##// END OF EJS Templates
cleanup: remove superfluous space after space after equals (C)
timeless -
r27638:90e3c512 default
parent child Browse files
Show More
@@ -1,2889 +1,2889 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 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 * Build a set of non-normal entries from the dirstate dmap
551 551 */
552 552 static PyObject *nonnormalentries(PyObject *self, PyObject *args)
553 553 {
554 554 PyObject *dmap, *nonnset = NULL, *fname, *v;
555 555 Py_ssize_t pos;
556 556
557 557 if (!PyArg_ParseTuple(args, "O!:nonnormalentries",
558 558 &PyDict_Type, &dmap))
559 559 goto bail;
560 560
561 561 nonnset = PySet_New(NULL);
562 562 if (nonnset == NULL)
563 563 goto bail;
564 564
565 565 pos = 0;
566 566 while (PyDict_Next(dmap, &pos, &fname, &v)) {
567 567 dirstateTupleObject *t;
568 568 if (!dirstate_tuple_check(v)) {
569 569 PyErr_SetString(PyExc_TypeError,
570 570 "expected a dirstate tuple");
571 571 goto bail;
572 572 }
573 573 t = (dirstateTupleObject *)v;
574 574
575 575 if (t->state == 'n' && t->mtime != -1)
576 576 continue;
577 577 if (PySet_Add(nonnset, fname) == -1)
578 578 goto bail;
579 579 }
580 580
581 581 return nonnset;
582 582 bail:
583 583 Py_XDECREF(nonnset);
584 584 return NULL;
585 585 }
586 586
587 587 /*
588 588 * Efficiently pack a dirstate object into its on-disk format.
589 589 */
590 590 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
591 591 {
592 592 PyObject *packobj = NULL;
593 593 PyObject *map, *copymap, *pl, *mtime_unset = NULL;
594 594 Py_ssize_t nbytes, pos, l;
595 595 PyObject *k, *v = NULL, *pn;
596 596 char *p, *s;
597 597 int now;
598 598
599 599 if (!PyArg_ParseTuple(args, "O!O!Oi:pack_dirstate",
600 600 &PyDict_Type, &map, &PyDict_Type, &copymap,
601 601 &pl, &now))
602 602 return NULL;
603 603
604 604 if (!PySequence_Check(pl) || PySequence_Size(pl) != 2) {
605 605 PyErr_SetString(PyExc_TypeError, "expected 2-element sequence");
606 606 return NULL;
607 607 }
608 608
609 609 /* Figure out how much we need to allocate. */
610 610 for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
611 611 PyObject *c;
612 612 if (!PyString_Check(k)) {
613 613 PyErr_SetString(PyExc_TypeError, "expected string key");
614 614 goto bail;
615 615 }
616 616 nbytes += PyString_GET_SIZE(k) + 17;
617 617 c = PyDict_GetItem(copymap, k);
618 618 if (c) {
619 619 if (!PyString_Check(c)) {
620 620 PyErr_SetString(PyExc_TypeError,
621 621 "expected string key");
622 622 goto bail;
623 623 }
624 624 nbytes += PyString_GET_SIZE(c) + 1;
625 625 }
626 626 }
627 627
628 628 packobj = PyString_FromStringAndSize(NULL, nbytes);
629 629 if (packobj == NULL)
630 630 goto bail;
631 631
632 632 p = PyString_AS_STRING(packobj);
633 633
634 634 pn = PySequence_ITEM(pl, 0);
635 635 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
636 636 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
637 637 goto bail;
638 638 }
639 639 memcpy(p, s, l);
640 640 p += 20;
641 641 pn = PySequence_ITEM(pl, 1);
642 642 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
643 643 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
644 644 goto bail;
645 645 }
646 646 memcpy(p, s, l);
647 647 p += 20;
648 648
649 649 for (pos = 0; PyDict_Next(map, &pos, &k, &v); ) {
650 650 dirstateTupleObject *tuple;
651 651 char state;
652 652 int mode, size, mtime;
653 653 Py_ssize_t len, l;
654 654 PyObject *o;
655 655 char *t;
656 656
657 657 if (!dirstate_tuple_check(v)) {
658 658 PyErr_SetString(PyExc_TypeError,
659 659 "expected a dirstate tuple");
660 660 goto bail;
661 661 }
662 662 tuple = (dirstateTupleObject *)v;
663 663
664 664 state = tuple->state;
665 665 mode = tuple->mode;
666 666 size = tuple->size;
667 667 mtime = tuple->mtime;
668 668 if (state == 'n' && mtime == now) {
669 669 /* See pure/parsers.py:pack_dirstate for why we do
670 670 * this. */
671 671 mtime = -1;
672 672 mtime_unset = (PyObject *)make_dirstate_tuple(
673 673 state, mode, size, mtime);
674 674 if (!mtime_unset)
675 675 goto bail;
676 676 if (PyDict_SetItem(map, k, mtime_unset) == -1)
677 677 goto bail;
678 678 Py_DECREF(mtime_unset);
679 679 mtime_unset = NULL;
680 680 }
681 681 *p++ = state;
682 682 putbe32((uint32_t)mode, p);
683 683 putbe32((uint32_t)size, p + 4);
684 684 putbe32((uint32_t)mtime, p + 8);
685 685 t = p + 12;
686 686 p += 16;
687 687 len = PyString_GET_SIZE(k);
688 688 memcpy(p, PyString_AS_STRING(k), len);
689 689 p += len;
690 690 o = PyDict_GetItem(copymap, k);
691 691 if (o) {
692 692 *p++ = '\0';
693 693 l = PyString_GET_SIZE(o);
694 694 memcpy(p, PyString_AS_STRING(o), l);
695 695 p += l;
696 696 len += l + 1;
697 697 }
698 698 putbe32((uint32_t)len, t);
699 699 }
700 700
701 701 pos = p - PyString_AS_STRING(packobj);
702 702 if (pos != nbytes) {
703 703 PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
704 704 (long)pos, (long)nbytes);
705 705 goto bail;
706 706 }
707 707
708 708 return packobj;
709 709 bail:
710 710 Py_XDECREF(mtime_unset);
711 711 Py_XDECREF(packobj);
712 712 Py_XDECREF(v);
713 713 return NULL;
714 714 }
715 715
716 716 /*
717 717 * A base-16 trie for fast node->rev mapping.
718 718 *
719 719 * Positive value is index of the next node in the trie
720 720 * Negative value is a leaf: -(rev + 1)
721 721 * Zero is empty
722 722 */
723 723 typedef struct {
724 724 int children[16];
725 725 } nodetree;
726 726
727 727 /*
728 728 * This class has two behaviors.
729 729 *
730 730 * When used in a list-like way (with integer keys), we decode an
731 731 * entry in a RevlogNG index file on demand. Our last entry is a
732 732 * sentinel, always a nullid. We have limited support for
733 733 * integer-keyed insert and delete, only at elements right before the
734 734 * sentinel.
735 735 *
736 736 * With string keys, we lazily perform a reverse mapping from node to
737 737 * rev, using a base-16 trie.
738 738 */
739 739 typedef struct {
740 740 PyObject_HEAD
741 741 /* Type-specific fields go here. */
742 742 PyObject *data; /* raw bytes of index */
743 743 PyObject **cache; /* cached tuples */
744 744 const char **offsets; /* populated on demand */
745 745 Py_ssize_t raw_length; /* original number of elements */
746 746 Py_ssize_t length; /* current number of elements */
747 747 PyObject *added; /* populated on demand */
748 748 PyObject *headrevs; /* cache, invalidated on changes */
749 749 PyObject *filteredrevs;/* filtered revs set */
750 750 nodetree *nt; /* base-16 trie */
751 751 unsigned ntlength; /* # nodes in use */
752 752 unsigned ntcapacity; /* # nodes allocated */
753 753 int ntdepth; /* maximum depth of tree */
754 754 int ntsplits; /* # splits performed */
755 755 int ntrev; /* last rev scanned */
756 756 int ntlookups; /* # lookups */
757 757 int ntmisses; /* # lookups that miss the cache */
758 758 int inlined;
759 759 } indexObject;
760 760
761 761 static Py_ssize_t index_length(const indexObject *self)
762 762 {
763 763 if (self->added == NULL)
764 764 return self->length;
765 765 return self->length + PyList_GET_SIZE(self->added);
766 766 }
767 767
768 768 static PyObject *nullentry;
769 769 static const char nullid[20];
770 770
771 771 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
772 772
773 773 #if LONG_MAX == 0x7fffffffL
774 774 static char *tuple_format = "Kiiiiiis#";
775 775 #else
776 776 static char *tuple_format = "kiiiiiis#";
777 777 #endif
778 778
779 779 /* A RevlogNG v1 index entry is 64 bytes long. */
780 780 static const long v1_hdrsize = 64;
781 781
782 782 /*
783 783 * Return a pointer to the beginning of a RevlogNG record.
