 0, /* tp_weaklistoffset */
2491 2495 0, /* tp_iter */
2492 2496 0, /* tp_iternext */
2493 2497 index_methods, /* tp_methods */
2494 2498 0, /* tp_members */
2495 2499 index_getset, /* tp_getset */
2496 2500 0, /* tp_base */
2497 2501 0, /* tp_dict */
2498 2502 0, /* tp_descr_get */
2499 2503 0, /* tp_descr_set */
2500 2504 0, /* tp_dictoffset */
2501 2505 (initproc)index_init, /* tp_init */
2502 2506 0, /* tp_alloc */
2503 2507 };
2504 2508
2505 2509 /*
2506 2510 * returns a tuple of the form (index, index, cache) with elements as
2507 2511 * follows:
2508 2512 *
2509 2513 * index: an index object that lazily parses RevlogNG records
2510 2514 * cache: if data is inlined, a tuple (index_file_content, 0), else None
2511 2515 *
2512 2516 * added complications are for backwards compatibility
2513 2517 */
2514 2518 static PyObject *parse_index2(PyObject *self, PyObject *args)
2515 2519 {
2516 2520 PyObject *tuple = NULL, *cache = NULL;
2517 2521 indexObject *idx;
2518 2522 int ret;
2519 2523
2520 2524 idx = PyObject_New(indexObject, &indexType);
2521 2525 if (idx == NULL)
2522 2526 goto bail;
2523 2527
2524 2528 ret = index_init(idx, args);
2525 2529 if (ret == -1)
2526 2530 goto bail;
2527 2531
2528 2532 if (idx->inlined) {
2529 2533 cache = Py_BuildValue("iO", 0, idx->data);
2530 2534 if (cache == NULL)
2531 2535 goto bail;
2532 2536 } else {
2533 2537 cache = Py_None;
2534 2538 Py_INCREF(cache);
2535 2539 }
2536 2540
2537 2541 tuple = Py_BuildValue("NN", idx, cache);
2538 2542 if (!tuple)
2539 2543 goto bail;
2540 2544 return tuple;
2541 2545
2542 2546 bail:
2543 2547 Py_XDECREF(idx);
2544 2548 Py_XDECREF(cache);
2545 2549 Py_XDECREF(tuple);
2546 2550 return NULL;
2547 2551 }
2548 2552
2549 2553 #define BUMPED_FIX 1
2550 2554 #define USING_SHA_256 2
2551 2555
2552 2556 static PyObject *readshas(
2553 2557 const char *source, unsigned char num, Py_ssize_t hashwidth)
2554 2558 {
2555 2559 int i;
2556 2560 PyObject *list = PyTuple_New(num);
2557 2561 if (list == NULL) {
2558 2562 return NULL;
2559 2563 }
2560 2564 for (i = 0; i < num; i++) {
2561 2565 PyObject *hash = PyString_FromStringAndSize(source, hashwidth);
2562 2566 if (hash == NULL) {
2563 2567 Py_DECREF(list);
2564 2568 return NULL;
2565 2569 }
2566 2570 PyTuple_SetItem(list, i, hash);
2567 2571 source += hashwidth;
2568 2572 }
2569 2573 return list;
2570 2574 }
2571 2575
2572 2576 static PyObject *fm1readmarker(const char *data, uint32_t *msize)
2573 2577 {
2574 2578 const char *meta;
2575 2579
2576 2580 double mtime;
2577 2581 int16_t tz;
2578 2582 uint16_t flags;
2579 2583 unsigned char nsuccs, nparents, nmetadata;
2580 2584 Py_ssize_t hashwidth = 20;
2581 2585
2582 2586 PyObject *prec = NULL, *parents = NULL, *succs = NULL;
2583 2587 PyObject *metadata = NULL, *ret = NULL;
2584 2588 int i;
2585 2589
2586 2590 *msize = getbe32(data);
2587 2591 data += 4;
2588 2592 mtime = getbefloat64(data);
2589 2593 data += 8;
2590 2594 tz = getbeint16(data);
2591 2595 data += 2;
2592 2596 flags = getbeuint16(data);
2593 2597 data += 2;
2594 2598
2595 2599 if (flags & USING_SHA_256) {
2596 2600 hashwidth = 32;
2597 2601 }
2598 2602
2599 2603 nsuccs = (unsigned char)(*data++);
2600 2604 nparents = (unsigned char)(*data++);
2601 2605 nmetadata = (unsigned char)(*data++);
2602 2606
2603 2607 prec = PyString_FromStringAndSize(data, hashwidth);
2604 2608 data += hashwidth;
2605 2609 if (prec == NULL) {
2606 2610 goto bail;
2607 2611 }
2608 2612
2609 2613 succs = readshas(data, nsuccs, hashwidth);
2610 2614 if (succs == NULL) {
2611 2615 goto bail;
2612 2616 }
2613 2617 data += nsuccs * hashwidth;
2614 2618
2615 2619 if (nparents == 1 || nparents == 2) {