784 784 */
785 785 static const char *index_deref(indexObject *self, Py_ssize_t pos)
786 786 {
787 787 if (self->inlined && pos > 0) {
788 788 if (self->offsets == NULL) {
789 789 self->offsets = malloc(self->raw_length *
790 790 sizeof(*self->offsets));
791 791 if (self->offsets == NULL)
792 792 return (const char *)PyErr_NoMemory();
793 793 inline_scan(self, self->offsets);
794 794 }
795 795 return self->offsets[pos];
796 796 }
797 797
798 798 return PyString_AS_STRING(self->data) + pos * v1_hdrsize;
799 799 }
800 800
801 801 static inline int index_get_parents(indexObject *self, Py_ssize_t rev,
802 802 int *ps, int maxrev)
803 803 {
804 804 if (rev >= self->length - 1) {
805 805 PyObject *tuple = PyList_GET_ITEM(self->added,
806 806 rev - self->length + 1);
807 807 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
808 808 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
809 809 } else {
810 810 const char *data = index_deref(self, rev);
811 811 ps[0] = getbe32(data + 24);
812 812 ps[1] = getbe32(data + 28);
813 813 }
814 814 /* If index file is corrupted, ps[] may point to invalid revisions. So
815 815 * there is a risk of buffer overflow to trust them unconditionally. */
816 816 if (ps[0] > maxrev || ps[1] > maxrev) {
817 817 PyErr_SetString(PyExc_ValueError, "parent out of range");
818 818 return -1;
819 819 }
820 820 return 0;
821 821 }
822 822
823 823
824 824 /*
825 825 * RevlogNG format (all in big endian, data may be inlined):
826 826 * 6 bytes: offset
827 827 * 2 bytes: flags
828 828 * 4 bytes: compressed length
829 829 * 4 bytes: uncompressed length
830 830 * 4 bytes: base revision
831 831 * 4 bytes: link revision
832 832 * 4 bytes: parent 1 revision
833 833 * 4 bytes: parent 2 revision
834 834 * 32 bytes: nodeid (only 20 bytes used)
835 835 */
836 836 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
837 837 {
838 838 uint64_t offset_flags;
839 839 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
840 840 const char *c_node_id;
841 841 const char *data;
842 842 Py_ssize_t length = index_length(self);
843 843 PyObject *entry;
844 844
845 845 if (pos < 0)
846 846 pos += length;
847 847
848 848 if (pos < 0 || pos >= length) {
849 849 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
850 850 return NULL;
851 851 }
852 852
853 853 if (pos == length - 1) {
854 854 Py_INCREF(nullentry);
855 855 return nullentry;
856 856 }
857 857
858 858 if (pos >= self->length - 1) {
859 859 PyObject *obj;
860 860 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
861 861 Py_INCREF(obj);
862 862 return obj;
863 863 }
864 864
865 865 if (self->cache) {
866 866 if (self->cache[pos]) {
867 867 Py_INCREF(self->cache[pos]);
868 868 return self->cache[pos];
869 869 }
870 870 } else {
871 871 self->cache = calloc(self->raw_length, sizeof(PyObject *));
872 872 if (self->cache == NULL)
873 873 return PyErr_NoMemory();
874 874 }
875 875
876 876 data = index_deref(self, pos);
877 877 if (data == NULL)
878 878 return NULL;
879 879
880 880 offset_flags = getbe32(data + 4);
881 881 if (pos == 0) /* mask out version number for the first entry */
882 882 offset_flags &= 0xFFFF;
883 883 else {
884 884 uint32_t offset_high = getbe32(data);
885 885 offset_flags |= ((uint64_t)offset_high) << 32;
886 886 }
887 887
888 888 comp_len = getbe32(data + 8);
889 889 uncomp_len = getbe32(data + 12);
890 890 base_rev = getbe32(data + 16);
891 891 link_rev = getbe32(data + 20);
892 892 parent_1 = getbe32(data + 24);
893 893 parent_2 = getbe32(data + 28);
894 894 c_node_id = data + 32;
895 895
896 896 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
897 897 uncomp_len, base_rev, link_rev,
898 898 parent_1, parent_2, c_node_id, 20);
899 899
900 900 if (entry) {
901 901 PyObject_GC_UnTrack(entry);
902 902 Py_INCREF(entry);
903 903 }
904 904
905 905 self->cache[pos] = entry;
906 906
907 907 return entry;
908 908 }
909 909
910 910 /*
911 911 * Return the 20-byte SHA of the node corresponding to the given rev.
912 912 */
913 913 static const char *index_node(indexObject *self, Py_ssize_t pos)
914 914 {
915 915 Py_ssize_t length = index_length(self);
916 916 const char *data;
917 917
918 918 if (pos == length - 1 || pos == INT_MAX)
919 919 return nullid;
920 920
921 921 if (pos >= length)
922 922 return NULL;
923 923
924 924 if (pos >= self->length - 1) {
925 925 PyObject *tuple, *str;
926 926 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
927 927 str = PyTuple_GetItem(tuple, 7);
928 928 return str ? PyString_AS_STRING(str) : NULL;
929 929 }
930 930
931 931 data = index_deref(self, pos);
932 932 return data ? data + 32 : NULL;
933 933 }
934 934
935 935 static int nt_insert(indexObject *self, const char *node, int rev);
936 936
937 937 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
938 938 {
939 939 if (PyString_AsStringAndSize(obj, node, nodelen) == -1)
940 940 return -1;
941 941 if (*nodelen == 20)
942 942 return 0;
943 943 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
944 944 return -1;
945 945 }
946 946
947 947 static PyObject *index_insert(indexObject *self, PyObject *args)
948 948 {
949 949 PyObject *obj;
950 950 char *node;
951 951 int index;
952 952 Py_ssize_t len, nodelen;
953 953
954 954 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
955 955 return NULL;
956 956
957 957 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
958 958 PyErr_SetString(PyExc_TypeError, "8-tuple required");
959 959 return NULL;
960 960 }
961 961
962 962 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
963 963 return NULL;
964 964
965 965 len = index_length(self);
966 966
967 967 if (index < 0)
968 968 index += len;
969 969
970 970 if (index != len - 1) {
971 971 PyErr_SetString(PyExc_IndexError,
972 972 "insert only supported at index -1");
973 973 return NULL;
974 974 }
975 975
976 976 if (self->added == NULL) {
977 977 self->added = PyList_New(0);
978 978 if (self->added == NULL)
979 979 return NULL;
980 980 }
981 981
982 982 if (PyList_Append(self->added, obj) == -1)
983 983 return NULL;
984 984
985 985 if (self->nt)
986 986 nt_insert(self, node, index);
987 987
988 988 Py_CLEAR(self->headrevs);
989 989 Py_RETURN_NONE;
990 990 }
991 991
992 992 static void _index_clearcaches(indexObject *self)
993 993 {
994 994 if (self->cache) {
995 995 Py_ssize_t i;
996 996
997 997 for (i = 0; i < self->raw_length; i++)
998 998 Py_CLEAR(self->cache[i]);
999 999 free(self->cache);
1000 1000 self->cache = NULL;
1001 1001 }
1002 1002 if (self->offsets) {
1003 1003 free(self->offsets);
1004 1004 self->offsets = NULL;
1005 1005 }
1006 1006 if (self->nt) {
1007 1007 free(self->nt);
1008 1008 self->nt = NULL;
1009 1009 }
1010 1010 Py_CLEAR(self->headrevs);
1011 1011 }
1012 1012
1013 1013 static PyObject *index_clearcaches(indexObject *self)
1014 1014 {
1015 1015 _index_clearcaches(self);
1016 1016 self->ntlength = self->ntcapacity = 0;
1017 1017 self->ntdepth = self->ntsplits = 0;
1018 1018 self->ntrev = -1;
1019 1019 self->ntlookups = self->ntmisses = 0;
1020 1020 Py_RETURN_NONE;
1021 1021 }
1022 1022
1023 1023 static PyObject *index_stats(indexObject *self)
1024 1024 {
1025 1025 PyObject *obj = PyDict_New();
1026 1026 PyObject *t = NULL;
1027 1027
1028 1028 if (obj == NULL)
1029 1029 return NULL;
1030 1030
1031 1031 #define istat(__n, __d) \
1032 1032 t = PyInt_FromSsize_t(self->__n); \
1033 1033 if (!t) \
1034 1034 goto bail; \
1035 1035 if (PyDict_SetItemString(obj, __d, t) == -1) \
1036 1036 goto bail; \
1037 1037 Py_DECREF(t);
1038 1038
1039 1039 if (self->added) {
1040 1040 Py_ssize_t len = PyList_GET_SIZE(self->added);
1041 1041 t = PyInt_FromSsize_t(len);
1042 1042 if (!t)
1043 1043 goto bail;
1044 1044 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
1045 1045 goto bail;
1046 1046 Py_DECREF(t);
1047 1047 }
1048 1048
1049 1049 if (self->raw_length != self->length - 1)
1050 1050 istat(raw_length, "revs on disk");
1051 1051 istat(length, "revs in memory");
1052 1052 istat(ntcapacity, "node trie capacity");
1053 1053 istat(ntdepth, "node trie depth");
1054 1054 istat(ntlength, "node trie count");
1055 1055 istat(ntlookups, "node trie lookups");
1056 1056 istat(ntmisses, "node trie misses");
1057 1057 istat(ntrev, "node trie last rev scanned");
1058 1058 istat(ntsplits, "node trie splits");
1059 1059
1060 1060 #undef istat
1061 1061
1062 1062 return obj;
1063 1063
1064 1064 bail:
1065 1065 Py_XDECREF(obj);
1066 1066 Py_XDECREF(t);
1067 1067 return NULL;
1068 1068 }
1069 1069
1070 1070 /*
1071 1071 * When we cache a list, we want to be sure the caller can't mutate
1072 1072 * the cached copy.
1073 1073 */
1074 1074 static PyObject *list_copy(PyObject *list)
1075 1075 {
1076 1076 Py_ssize_t len = PyList_GET_SIZE(list);
1077 1077 PyObject *newlist = PyList_New(len);
1078 1078 Py_ssize_t i;
1079 1079
1080 1080 if (newlist == NULL)
1081 1081 return NULL;
1082 1082
1083 1083 for (i = 0; i < len; i++) {
1084 1084 PyObject *obj = PyList_GET_ITEM(list, i);
1085 1085 Py_INCREF(obj);
1086 1086 PyList_SET_ITEM(newlist, i, obj);
1087 1087 }
1088 1088
1089 1089 return newlist;
1090 1090 }
1091 1091
1092 1092 static int check_filter(PyObject *filter, Py_ssize_t arg) {
1093 1093 if (filter) {
1094 1094 PyObject *arglist, *result;
1095 1095 int isfiltered;
1096 1096
1097 1097 arglist = Py_BuildValue("(n)", arg);
1098 1098 if (!arglist) {
1099 1099 return -1;
1100 1100 }
1101 1101
1102 1102 result = PyEval_CallObject(filter, arglist);
1103 1103 Py_DECREF(arglist);
1104 1104 if (!result) {
1105 1105 return -1;
1106 1106 }
1107 1107
1108 1108 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
1109 1109 * same as this function, so we can just return it directly.*/
1110 1110 isfiltered = PyObject_IsTrue(result);
1111 1111 Py_DECREF(result);
1112 1112 return isfiltered;
1113 1113 } else {
1114 1114 return 0;
1115 1115 }
1116 1116 }
1117 1117
1118 1118 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
1119 1119 Py_ssize_t marker, char *phases)
1120 1120 {
1121 1121 PyObject *iter = NULL;
1122 1122 PyObject *iter_item = NULL;
1123 1123 Py_ssize_t min_idx = index_length(self) + 1;
1124 1124 long iter_item_long;
1125 1125
1126 1126 if (PyList_GET_SIZE(list) != 0) {
1127 1127 iter = PyObject_GetIter(list);
1128 1128 if (iter == NULL)
1129 1129 return -2;
1130 1130 while ((iter_item = PyIter_Next(iter)))
1131 1131 {
1132 1132 iter_item_long = PyInt_AS_LONG(iter_item);
1133 1133 Py_DECREF(iter_item);
1134 1134 if (iter_item_long < min_idx)
1135 1135 min_idx = iter_item_long;
1136 1136 phases[iter_item_long] = marker;
1137 1137 }
1138 1138 Py_DECREF(iter);
1139 1139 }
1140 1140
1141 1141 return min_idx;
1142 1142 }
1143 1143
1144 1144 static inline void set_phase_from_parents(char *phases, int parent_1,
1145 1145 int parent_2, Py_ssize_t i)
1146 1146 {
1147 1147 if (parent_1 >= 0 && phases[parent_1] > phases[i])
1148 1148 phases[i] = phases[parent_1];
1149 1149 if (parent_2 >= 0 && phases[parent_2] > phases[i])
1150 1150 phases[i] = phases[parent_2];
1151 1151 }
1152 1152
1153 1153 static PyObject *reachableroots2(indexObject *self, PyObject *args)
1154 1154 {
1155 1155
1156 1156 /* Input */
1157 1157 long minroot;
1158 1158 PyObject *includepatharg = NULL;
1159 1159 int includepath = 0;
1160 1160 /* heads and roots are lists */
1161 1161 PyObject *heads = NULL;
1162 1162 PyObject *roots = NULL;
1163 1163 PyObject *reachable = NULL;
1164 1164
1165 1165 PyObject *val;
1166 1166 Py_ssize_t len = index_length(self) - 1;
1167 1167 long revnum;
1168 1168 Py_ssize_t k;
1169 1169 Py_ssize_t i;
1170 1170 Py_ssize_t l;
1171 1171 int r;
1172 1172 int parents[2];
1173 1173
1174 1174 /* Internal data structure:
1175 1175 * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit
1176 1176 * revstates: array of length len+1 (all revs + nullrev) */
1177 1177 int *tovisit = NULL;
1178 1178 long lentovisit = 0;
1179 1179 enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 };
1180 1180 char *revstates = NULL;
1181 1181
1182 1182 /* Get arguments */
1183 1183 if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
1184 1184 &PyList_Type, &roots,
1185 1185 &PyBool_Type, &includepatharg))
1186 1186 goto bail;
1187 1187
1188 1188 if (includepatharg == Py_True)
1189 1189 includepath = 1;
1190 1190
1191 1191 /* Initialize return set */
1192 1192 reachable = PyList_New(0);
1193 1193 if (reachable == NULL)
1194 1194 goto bail;
1195 1195
1196 1196 /* Initialize internal datastructures */
1197 1197 tovisit = (int *)malloc((len + 1) * sizeof(int));
1198 1198 if (tovisit == NULL) {
1199 1199 PyErr_NoMemory();
1200 1200 goto bail;
1201 1201 }
1202 1202
1203 1203 revstates = (char *)calloc(len + 1, 1);
1204 1204 if (revstates == NULL) {
1205 1205 PyErr_NoMemory();
1206 1206 goto bail;
1207 1207 }
1208 1208
1209 1209 l = PyList_GET_SIZE(roots);
1210 1210 for (i = 0; i < l; i++) {
1211 1211 revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i));
1212 1212 if (revnum == -1 && PyErr_Occurred())
1213 1213 goto bail;
1214 1214 /* If root is out of range, e.g. wdir(), it must be unreachable
1215 1215 * from heads. So we can just ignore it. */
1216 1216 if (revnum + 1 < 0 || revnum + 1 >= len + 1)
1217 1217 continue;
1218 1218 revstates[revnum + 1] |= RS_ROOT;
1219 1219 }
1220 1220
1221 1221 /* Populate tovisit with all the heads */
1222 1222 l = PyList_GET_SIZE(heads);
1223 1223 for (i = 0; i < l; i++) {
1224 1224 revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
1225 1225 if (revnum == -1 && PyErr_Occurred())
1226 1226 goto bail;
1227 1227 if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
1228 1228 PyErr_SetString(PyExc_IndexError, "head out of range");
1229 1229 goto bail;
1230 1230 }
1231 1231 if (!(revstates[revnum + 1] & RS_SEEN)) {
1232 1232 tovisit[lentovisit++] = (int)revnum;
1233 1233 revstates[revnum + 1] |= RS_SEEN;
1234 1234 }
1235 1235 }
1236 1236
1237 1237 /* Visit the tovisit list and find the reachable roots */
1238 1238 k = 0;
1239 1239 while (k < lentovisit) {
1240 1240 /* Add the node to reachable if it is a root*/
1241 1241 revnum = tovisit[k++];
1242 1242 if (revstates[revnum + 1] & RS_ROOT) {
1243 1243 revstates[revnum + 1] |= RS_REACHABLE;
1244 1244 val = PyInt_FromLong(revnum);
1245 1245 if (val == NULL)
1246 1246 goto bail;
1247 1247 r = PyList_Append(reachable, val);
1248 1248 Py_DECREF(val);
1249 1249 if (r < 0)
1250 1250 goto bail;
1251 1251 if (includepath == 0)
1252 1252 continue;
1253 1253 }
1254 1254
1255 1255 /* Add its parents to the list of nodes to visit */
1256 1256 if (revnum == -1)
1257 1257 continue;
1258 1258 r = index_get_parents(self, revnum, parents, (int)len - 1);
1259 1259 if (r < 0)
1260 1260 goto bail;
1261 1261 for (i = 0; i < 2; i++) {
1262 1262 if (!(revstates[parents[i] + 1] & RS_SEEN)
1263 1263 && parents[i] >= minroot) {
1264 1264 tovisit[lentovisit++] = parents[i];
1265 1265 revstates[parents[i] + 1] |= RS_SEEN;
1266 1266 }
1267 1267 }
1268 1268 }
1269 1269
1270 1270 /* Find all the nodes in between the roots we found and the heads
1271 1271 * and add them to the reachable set */
1272 1272 if (includepath == 1) {
1273 1273 long minidx = minroot;
1274 1274 if (minidx < 0)
1275 1275 minidx = 0;
1276 1276 for (i = minidx; i < len; i++) {
1277 1277 if (!(revstates[i + 1] & RS_SEEN))
1278 1278 continue;
1279 1279 r = index_get_parents(self, i, parents, (int)len - 1);
1280 1280 /* Corrupted index file, error is set from
1281 1281 * index_get_parents */
1282 1282 if (r < 0)
1283 1283 goto bail;
1284 1284 if (((revstates[parents[0] + 1] |
1285 1285 revstates[parents[1] + 1]) & RS_REACHABLE)
1286 1286 && !(revstates[i + 1] & RS_REACHABLE)) {
1287 1287 revstates[i + 1] |= RS_REACHABLE;
1288 1288 val = PyInt_FromLong(i);
1289 1289 if (val == NULL)
1290 1290 goto bail;
1291 1291 r = PyList_Append(reachable, val);
1292 1292 Py_DECREF(val);
1293 1293 if (r < 0)
1294 1294 goto bail;
1295 1295 }
1296 1296 }
1297 1297 }
1298 1298
1299 1299 free(revstates);
1300 1300 free(tovisit);
1301 1301 return reachable;
1302 1302 bail:
1303 1303 Py_XDECREF(reachable);
1304 1304 free(revstates);
1305 1305 free(tovisit);
1306 1306 return NULL;
1307 1307 }
1308 1308
1309 1309 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
1310 1310 {
1311 1311 PyObject *roots = Py_None;
1312 1312 PyObject *ret = NULL;
1313 1313 PyObject *phaseslist = NULL;
1314 1314 PyObject *phaseroots = NULL;
1315 1315 PyObject *phaseset = NULL;
1316 1316 PyObject *phasessetlist = NULL;
1317 1317 PyObject *rev = NULL;
1318 1318 Py_ssize_t len = index_length(self) - 1;
1319 1319 Py_ssize_t numphase = 0;
1320 1320 Py_ssize_t minrevallphases = 0;
1321 1321 Py_ssize_t minrevphase = 0;
1322 1322 Py_ssize_t i = 0;
1323 1323 char *phases = NULL;
1324 1324 long phase;
1325 1325
1326 1326 if (!PyArg_ParseTuple(args, "O", &roots))
1327 1327 goto done;
1328 1328 if (roots == NULL || !PyList_Check(roots))
1329 1329 goto done;
1330 1330
1331 1331 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
1332 1332 if (phases == NULL) {
1333 1333 PyErr_NoMemory();
1334 1334 goto done;
1335 1335 }
1336 1336 /* Put the phase information of all the roots in phases */
1337 1337 numphase = PyList_GET_SIZE(roots)+1;
1338 1338 minrevallphases = len + 1;
1339 1339 phasessetlist = PyList_New(numphase);
1340 1340 if (phasessetlist == NULL)
1341 1341 goto done;
1342 1342
1343 1343 PyList_SET_ITEM(phasessetlist, 0, Py_None);
1344 1344 Py_INCREF(Py_None);
1345 1345
1346 1346 for (i = 0; i < numphase-1; i++) {
1347 1347 phaseroots = PyList_GET_ITEM(roots, i);
1348 1348 phaseset = PySet_New(NULL);
1349 1349 if (phaseset == NULL)
1350 1350 goto release;
1351 1351 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
1352 1352 if (!PyList_Check(phaseroots))
1353 1353 goto release;
1354 1354 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
1355 1355 if (minrevphase == -2) /* Error from add_roots_get_min */
1356 1356 goto release;
1357 1357 minrevallphases = MIN(minrevallphases, minrevphase);
1358 1358 }