2616 2620 parents = readshas(data, nparents, hashwidth);
2617 2621 if (parents == NULL) {
2618 2622 goto bail;
2619 2623 }
2620 2624 data += nparents * hashwidth;
2621 2625 } else {
2622 2626 parents = Py_None;
2623 2627 }
2624 2628
2625 2629 meta = data + (2 * nmetadata);
2626 2630 metadata = PyTuple_New(nmetadata);
2627 2631 if (metadata == NULL) {
2628 2632 goto bail;
2629 2633 }
2630 2634 for (i = 0; i < nmetadata; i++) {
2631 2635 PyObject *tmp, *left = NULL, *right = NULL;
2632 2636 Py_ssize_t metasize = (unsigned char)(*data++);
2633 2637 left = PyString_FromStringAndSize(meta, metasize);
2634 2638 meta += metasize;
2635 2639 metasize = (unsigned char)(*data++);
2636 2640 right = PyString_FromStringAndSize(meta, metasize);
2637 2641 meta += metasize;
2638 2642 if (!left || !right) {
2639 2643 Py_XDECREF(left);
2640 2644 Py_XDECREF(right);
2641 2645 goto bail;
2642 2646 }
2643 2647 tmp = PyTuple_Pack(2, left, right);
2644 2648 Py_DECREF(left);
2645 2649 Py_DECREF(right);
2646 2650 if (!tmp) {
2647 2651 goto bail;
2648 2652 }
2649 2653 PyTuple_SetItem(metadata, i, tmp);
2650 2654 }
2651 2655 ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags,
2652 2656 metadata, mtime, (int)tz * 60, parents);
2653 2657 bail:
2654 2658 Py_XDECREF(prec);
2655 2659 Py_XDECREF(succs);
2656 2660 Py_XDECREF(metadata);
2657 2661 if (parents != Py_None)
2658 2662 Py_XDECREF(parents);
2659 2663 return ret;
2660 2664 }
2661 2665
2662 2666
2663 2667 static PyObject *fm1readmarkers(PyObject *self, PyObject *args) {
2664 2668 const char *data;
2665 2669 Py_ssize_t datalen;
2666 2670 /* only unsigned long because python 2.4, should be Py_ssize_t */
2667 2671 unsigned long offset, stop;
2668 2672 PyObject *markers = NULL;
2669 2673
2670 2674 /* replace kk with nn when we drop Python 2.4 */
2671 2675 if (!PyArg_ParseTuple(args, "s#kk", &data, &datalen, &offset, &stop)) {
2672 2676 return NULL;
2673 2677 }
2674 2678 data += offset;
2675 2679 markers = PyList_New(0);
2676 2680 if (!markers) {
2677 2681 return NULL;
2678 2682 }
2679 2683 while (offset < stop) {
2680 2684 uint32_t msize;
2681 2685 int error;
2682 2686 PyObject *record = fm1readmarker(data, &msize);
2683 2687 if (!record) {
2684 2688 goto bail;
2685 2689 }
2686 2690 error = PyList_Append(markers, record);
2687 2691 Py_DECREF(record);
2688 2692 if (error) {
2689 2693 goto bail;
2690 2694 }
2691 2695 data += msize;
2692 2696 offset += msize;
2693 2697 }
2694 2698 return markers;
2695 2699 bail:
2696 2700 Py_DECREF(markers);
2697 2701 return NULL;
2698 2702 }
2699 2703
2700 2704 static char parsers_doc[] = "Efficient content parsing.";
2701 2705
2702 2706 PyObject *encodedir(PyObject *self, PyObject *args);
2703 2707 PyObject *pathencode(PyObject *self, PyObject *args);
2704 2708 PyObject *lowerencode(PyObject *self, PyObject *args);
2705 2709
2706 2710 static PyMethodDef methods[] = {
2707 2711 {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
2708 2712 {"parse_manifest", parse_manifest, METH_VARARGS, "parse a manifest\n"},
2709 2713 {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
2710 2714 {"parse_index2", parse_index2, METH_VARARGS, "parse a revlog index\n"},
2711 2715 {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
2712 2716 {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
2713 2717 {"dict_new_presized", dict_new_presized, METH_VARARGS,
2714 2718 "construct a dict with an expected size\n"},
2715 2719 {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
2716 2720 "make file foldmap\n"},
2717 2721 {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
2718 2722 {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