1359 1359 /* Propagate the phase information from the roots to the revs */
1360 1360 if (minrevallphases != -1) {
1361 1361 int parents[2];
1362 1362 for (i = minrevallphases; i < len; i++) {
1363 1363 if (index_get_parents(self, i, parents,
1364 1364 (int)len - 1) < 0)
1365 1365 goto release;
1366 1366 set_phase_from_parents(phases, parents[0], parents[1], i);
1367 1367 }
1368 1368 }
1369 1369 /* Transform phase list to a python list */
1370 1370 phaseslist = PyList_New(len);
1371 1371 if (phaseslist == NULL)
1372 1372 goto release;
1373 1373 for (i = 0; i < len; i++) {
1374 1374 PyObject *phaseval;
1375 1375
1376 1376 phase = phases[i];
1377 1377 /* We only store the sets of phase for non public phase, the public phase
1378 1378 * is computed as a difference */
1379 1379 if (phase != 0) {
1380 1380 phaseset = PyList_GET_ITEM(phasessetlist, phase);
1381 1381 rev = PyInt_FromLong(i);
1382 1382 if (rev == NULL)
1383 1383 goto release;
1384 1384 PySet_Add(phaseset, rev);
1385 1385 Py_XDECREF(rev);
1386 1386 }
1387 1387 phaseval = PyInt_FromLong(phase);
1388 1388 if (phaseval == NULL)
1389 1389 goto release;
1390 1390 PyList_SET_ITEM(phaseslist, i, phaseval);
1391 1391 }
1392 1392 ret = PyTuple_Pack(2, phaseslist, phasessetlist);
1393 1393
1394 1394 release:
1395 1395 Py_XDECREF(phaseslist);
1396 1396 Py_XDECREF(phasessetlist);
1397 1397 done:
1398 1398 free(phases);
1399 1399 return ret;
1400 1400 }
1401 1401
1402 1402 static PyObject *index_headrevs(indexObject *self, PyObject *args)
1403 1403 {
1404 1404 Py_ssize_t i, j, len;
1405 1405 char *nothead = NULL;
1406 1406 PyObject *heads = NULL;
1407 1407 PyObject *filter = NULL;
1408 1408 PyObject *filteredrevs = Py_None;
1409 1409
1410 1410 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
1411 1411 return NULL;
1412 1412 }
1413 1413
1414 1414 if (self->headrevs && filteredrevs == self->filteredrevs)
1415 1415 return list_copy(self->headrevs);
1416 1416
1417 1417 Py_DECREF(self->filteredrevs);
1418 1418 self->filteredrevs = filteredrevs;
1419 1419 Py_INCREF(filteredrevs);
1420 1420
1421 1421 if (filteredrevs != Py_None) {
1422 1422 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
1423 1423 if (!filter) {
1424 1424 PyErr_SetString(PyExc_TypeError,
1425 1425 "filteredrevs has no attribute __contains__");
1426 1426 goto bail;
1427 1427 }
1428 1428 }
1429 1429
1430 1430 len = index_length(self) - 1;
1431 1431 heads = PyList_New(0);
1432 1432 if (heads == NULL)
1433 1433 goto bail;
1434 1434 if (len == 0) {
1435 1435 PyObject *nullid = PyInt_FromLong(-1);
1436 1436 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
1437 1437 Py_XDECREF(nullid);
1438 1438 goto bail;
1439 1439 }
1440 1440 goto done;
1441 1441 }
1442 1442
1443 1443 nothead = calloc(len, 1);
1444 1444 if (nothead == NULL) {
1445 1445 PyErr_NoMemory();
1446 1446 goto bail;
1447 1447 }
1448 1448
1449 1449 for (i = 0; i < len; i++) {
1450 1450 int isfiltered;
1451 1451 int parents[2];
1452 1452
1453 1453 isfiltered = check_filter(filter, i);
1454 1454 if (isfiltered == -1) {
1455 1455 PyErr_SetString(PyExc_TypeError,
1456 1456 "unable to check filter");
1457 1457 goto bail;
1458 1458 }
1459 1459
1460 1460 if (isfiltered) {
1461 1461 nothead[i] = 1;
1462 1462 continue;
1463 1463 }
1464 1464
1465 1465 if (index_get_parents(self, i, parents, (int)len - 1) < 0)
1466 1466 goto bail;
1467 1467 for (j = 0; j < 2; j++) {
1468 1468 if (parents[j] >= 0)
1469 1469 nothead[parents[j]] = 1;
1470 1470 }
1471 1471 }
1472 1472
1473 1473 for (i = 0; i < len; i++) {
1474 1474 PyObject *head;
1475 1475
1476 1476 if (nothead[i])
1477 1477 continue;
1478 1478 head = PyInt_FromSsize_t(i);
1479 1479 if (head == NULL || PyList_Append(heads, head) == -1) {
1480 1480 Py_XDECREF(head);
1481 1481 goto bail;
1482 1482 }
1483 1483 }
1484 1484
1485 1485 done:
1486 1486 self->headrevs = heads;
1487 1487 Py_XDECREF(filter);
1488 1488 free(nothead);
1489 1489 return list_copy(self->headrevs);
1490 1490 bail:
1491 1491 Py_XDECREF(filter);
1492 1492 Py_XDECREF(heads);
1493 1493 free(nothead);
1494 1494 return NULL;
1495 1495 }
1496 1496
1497 1497 static inline int nt_level(const char *node, Py_ssize_t level)
1498 1498 {
1499 1499 int v = node[level>>1];
1500 1500 if (!(level & 1))
1501 1501 v >>= 4;
1502 1502 return v & 0xf;
1503 1503 }
1504 1504
1505 1505 /*
1506 1506 * Return values:
1507 1507 *
1508 1508 * -4: match is ambiguous (multiple candidates)
1509 1509 * -2: not found
1510 1510 * rest: valid rev
1511 1511 */
1512 1512 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
1513 1513 int hex)
1514 1514 {
1515 1515 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
1516 1516 int level, maxlevel, off;
1517 1517
1518 1518 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
1519 1519 return -1;
1520 1520
1521 1521 if (self->nt == NULL)
1522 1522 return -2;
1523 1523
1524 1524 if (hex)
1525 1525 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
1526 1526 else
1527 1527 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
1528 1528
1529 1529 for (level = off = 0; level < maxlevel; level++) {
1530 1530 int k = getnybble(node, level);
1531 1531 nodetree *n = &self->nt[off];
1532 1532 int v = n->children[k];
1533 1533
1534 1534 if (v < 0) {
1535 1535 const char *n;
1536 1536 Py_ssize_t i;
1537 1537
1538 1538 v = -(v + 1);
1539 1539 n = index_node(self, v);
1540 1540 if (n == NULL)
1541 1541 return -2;
1542 1542 for (i = level; i < maxlevel; i++)
1543 1543 if (getnybble(node, i) != nt_level(n, i))
1544 1544 return -2;
1545 1545 return v;
1546 1546 }
1547 1547 if (v == 0)
1548 1548 return -2;
1549 1549 off = v;
1550 1550 }
1551 1551 /* multiple matches against an ambiguous prefix */
1552 1552 return -4;
1553 1553 }
1554 1554
1555 1555 static int nt_new(indexObject *self)
1556 1556 {
1557 1557 if (self->ntlength == self->ntcapacity) {
1558 1558 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1559 1559 PyErr_SetString(PyExc_MemoryError,
1560 1560 "overflow in nt_new");
1561 1561 return -1;
1562 1562 }
1563 1563 self->ntcapacity *= 2;
1564 1564 self->nt = realloc(self->nt,
1565 1565 self->ntcapacity * sizeof(nodetree));
1566 1566 if (self->nt == NULL) {
1567 1567 PyErr_SetString(PyExc_MemoryError, "out of memory");
1568 1568 return -1;
1569 1569 }
1570 1570 memset(&self->nt[self->ntlength], 0,
1571 1571 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1572 1572 }
1573 1573 return self->ntlength++;
1574 1574 }
1575 1575
1576 1576 static int nt_insert(indexObject *self, const char *node, int rev)
1577 1577 {
1578 1578 int level = 0;
1579 1579 int off = 0;
1580 1580
1581 1581 while (level < 40) {
1582 1582 int k = nt_level(node, level);
1583 1583 nodetree *n;
1584 1584 int v;
1585 1585
1586 1586 n = &self->nt[off];
1587 1587 v = n->children[k];
1588 1588
1589 1589 if (v == 0) {
1590 1590 n->children[k] = -rev - 1;
1591 1591 return 0;
1592 1592 }
1593 1593 if (v < 0) {
1594 1594 const char *oldnode = index_node(self, -(v + 1));
1595 1595 int noff;
1596 1596
1597 1597 if (!oldnode || !memcmp(oldnode, node, 20)) {
1598 1598 n->children[k] = -rev - 1;
1599 1599 return 0;
1600 1600 }
1601 1601 noff = nt_new(self);
1602 1602 if (noff == -1)
1603 1603 return -1;
1604 1604 /* self->nt may have been changed by realloc */
1605 1605 self->nt[off].children[k] = noff;
1606 1606 off = noff;
1607 1607 n = &self->nt[off];
1608 1608 n->children[nt_level(oldnode, ++level)] = v;
1609 1609 if (level > self->ntdepth)
1610 1610 self->ntdepth = level;
1611 1611 self->ntsplits += 1;
1612 1612 } else {
1613 1613 level += 1;
1614 1614 off = v;
1615 1615 }
1616 1616 }
1617 1617
1618 1618 return -1;
1619 1619 }
1620 1620
1621 1621 static int nt_init(indexObject *self)
1622 1622 {
1623 1623 if (self->nt == NULL) {
1624 1624 if ((size_t)self->raw_length > INT_MAX / sizeof(nodetree)) {
1625 1625 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1626 1626 return -1;
1627 1627 }
1628 1628 self->ntcapacity = self->raw_length < 4
1629 1629 ? 4 : (int)self->raw_length / 2;
1630 1630
1631 1631 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1632 1632 if (self->nt == NULL) {
1633 1633 PyErr_NoMemory();
1634 1634 return -1;
1635 1635 }
1636 1636 self->ntlength = 1;
1637 1637 self->ntrev = (int)index_length(self) - 1;
1638 1638 self->ntlookups = 1;
1639 1639 self->ntmisses = 0;
1640 1640 if (nt_insert(self, nullid, INT_MAX) == -1)
1641 1641 return -1;
1642 1642 }
1643 1643 return 0;
1644 1644 }
1645 1645
1646 1646 /*
1647 1647 * Return values:
1648 1648 *
1649 1649 * -3: error (exception set)
1650 1650 * -2: not found (no exception set)
1651 1651 * rest: valid rev
1652 1652 */
1653 1653 static int index_find_node(indexObject *self,
1654 1654 const char *node, Py_ssize_t nodelen)
1655 1655 {
1656 1656 int rev;
1657 1657
1658 1658 self->ntlookups++;
1659 1659 rev = nt_find(self, node, nodelen, 0);
1660 1660 if (rev >= -1)
1661 1661 return rev;
1662 1662
1663 1663 if (nt_init(self) == -1)
1664 1664 return -3;
1665 1665
1666 1666 /*
1667 1667 * For the first handful of lookups, we scan the entire index,
1668 1668 * and cache only the matching nodes. This optimizes for cases
1669 1669 * like "hg tip", where only a few nodes are accessed.
1670 1670 *
1671 1671 * After that, we cache every node we visit, using a single
1672 1672 * scan amortized over multiple lookups. This gives the best
1673 1673 * bulk performance, e.g. for "hg log".
1674 1674 */
1675 1675 if (self->ntmisses++ < 4) {
1676 1676 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1677 1677 const char *n = index_node(self, rev);
1678 1678 if (n == NULL)
1679 1679 return -2;
1680 1680 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1681 1681 if (nt_insert(self, n, rev) == -1)
1682 1682 return -3;
1683 1683 break;
1684 1684 }
1685 1685 }
1686 1686 } else {
1687 1687 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1688 1688 const char *n = index_node(self, rev);
1689 1689 if (n == NULL) {
1690 1690 self->ntrev = rev + 1;
1691 1691 return -2;
1692 1692 }
1693 1693 if (nt_insert(self, n, rev) == -1) {
1694 1694 self->ntrev = rev + 1;
1695 1695 return -3;
1696 1696 }
1697 1697 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1698 1698 break;
1699 1699 }
1700 1700 }
1701 1701 self->ntrev = rev;
1702 1702 }
1703 1703
1704 1704 if (rev >= 0)
1705 1705 return rev;
1706 1706 return -2;
1707 1707 }
1708 1708
1709 1709 static void raise_revlog_error(void)
1710 1710 {
1711 1711 PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
1712 1712
1713 1713 mod = PyImport_ImportModule("mercurial.error");
1714 1714 if (mod == NULL) {
1715 1715 goto cleanup;
1716 1716 }
1717 1717
1718 1718 dict = PyModule_GetDict(mod);
1719 1719 if (dict == NULL) {
1720 1720 goto cleanup;
1721 1721 }
1722 1722 Py_INCREF(dict);
1723 1723
1724 1724 errclass = PyDict_GetItemString(dict, "RevlogError");
1725 1725 if (errclass == NULL) {
1726 1726 PyErr_SetString(PyExc_SystemError,
1727 1727 "could not find RevlogError");
1728 1728 goto cleanup;
1729 1729 }
1730 1730
1731 1731 /* value of exception is ignored by callers */
1732 1732 PyErr_SetString(errclass, "RevlogError");
1733 1733
1734 1734 cleanup:
1735 1735 Py_XDECREF(dict);
1736 1736 Py_XDECREF(mod);
1737 1737 }
1738 1738
1739 1739 static PyObject *index_getitem(indexObject *self, PyObject *value)
1740 1740 {
1741 1741 char *node;
1742 1742 Py_ssize_t nodelen;
1743 1743 int rev;
1744 1744
1745 1745 if (PyInt_Check(value))
1746 1746 return index_get(self, PyInt_AS_LONG(value));
1747 1747
1748 1748 if (node_check(value, &node, &nodelen) == -1)
1749 1749 return NULL;
1750 1750 rev = index_find_node(self, node, nodelen);
1751 1751 if (rev >= -1)
1752 1752 return PyInt_FromLong(rev);
1753 1753 if (rev == -2)
1754 1754 raise_revlog_error();
1755 1755 return NULL;
1756 1756 }
1757 1757
1758 1758 static int nt_partialmatch(indexObject *self, const char *node,
1759 1759 Py_ssize_t nodelen)
1760 1760 {
1761 1761 int rev;
1762 1762
1763 1763 if (nt_init(self) == -1)
1764 1764 return -3;
1765 1765
1766 1766 if (self->ntrev > 0) {
1767 1767 /* ensure that the radix tree is fully populated */
1768 1768 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1769 1769 const char *n = index_node(self, rev);
1770 1770 if (n == NULL)
1771 1771 return -2;
1772 1772 if (nt_insert(self, n, rev) == -1)
1773 1773 return -3;
1774 1774 }
1775 1775 self->ntrev = rev;
1776 1776 }
1777 1777
1778 1778 return nt_find(self, node, nodelen, 1);
1779 1779 }
1780 1780
1781 1781 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1782 1782 {
1783 1783 const char *fullnode;
1784 1784 int nodelen;
1785 1785 char *node;
1786 1786 int rev, i;
1787 1787
1788 1788 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1789 1789 return NULL;
1790 1790
1791 1791 if (nodelen < 4) {
1792 1792 PyErr_SetString(PyExc_ValueError, "key too short");
1793 1793 return NULL;
1794 1794 }
1795 1795
1796 1796 if (nodelen > 40) {
1797 1797 PyErr_SetString(PyExc_ValueError, "key too long");
1798 1798 return NULL;
1799 1799 }
1800 1800
1801 1801 for (i = 0; i < nodelen; i++)
1802 1802 hexdigit(node, i);
1803 1803 if (PyErr_Occurred()) {
1804 1804 /* input contains non-hex characters */
1805 1805 PyErr_Clear();
1806 1806 Py_RETURN_NONE;
1807 1807 }
1808 1808
1809 1809 rev = nt_partialmatch(self, node, nodelen);
1810 1810
1811 1811 switch (rev) {
1812 1812 case -4:
1813 1813 raise_revlog_error();
1814 1814 case -3:
1815 1815 return NULL;
1816 1816 case -2:
1817 1817 Py_RETURN_NONE;
1818 1818 case -1:
1819 1819 return PyString_FromStringAndSize(nullid, 20);
1820 1820 }
1821 1821
1822 1822 fullnode = index_node(self, rev);
1823 1823 if (fullnode == NULL) {
1824 1824 PyErr_Format(PyExc_IndexError,
1825 1825 "could not access rev %d", rev);
1826 1826 return NULL;
1827 1827 }
1828 1828 return PyString_FromStringAndSize(fullnode, 20);
1829 1829 }
1830 1830
1831 1831 static PyObject *index_m_get(indexObject *self, PyObject *args)
1832 1832 {
1833 1833 Py_ssize_t nodelen;
1834 1834 PyObject *val;
1835 1835 char *node;
1836 1836 int rev;
1837 1837
1838 1838 if (!PyArg_ParseTuple(args, "O", &val))
1839 1839 return NULL;
1840 1840 if (node_check(val, &node, &nodelen) == -1)
1841 1841 return NULL;
1842 1842 rev = index_find_node(self, node, nodelen);
1843 if (rev == -3)
1843 if (rev == -3)
1844 1844 return NULL;
1845 1845 if (rev == -2)
1846 1846 Py_RETURN_NONE;
1847 1847 return PyInt_FromLong(rev);
1848 1848 }
1849 1849
1850 1850 static int index_contains(indexObject *self, PyObject *value)
1851 1851 {
1852 1852 char *node;
1853 1853 Py_ssize_t nodelen;
1854 1854
1855 1855 if (PyInt_Check(value)) {
1856 1856 long rev = PyInt_AS_LONG(value);
1857 1857 return rev >= -1 && rev < index_length(self);
1858 1858 }
1859 1859
1860 1860 if (node_check(value, &node, &nodelen) == -1)
1861 1861 return -1;
1862 1862
1863 1863 switch (index_find_node(self, node, nodelen)) {
1864 1864 case -3:
1865 1865 return -1;
1866 1866 case -2:
1867 1867 return 0;
1868 1868 default:
1869 1869 return 1;
1870 1870 }
1871 1871 }
1872 1872
1873 1873 typedef uint64_t bitmask;
1874 1874
1875 1875 /*
1876 1876 * Given a disjoint set of revs, return all candidates for the
1877 1877 * greatest common ancestor. In revset notation, this is the set
1878 1878 * "heads(::a and ::b and ...)"
1879 1879 */
1880 1880 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1881 1881 int revcount)
1882 1882 {
1883 1883 const bitmask allseen = (1ull << revcount) - 1;
1884 1884 const bitmask poison = 1ull << revcount;
1885 1885 PyObject *gca = PyList_New(0);
1886 1886 int i, v, interesting;
1887 1887 int maxrev = -1;
1888 1888 bitmask sp;
1889 1889 bitmask *seen;
1890 1890
1891 1891 if (gca == NULL)
1892 1892 return PyErr_NoMemory();
1893 1893
1894 1894 for (i = 0; i < revcount; i++) {
1895 1895 if (revs[i] > maxrev)
1896 1896 maxrev = revs[i];
1897 1897 }
1898 1898
1899 1899 seen = calloc(sizeof(*seen), maxrev + 1);
1900 1900 if (seen == NULL) {
1901 1901 Py_DECREF(gca);
1902 1902 return PyErr_NoMemory();
1903 1903 }
1904 1904
1905 1905 for (i = 0; i < revcount; i++)
1906 1906 seen[revs[i]] = 1ull << i;
1907 1907
1908 1908 interesting = revcount;
1909 1909
1910 1910 for (v = maxrev; v >= 0 && interesting; v--) {
1911 1911 bitmask sv = seen[v];
1912 1912 int parents[2];
1913 1913
1914 1914 if (!sv)
1915 1915 continue;
1916 1916
1917 1917 if (sv < poison) {
1918 1918 interesting -= 1;
1919 1919 if (sv == allseen) {
1920 1920 PyObject *obj = PyInt_FromLong(v);
1921 1921 if (obj == NULL)
1922 1922 goto bail;
1923 1923 if (PyList_Append(gca, obj) == -1) {
1924 1924 Py_DECREF(obj);
1925 1925 goto bail;
1926 1926 }
1927 1927 sv |= poison;
1928 1928 for (i = 0; i < revcount; i++) {
1929 1929 if (revs[i] == v)
1930 1930 goto done;
1931 1931 }
1932 1932 }
1933 1933 }
1934 1934 if (index_get_parents(self, v, parents, maxrev) < 0)
1935 1935 goto bail;
1936 1936
1937 1937 for (i = 0; i < 2; i++) {
1938 1938 int p = parents[i];
1939 1939 if (p == -1)
1940 1940 continue;
1941 1941 sp = seen[p];
1942 1942 if (sv < poison) {
1943 1943 if (sp == 0) {
1944 1944 seen[p] = sv;
1945 1945 interesting++;
1946 1946 }
1947 1947 else if (sp != sv)
1948 1948 seen[p] |= sv;
1949 1949 } else {
1950 1950 if (sp && sp < poison)
1951 1951 interesting--;
1952 1952 seen[p] = sv;
1953 1953 }
1954 1954 }
1955 1955 }
1956 1956
1957 1957 done:
1958 1958 free(seen);
1959 1959 return gca;
1960 1960 bail:
1961 1961 free(seen);
1962 1962 Py_XDECREF(gca);
1963 1963 return NULL;
1964 1964 }
1965 1965
1966 1966 /*
1967 1967 * Given a disjoint set of revs, return the subset with the longest
1968 1968 * path to the root.