2719 2723 {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
2720 2724 {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
2721 2725 "parse v1 obsolete markers\n"},
2722 2726 {NULL, NULL}
2723 2727 };
2724 2728
2725 2729 void dirs_module_init(PyObject *mod);
2726 2730 void manifest_module_init(PyObject *mod);
2727 2731
2728 2732 static void module_init(PyObject *mod)
2729 2733 {
2730 2734 /* This module constant has two purposes. First, it lets us unit test
2731 2735 * the ImportError raised without hard-coding any error text. This
2732 2736 * means we can change the text in the future without breaking tests,
2733 2737 * even across changesets without a recompile. Second, its presence
2734 2738 * can be used to determine whether the version-checking logic is
2735 2739 * present, which also helps in testing across changesets without a
2736 2740 * recompile. Note that this means the pure-Python version of parsers
2737 2741 * should not have this module constant. */
2738 2742 PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
2739 2743
2740 2744 dirs_module_init(mod);
2741 2745 manifest_module_init(mod);
2742 2746
2743 2747 indexType.tp_new = PyType_GenericNew;
2744 2748 if (PyType_Ready(&indexType) < 0 ||
2745 2749 PyType_Ready(&dirstateTupleType) < 0)
2746 2750 return;
2747 2751 Py_INCREF(&indexType);
2748 2752 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2749 2753 Py_INCREF(&dirstateTupleType);
2750 2754 PyModule_AddObject(mod, "dirstatetuple",
2751 2755 (PyObject *)&dirstateTupleType);
2752 2756
2753 2757 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2754 2758 -1, -1, -1, -1, nullid, 20);
2755 2759 if (nullentry)
2756 2760 PyObject_GC_UnTrack(nullentry);
2757 2761 }
2758 2762
2759 2763 static int check_python_version(void)
2760 2764 {
2761 2765 PyObject *sys = PyImport_ImportModule("sys"), *ver;
2762 2766 long hexversion;
2763 2767 if (!sys)
2764 2768 return -1;
2765 2769 ver = PyObject_GetAttrString(sys, "hexversion");
2766 2770 Py_DECREF(sys);
2767 2771 if (!ver)
2768 2772 return -1;
2769 2773 hexversion = PyInt_AsLong(ver);
2770 2774 Py_DECREF(ver);
2771 2775 /* sys.hexversion is a 32-bit number by default, so the -1 case
2772 2776 * should only occur in unusual circumstances (e.g. if sys.hexversion
2773 2777 * is manually set to an invalid value). */
2774 2778 if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
2775 2779 PyErr_Format(PyExc_ImportError, "%s: The Mercurial extension "
2776 2780 "modules were compiled with Python " PY_VERSION ", but "
2777 2781 "Mercurial is currently using Python with sys.hexversion=%ld: "
2778 2782 "Python %s\n at: %s", versionerrortext, hexversion,
2779 2783 Py_GetVersion(), Py_GetProgramFullPath());
2780 2784 return -1;
2781 2785 }
2782 2786 return 0;
2783 2787 }
2784 2788
2785 2789 #ifdef IS_PY3K
2786 2790 static struct PyModuleDef parsers_module = {
2787 2791 PyModuleDef_HEAD_INIT,
2788 2792 "parsers",
2789 2793 parsers_doc,
2790 2794 -1,
2791 2795 methods
2792 2796 };
2793 2797
2794 2798 PyMODINIT_FUNC PyInit_parsers(void)
2795 2799 {
2796 2800 PyObject *mod;
2797 2801
2798 2802 if (check_python_version() == -1)
2799 2803 return;
2800 2804 mod = PyModule_Create(&parsers_module);
2801 2805 module_init(mod);
2802 2806 return mod;
2803 2807 }
2804 2808 #else
2805 2809 PyMODINIT_FUNC initparsers(void)
2806 2810 {
2807 2811 PyObject *mod;
2808 2812
2809 2813 if (check_python_version() == -1)
2810 2814 return;
2811 2815 mod = Py_InitModule3("parsers", methods, parsers_doc);
2812 2816 module_init(mod);
2813 2817 }
2814 2818 #endif
General Comments 0
You need to be logged in to leave comments. Login now