1969 1969 */
1970 1970 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1971 1971 {
1972 1972 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1973 1973 static const Py_ssize_t capacity = 24;
1974 1974 int *depth, *interesting = NULL;
1975 1975 int i, j, v, ninteresting;
1976 1976 PyObject *dict = NULL, *keys = NULL;
1977 1977 long *seen = NULL;
1978 1978 int maxrev = -1;
1979 1979 long final;
1980 1980
1981 1981 if (revcount > capacity) {
1982 1982 PyErr_Format(PyExc_OverflowError,
1983 1983 "bitset size (%ld) > capacity (%ld)",
1984 1984 (long)revcount, (long)capacity);
1985 1985 return NULL;
1986 1986 }
1987 1987
1988 1988 for (i = 0; i < revcount; i++) {
1989 1989 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1990 1990 if (n > maxrev)
1991 1991 maxrev = n;
1992 1992 }
1993 1993
1994 1994 depth = calloc(sizeof(*depth), maxrev + 1);
1995 1995 if (depth == NULL)
1996 1996 return PyErr_NoMemory();
1997 1997
1998 1998 seen = calloc(sizeof(*seen), maxrev + 1);
1999 1999 if (seen == NULL) {
2000 2000 PyErr_NoMemory();
2001 2001 goto bail;
2002 2002 }
2003 2003
2004 2004 interesting = calloc(sizeof(*interesting), 2 << revcount);
2005 2005 if (interesting == NULL) {
2006 2006 PyErr_NoMemory();
2007 2007 goto bail;
2008 2008 }
2009 2009
2010 2010 if (PyList_Sort(revs) == -1)
2011 2011 goto bail;
2012 2012
2013 2013 for (i = 0; i < revcount; i++) {
2014 2014 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
2015 2015 long b = 1l << i;
2016 2016 depth[n] = 1;
2017 2017 seen[n] = b;
2018 2018 interesting[b] = 1;
2019 2019 }
2020 2020
2021 2021 ninteresting = (int)revcount;
2022 2022
2023 2023 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
2024 2024 int dv = depth[v];
2025 2025 int parents[2];
2026 2026 long sv;
2027 2027
2028 2028 if (dv == 0)
2029 2029 continue;
2030 2030
2031 2031 sv = seen[v];
2032 2032 if (index_get_parents(self, v, parents, maxrev) < 0)
2033 2033 goto bail;
2034 2034
2035 2035 for (i = 0; i < 2; i++) {
2036 2036 int p = parents[i];
2037 2037 long sp;
2038 2038 int dp;
2039 2039
2040 2040 if (p == -1)
2041 2041 continue;
2042 2042
2043 2043 dp = depth[p];
2044 2044 sp = seen[p];
2045 2045 if (dp <= dv) {
2046 2046 depth[p] = dv + 1;
2047 2047 if (sp != sv) {
2048 2048 interesting[sv] += 1;
2049 2049 seen[p] = sv;
2050 2050 if (sp) {
2051 2051 interesting[sp] -= 1;
2052 2052 if (interesting[sp] == 0)
2053 2053 ninteresting -= 1;
2054 2054 }
2055 2055 }
2056 2056 }
2057 2057 else if (dv == dp - 1) {
2058 2058 long nsp = sp | sv;
2059 2059 if (nsp == sp)
2060 2060 continue;
2061 2061 seen[p] = nsp;
2062 2062 interesting[sp] -= 1;
2063 2063 if (interesting[sp] == 0 && interesting[nsp] > 0)
2064 2064 ninteresting -= 1;
2065 2065 interesting[nsp] += 1;
2066 2066 }
2067 2067 }
2068 2068 interesting[sv] -= 1;
2069 2069 if (interesting[sv] == 0)
2070 2070 ninteresting -= 1;
2071 2071 }
2072 2072
2073 2073 final = 0;
2074 2074 j = ninteresting;
2075 2075 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
2076 2076 if (interesting[i] == 0)
2077 2077 continue;
2078 2078 final |= i;
2079 2079 j -= 1;
2080 2080 }
2081 2081 if (final == 0) {
2082 2082 keys = PyList_New(0);
2083 2083 goto bail;
2084 2084 }
2085 2085
2086 2086 dict = PyDict_New();
2087 2087 if (dict == NULL)
2088 2088 goto bail;
2089 2089
2090 2090 for (i = 0; i < revcount; i++) {
2091 2091 PyObject *key;
2092 2092
2093 2093 if ((final & (1 << i)) == 0)
2094 2094 continue;
2095 2095
2096 2096 key = PyList_GET_ITEM(revs, i);
2097 2097 Py_INCREF(key);
2098 2098 Py_INCREF(Py_None);
2099 2099 if (PyDict_SetItem(dict, key, Py_None) == -1) {
2100 2100 Py_DECREF(key);
2101 2101 Py_DECREF(Py_None);
2102 2102 goto bail;
2103 2103 }
2104 2104 }
2105 2105
2106 2106 keys = PyDict_Keys(dict);
2107 2107
2108 2108 bail:
2109 2109 free(depth);
2110 2110 free(seen);
2111 2111 free(interesting);
2112 2112 Py_XDECREF(dict);
2113 2113
2114 2114 return keys;
2115 2115 }
2116 2116
2117 2117 /*
2118 2118 * Given a (possibly overlapping) set of revs, return all the
2119 2119 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
2120 2120 */
2121 2121 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
2122 2122 {
2123 2123 PyObject *ret = NULL;
2124 2124 Py_ssize_t argcount, i, len;
2125 2125 bitmask repeat = 0;
2126 2126 int revcount = 0;
2127 2127 int *revs;
2128 2128
2129 2129 argcount = PySequence_Length(args);
2130 2130 revs = malloc(argcount * sizeof(*revs));
2131 2131 if (argcount > 0 && revs == NULL)
2132 2132 return PyErr_NoMemory();
2133 2133 len = index_length(self) - 1;
2134 2134
2135 2135 for (i = 0; i < argcount; i++) {
2136 2136 static const int capacity = 24;
2137 2137 PyObject *obj = PySequence_GetItem(args, i);
2138 2138 bitmask x;
2139 2139 long val;
2140 2140
2141 2141 if (!PyInt_Check(obj)) {
2142 2142 PyErr_SetString(PyExc_TypeError,
2143 2143 "arguments must all be ints");
2144 2144 Py_DECREF(obj);
2145 2145 goto bail;
2146 2146 }
2147 2147 val = PyInt_AsLong(obj);
2148 2148 Py_DECREF(obj);
2149 2149 if (val == -1) {
2150 2150 ret = PyList_New(0);
2151 2151 goto done;
2152 2152 }
2153 2153 if (val < 0 || val >= len) {
2154 2154 PyErr_SetString(PyExc_IndexError,
2155 2155 "index out of range");
2156 2156 goto bail;
2157 2157 }
2158 2158 /* this cheesy bloom filter lets us avoid some more
2159 2159 * expensive duplicate checks in the common set-is-disjoint
2160 2160 * case */
2161 2161 x = 1ull << (val & 0x3f);
2162 2162 if (repeat & x) {
2163 2163 int k;
2164 2164 for (k = 0; k < revcount; k++) {
2165 2165 if (val == revs[k])
2166 2166 goto duplicate;
2167 2167 }
2168 2168 }
2169 2169 else repeat |= x;
2170 2170 if (revcount >= capacity) {
2171 2171 PyErr_Format(PyExc_OverflowError,
2172 2172 "bitset size (%d) > capacity (%d)",
2173 2173 revcount, capacity);
2174 2174 goto bail;
2175 2175 }
2176 2176 revs[revcount++] = (int)val;
2177 2177 duplicate:;
2178 2178 }
2179 2179
2180 2180 if (revcount == 0) {
2181 2181 ret = PyList_New(0);
2182 2182 goto done;
2183 2183 }
2184 2184 if (revcount == 1) {
2185 2185 PyObject *obj;
2186 2186 ret = PyList_New(1);
2187 2187 if (ret == NULL)
2188 2188 goto bail;
2189 2189 obj = PyInt_FromLong(revs[0]);
2190 2190 if (obj == NULL)
2191 2191 goto bail;
2192 2192 PyList_SET_ITEM(ret, 0, obj);
2193 2193 goto done;
2194 2194 }
2195 2195
2196 2196 ret = find_gca_candidates(self, revs, revcount);
2197 2197 if (ret == NULL)
2198 2198 goto bail;
2199 2199
2200 2200 done:
2201 2201 free(revs);
2202 2202 return ret;
2203 2203
2204 2204 bail:
2205 2205 free(revs);
2206 2206 Py_XDECREF(ret);
2207 2207 return NULL;
2208 2208 }
2209 2209
2210 2210 /*
2211 2211 * Given a (possibly overlapping) set of revs, return the greatest
2212 2212 * common ancestors: those with the longest path to the root.
2213 2213 */
2214 2214 static PyObject *index_ancestors(indexObject *self, PyObject *args)
2215 2215 {
2216 2216 PyObject *ret;
2217 2217 PyObject *gca = index_commonancestorsheads(self, args);
2218 2218 if (gca == NULL)
2219 2219 return NULL;
2220 2220
2221 2221 if (PyList_GET_SIZE(gca) <= 1) {
2222 2222 return gca;
2223 2223 }
2224 2224
2225 2225 ret = find_deepest(self, gca);
2226 2226 Py_DECREF(gca);
2227 2227 return ret;
2228 2228 }
2229 2229
2230 2230 /*
2231 2231 * Invalidate any trie entries introduced by added revs.
2232 2232 */
2233 2233 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
2234 2234 {
2235 2235 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
2236 2236
2237 2237 for (i = start; i < len; i++) {
2238 2238 PyObject *tuple = PyList_GET_ITEM(self->added, i);
2239 2239 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
2240 2240
2241 2241 nt_insert(self, PyString_AS_STRING(node), -1);
2242 2242 }
2243 2243
2244 2244 if (start == 0)
2245 2245 Py_CLEAR(self->added);
2246 2246 }
2247 2247
2248 2248 /*
2249 2249 * Delete a numeric range of revs, which must be at the end of the
2250 2250 * range, but exclude the sentinel nullid entry.
2251 2251 */
2252 2252 static int index_slice_del(indexObject *self, PyObject *item)
2253 2253 {
2254 2254 Py_ssize_t start, stop, step, slicelength;
2255 2255 Py_ssize_t length = index_length(self);
2256 2256 int ret = 0;
2257 2257
2258 2258 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
2259 2259 &start, &stop, &step, &slicelength) < 0)
2260 2260 return -1;
2261 2261
2262 2262 if (slicelength <= 0)
2263 2263 return 0;
2264 2264
2265 2265 if ((step < 0 && start < stop) || (step > 0 && start > stop))
2266 2266 stop = start;
2267 2267
2268 2268 if (step < 0) {
2269 2269 stop = start + 1;
2270 2270 start = stop + step*(slicelength - 1) - 1;
2271 2271 step = -step;
2272 2272 }
2273 2273
2274 2274 if (step != 1) {
2275 2275 PyErr_SetString(PyExc_ValueError,
2276 2276 "revlog index delete requires step size of 1");
2277 2277 return -1;
2278 2278 }
2279 2279
2280 2280 if (stop != length - 1) {
2281 2281 PyErr_SetString(PyExc_IndexError,
2282 2282 "revlog index deletion indices are invalid");
2283 2283 return -1;
2284 2284 }
2285 2285
2286 2286 if (start < self->length - 1) {
2287 2287 if (self->nt) {
2288 2288 Py_ssize_t i;
2289 2289
2290 2290 for (i = start + 1; i < self->length - 1; i++) {
2291 2291 const char *node = index_node(self, i);
2292 2292
2293 2293 if (node)
2294 2294 nt_insert(self, node, -1);
2295 2295 }
2296 2296 if (self->added)
2297 2297 nt_invalidate_added(self, 0);
2298 2298 if (self->ntrev > start)
2299 2299 self->ntrev = (int)start;
2300 2300 }
2301 2301 self->length = start + 1;
2302 2302 if (start < self->raw_length) {
2303 2303 if (self->cache) {
2304 2304 Py_ssize_t i;
2305 2305 for (i = start; i < self->raw_length; i++)
2306 2306 Py_CLEAR(self->cache[i]);
2307 2307 }
2308 2308 self->raw_length = start;
2309 2309 }
2310 2310 goto done;
2311 2311 }
2312 2312
2313 2313 if (self->nt) {
2314 2314 nt_invalidate_added(self, start - self->length + 1);
2315 2315 if (self->ntrev > start)
2316 2316 self->ntrev = (int)start;
2317 2317 }
2318 2318 if (self->added)
2319 2319 ret = PyList_SetSlice(self->added, start - self->length + 1,
2320 2320 PyList_GET_SIZE(self->added), NULL);
2321 2321 done:
2322 2322 Py_CLEAR(self->headrevs);
2323 2323 return ret;
2324 2324 }
2325 2325
2326 2326 /*
2327 2327 * Supported ops:
2328 2328 *
2329 2329 * slice deletion
2330 2330 * string assignment (extend node->rev mapping)
2331 2331 * string deletion (shrink node->rev mapping)
2332 2332 */
2333 2333 static int index_assign_subscript(indexObject *self, PyObject *item,
2334 2334 PyObject *value)
2335 2335 {
2336 2336 char *node;
2337 2337 Py_ssize_t nodelen;
2338 2338 long rev;
2339 2339
2340 2340 if (PySlice_Check(item) && value == NULL)
2341 2341 return index_slice_del(self, item);
2342 2342
2343 2343 if (node_check(item, &node, &nodelen) == -1)
2344 2344 return -1;
2345 2345
2346 2346 if (value == NULL)
2347 2347 return self->nt ? nt_insert(self, node, -1) : 0;
2348 2348 rev = PyInt_AsLong(value);
2349 2349 if (rev > INT_MAX || rev < 0) {
2350 2350 if (!PyErr_Occurred())
2351 2351 PyErr_SetString(PyExc_ValueError, "rev out of range");
2352 2352 return -1;
2353 2353 }
2354 2354
2355 2355 if (nt_init(self) == -1)
2356 2356 return -1;
2357 2357 return nt_insert(self, node, (int)rev);
2358 2358 }
2359 2359
2360 2360 /*
2361 2361 * Find all RevlogNG entries in an index that has inline data. Update
2362 2362 * the optional "offsets" table with those entries.
2363 2363 */
2364 2364 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
2365 2365 {
2366 2366 const char *data = PyString_AS_STRING(self->data);
2367 2367 Py_ssize_t pos = 0;
2368 2368 Py_ssize_t end = PyString_GET_SIZE(self->data);
2369 2369 long incr = v1_hdrsize;
2370 2370 Py_ssize_t len = 0;
2371 2371
2372 2372 while (pos + v1_hdrsize <= end && pos >= 0) {
2373 2373 uint32_t comp_len;
2374 2374 /* 3rd element of header is length of compressed inline data */
2375 2375 comp_len = getbe32(data + pos + 8);
2376 2376 incr = v1_hdrsize + comp_len;
2377 2377 if (offsets)
2378 2378 offsets[len] = data + pos;
2379 2379 len++;
2380 2380 pos += incr;
2381 2381 }
2382 2382
2383 2383 if (pos != end) {
2384 2384 if (!PyErr_Occurred())
2385 2385 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2386 2386 return -1;
2387 2387 }
2388 2388
2389 2389 return len;
2390 2390 }
2391 2391
2392 2392 static int index_init(indexObject *self, PyObject *args)
2393 2393 {
2394 2394 PyObject *data_obj, *inlined_obj;
2395 2395 Py_ssize_t size;
2396 2396
2397 2397 /* Initialize before argument-checking to avoid index_dealloc() crash. */
2398 2398 self->raw_length = 0;
2399 2399 self->added = NULL;
2400 2400 self->cache = NULL;
2401 2401 self->data = NULL;
2402 2402 self->headrevs = NULL;
2403 2403 self->filteredrevs = Py_None;
2404 2404 Py_INCREF(Py_None);
2405 2405 self->nt = NULL;
2406 2406 self->offsets = NULL;
2407 2407
2408 2408 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
2409 2409 return -1;
2410 2410 if (!PyString_Check(data_obj)) {
2411 2411 PyErr_SetString(PyExc_TypeError, "data is not a string");
2412 2412 return -1;
2413 2413 }
2414 2414 size = PyString_GET_SIZE(data_obj);
2415 2415
2416 2416 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
2417 2417 self->data = data_obj;
2418 2418
2419 2419 self->ntlength = self->ntcapacity = 0;
2420 2420 self->ntdepth = self->ntsplits = 0;
2421 2421 self->ntlookups = self->ntmisses = 0;
2422 2422 self->ntrev = -1;
2423 2423 Py_INCREF(self->data);
2424 2424
2425 2425 if (self->inlined) {
2426 2426 Py_ssize_t len = inline_scan(self, NULL);
2427 2427 if (len == -1)
2428 2428 goto bail;
2429 2429 self->raw_length = len;
2430 2430 self->length = len + 1;
2431 2431 } else {
2432 2432 if (size % v1_hdrsize) {
2433 2433 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2434 2434 goto bail;
2435 2435 }
2436 2436 self->raw_length = size / v1_hdrsize;
2437 2437 self->length = self->raw_length + 1;
2438 2438 }
2439 2439
2440 2440 return 0;
2441 2441 bail:
2442 2442 return -1;
2443 2443 }
2444 2444
2445 2445 static PyObject *index_nodemap(indexObject *self)
2446 2446 {
2447 2447 Py_INCREF(self);
2448 2448 return (PyObject *)self;
2449 2449 }
2450 2450
2451 2451 static void index_dealloc(indexObject *self)
2452 2452 {
2453 2453 _index_clearcaches(self);
2454 2454 Py_XDECREF(self->filteredrevs);
2455 2455 Py_XDECREF(self->data);
2456 2456 Py_XDECREF(self->added);
2457 2457 PyObject_Del(self);
2458 2458 }
2459 2459
2460 2460 static PySequenceMethods index_sequence_methods = {
2461 2461 (lenfunc)index_length, /* sq_length */
2462 2462 0, /* sq_concat */
2463 2463 0, /* sq_repeat */
2464 2464 (ssizeargfunc)index_get, /* sq_item */
2465 2465 0, /* sq_slice */
2466 2466 0, /* sq_ass_item */
2467 2467 0, /* sq_ass_slice */
2468 2468 (objobjproc)index_contains, /* sq_contains */
2469 2469 };
2470 2470
2471 2471 static PyMappingMethods index_mapping_methods = {
2472 2472 (lenfunc)index_length, /* mp_length */
2473 2473 (binaryfunc)index_getitem, /* mp_subscript */
2474 2474 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
2475 2475 };
2476 2476
2477 2477 static PyMethodDef index_methods[] = {
2478 2478 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
2479 2479 "return the gca set of the given revs"},
2480 2480 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
2481 2481 METH_VARARGS,
2482 2482 "return the heads of the common ancestors of the given revs"},
2483 2483 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
2484 2484 "clear the index caches"},
2485 2485 {"get", (PyCFunction)index_m_get, METH_VARARGS,
2486 2486 "get an index entry"},
2487 2487 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
2488 2488 METH_VARARGS, "compute phases"},
2489 2489 {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS,
2490 2490 "reachableroots"},
2491 2491 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
2492 2492 "get head revisions"}, /* Can do filtering since 3.2 */
2493 2493 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
2494 2494 "get filtered head revisions"}, /* Can always do filtering */
2495 2495 {"insert", (PyCFunction)index_insert, METH_VARARGS,
2496 2496 "insert an index entry"},
2497 2497 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
2498 2498 "match a potentially ambiguous node ID"},
2499 2499 {"stats", (PyCFunction)index_stats, METH_NOARGS,
2500 2500 "stats for the index"},
2501 2501 {NULL} /* Sentinel */
2502 2502 };
2503 2503
2504 2504 static PyGetSetDef index_getset[] = {
2505 2505 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
2506 2506 {NULL} /* Sentinel */
2507 2507 };
2508 2508
2509 2509 static PyTypeObject indexType = {
2510 2510 PyObject_HEAD_INIT(NULL)
2511 2511 0, /* ob_size */
2512 2512 "parsers.index", /* tp_name */
2513 2513 sizeof(indexObject), /* tp_basicsize */
2514 2514 0, /* tp_itemsize */
2515 2515 (destructor)index_dealloc, /* tp_dealloc */
2516 2516 0, /* tp_print */
2517 2517 0, /* tp_getattr */
2518 2518 0, /* tp_setattr */
2519 2519 0, /* tp_compare */
2520 2520 0, /* tp_repr */
2521 2521 0, /* tp_as_number */
2522 2522 &index_sequence_methods, /* tp_as_sequence */
2523 2523 &index_mapping_methods, /* tp_as_mapping */
2524 2524 0, /* tp_hash */
2525 2525 0, /* tp_call */
2526 2526 0, /* tp_str */
2527 2527 0, /* tp_getattro */
2528 2528 0, /* tp_setattro */
2529 2529 0, /* tp_as_buffer */
2530 2530 Py_TPFLAGS_DEFAULT, /* tp_flags */
2531 2531 "revlog index", /* tp_doc */
2532 2532 0, /* tp_traverse */
2533 2533 0, /* tp_clear */
2534 2534 0, /* tp_richcompare */
2535 2535 0, /* tp_weaklistoffset */
2536 2536 0, /* tp_iter */
2537 2537 0, /* tp_iternext */
2538 2538 index_methods, /* tp_methods */
2539 2539 0, /* tp_members */
2540 2540 index_getset, /* tp_getset */
2541 2541 0, /* tp_base */
2542 2542 0, /* tp_dict */
2543 2543 0, /* tp_descr_get */
2544 2544 0, /* tp_descr_set */
2545 2545 0, /* tp_dictoffset */
2546 2546 (initproc)index_init, /* tp_init */
2547 2547 0, /* tp_alloc */
2548 2548 };
2549 2549
2550 2550 /*
2551 2551 * returns a tuple of the form (index, index, cache) with elements as
2552 2552 * follows:
2553 2553 *
2554 2554 * index: an index object that lazily parses RevlogNG records
2555 2555 * cache: if data is inlined, a tuple (index_file_content, 0), else None
2556 2556 *
2557 2557 * added complications are for backwards compatibility
2558 2558 */
2559 2559 static PyObject *parse_index2(PyObject *self, PyObject *args)
2560 2560 {
2561 2561 PyObject *tuple = NULL, *cache = NULL;
2562 2562 indexObject *idx;
2563 2563 int ret;
2564 2564
2565 2565 idx = PyObject_New(indexObject, &indexType);
2566 2566 if (idx == NULL)
2567 2567 goto bail;
2568 2568
2569 2569 ret = index_init(idx, args);
2570 2570 if (ret == -1)
2571 2571 goto bail;
2572 2572
2573 2573 if (idx->inlined) {
2574 2574 cache = Py_BuildValue("iO", 0, idx->data);
2575 2575 if (cache == NULL)
2576 2576 goto bail;
2577 2577 } else {
2578 2578 cache = Py_None;
2579 2579 Py_INCREF(cache);
2580 2580 }
2581 2581
2582 2582 tuple = Py_BuildValue("NN", idx, cache);
2583 2583 if (!tuple)
2584 2584 goto bail;
2585 2585 return tuple;
2586 2586
2587 2587 bail:
2588 2588 Py_XDECREF(idx);
2589 2589 Py_XDECREF(cache);
2590 2590 Py_XDECREF(tuple);
2591 2591 return NULL;
2592 2592 }
2593 2593
2594 2594 #define BUMPED_FIX 1
2595 2595 #define USING_SHA_256 2
2596 2596 #define FM1_HEADER_SIZE (4 + 8 + 2 + 2 + 1 + 1 + 1)
2597 2597
2598 2598 static PyObject *readshas(
2599 2599 const char *source, unsigned char num, Py_ssize_t hashwidth)
2600 2600 {
2601 2601 int i;
2602 2602 PyObject *list = PyTuple_New(num);
2603 2603 if (list == NULL) {
2604 2604 return NULL;
2605 2605 }
2606 2606 for (i = 0; i < num; i++) {
2607 2607 PyObject *hash = PyString_FromStringAndSize(source, hashwidth);
2608 2608 if (hash == NULL) {
2609 2609 Py_DECREF(list);
2610 2610 return NULL;
2611 2611 }
2612 2612 PyTuple_SET_ITEM(list, i, hash);
2613 2613 source += hashwidth;
2614 2614 }
2615 2615 return list;
2616 2616 }
2617 2617
2618 2618 static PyObject *fm1readmarker(const char *databegin, const char *dataend,
2619 2619 uint32_t *msize)
2620 2620 {
2621 2621 const char *data = databegin;
2622 2622 const char *meta;
2623 2623
2624 2624 double mtime;
2625 2625 int16_t tz;
2626 2626 uint16_t flags;
2627 2627 unsigned char nsuccs, nparents, nmetadata;
2628 2628 Py_ssize_t hashwidth = 20;
2629 2629
2630 2630 PyObject *prec = NULL, *parents = NULL, *succs = NULL;
2631 2631 PyObject *metadata = NULL, *ret = NULL;
2632 2632 int i;
2633 2633
2634 2634 if (data + FM1_HEADER_SIZE > dataend) {
2635 2635 goto overflow;
2636 2636 }
2637 2637
2638 2638 *msize = getbe32(data);
2639 2639 data += 4;
2640 2640 mtime = getbefloat64(data);
2641 2641 data += 8;
2642 2642 tz = getbeint16(data);
2643 2643 data += 2;
2644 2644 flags = getbeuint16(data);
2645 2645 data += 2;
2646 2646
2647 2647 if (flags & USING_SHA_256) {
2648 2648 hashwidth = 32;
2649 2649 }
2650 2650
2651 2651 nsuccs = (unsigned char)(*data++);
2652 2652 nparents = (unsigned char)(*data++);
2653 2653 nmetadata = (unsigned char)(*data++);
2654 2654
2655 2655 if (databegin + *msize > dataend) {
2656 2656 goto overflow;
2657 2657 }
2658 2658 dataend = databegin + *msize; /* narrow down to marker size */
2659 2659
2660 2660 if (data + hashwidth > dataend) {
2661 2661 goto overflow;
2662 2662 }
2663 2663 prec = PyString_FromStringAndSize(data, hashwidth);
2664 2664 data += hashwidth;
2665 2665 if (prec == NULL) {
2666 2666 goto bail;
2667 2667 }
2668 2668
2669 2669 if (data + nsuccs * hashwidth > dataend) {
2670 2670 goto overflow;
2671 2671 }
2672 2672 succs = readshas(data, nsuccs, hashwidth);
2673 2673 if (succs == NULL) {
2674 2674 goto bail;
2675 2675 }
2676 2676 data += nsuccs * hashwidth;
2677 2677
2678 2678 if (nparents == 1 || nparents == 2) {
2679 2679 if (data + nparents * hashwidth > dataend) {
2680 2680 goto overflow;
2681 2681 }
2682 2682 parents = readshas(data, nparents, hashwidth);
2683 2683 if (parents == NULL) {
2684 2684 goto bail;
2685 2685 }
2686 2686 data += nparents * hashwidth;
2687 2687 } else {
2688 2688 parents = Py_None;
2689 2689 }
2690 2690
2691 2691 if (data + 2 * nmetadata > dataend) {
2692 2692 goto overflow;
2693 2693 }
2694 2694 meta = data + (2 * nmetadata);
2695 2695 metadata = PyTuple_New(nmetadata);
2696 2696 if (metadata == NULL) {
2697 2697 goto bail;
2698 2698 }
2699 2699 for (i = 0; i < nmetadata; i++) {
2700 2700 PyObject *tmp, *left = NULL, *right = NULL;
2701 2701 Py_ssize_t leftsize = (unsigned char)(*data++);
2702 2702 Py_ssize_t rightsize = (unsigned char)(*data++);
2703 2703 if (meta + leftsize + rightsize > dataend) {
2704 2704 goto overflow;
2705 2705 }
2706 2706 left = PyString_FromStringAndSize(meta, leftsize);
2707 2707 meta += leftsize;
2708 2708 right = PyString_FromStringAndSize(meta, rightsize);
2709 2709 meta += rightsize;
2710 2710 tmp = PyTuple_New(2);
2711 2711 if (!left || !right || !tmp) {
2712 2712 Py_XDECREF(left);
2713 2713 Py_XDECREF(right);
2714 2714 Py_XDECREF(tmp);
2715 2715 goto bail;
2716 2716 }
2717 2717 PyTuple_SET_ITEM(tmp, 0, left);
2718 2718 PyTuple_SET_ITEM(tmp, 1, right);
2719 2719 PyTuple_SET_ITEM(metadata, i, tmp);
2720 2720 }
2721 2721 ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags,
2722 2722 metadata, mtime, (int)tz * 60, parents);
2723 2723 goto bail; /* return successfully */
2724 2724
2725 2725 overflow:
2726 2726 PyErr_SetString(PyExc_ValueError, "overflow in obsstore");
2727 2727 bail:
2728 2728 Py_XDECREF(prec);
2729 2729 Py_XDECREF(succs);
2730 2730 Py_XDECREF(metadata);
2731 2731 if (parents != Py_None)
2732 2732 Py_XDECREF(parents);
2733 2733 return ret;
2734 2734 }
2735 2735
2736 2736
2737 2737 static PyObject *fm1readmarkers(PyObject *self, PyObject *args) {
2738 2738 const char *data, *dataend;
2739 2739 int datalen;
2740 2740 Py_ssize_t offset, stop;
2741 2741 PyObject *markers = NULL;
2742 2742
2743 2743 if (!PyArg_ParseTuple(args, "s#nn", &data, &datalen, &offset, &stop)) {
2744 2744 return NULL;
2745 2745 }
2746 2746 dataend = data + datalen;
2747 2747 data += offset;
2748 2748 markers = PyList_New(0);
2749 2749 if (!markers) {
2750 2750 return NULL;
2751 2751 }
2752 2752 while (offset < stop) {
2753 2753 uint32_t msize;
2754 2754 int error;
2755 2755 PyObject *record = fm1readmarker(data, dataend, &msize);
2756 2756 if (!record) {
2757 2757 goto bail;
2758 2758 }
2759 2759 error = PyList_Append(markers, record);
2760 2760 Py_DECREF(record);
2761 2761 if (error) {
2762 2762 goto bail;
2763 2763 }
2764 2764 data += msize;
2765 2765 offset += msize;
2766 2766 }
2767 2767 return markers;
2768 2768 bail:
2769 2769 Py_DECREF(markers);
2770 2770 return NULL;
2771 2771 }
2772 2772
2773 2773 static char parsers_doc[] = "Efficient content parsing.";
2774 2774
2775 2775 PyObject *encodedir(PyObject *self, PyObject *args);
2776 2776 PyObject *pathencode(PyObject *self, PyObject *args);
2777 2777 PyObject *lowerencode(PyObject *self, PyObject *args);
2778 2778
2779 2779 static PyMethodDef methods[] = {
2780 2780 {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
2781 2781 {"nonnormalentries", nonnormalentries, METH_VARARGS,
2782 2782 "create a set containing non-normal entries of given dirstate\n"},
2783 2783 {"parse_manifest", parse_manifest, METH_VARARGS, "parse a manifest\n"},
2784 2784 {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
2785 2785 {"parse_index2", parse_index2, METH_VARARGS, "parse a revlog index\n"},
2786 2786 {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
2787 2787 {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
2788 2788 {"dict_new_presized", dict_new_presized, METH_VARARGS,
2789 2789 "construct a dict with an expected size\n"},
2790 2790 {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
2791 2791 "make file foldmap\n"},
2792 2792 {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
2793 2793 {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
2794 2794 {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
2795 2795 {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
2796 2796 "parse v1 obsolete markers\n"},
2797 2797 {NULL, NULL}
2798 2798 };
2799 2799
2800 2800 void dirs_module_init(PyObject *mod);
2801 2801 void manifest_module_init(PyObject *mod);
2802 2802
2803 2803 static void module_init(PyObject *mod)
2804 2804 {
2805 2805 /* This module constant has two purposes. First, it lets us unit test
2806 2806 * the ImportError raised without hard-coding any error text. This
2807 2807 * means we can change the text in the future without breaking tests,
2808 2808 * even across changesets without a recompile. Second, its presence
2809 2809 * can be used to determine whether the version-checking logic is
2810 2810 * present, which also helps in testing across changesets without a
2811 2811 * recompile. Note that this means the pure-Python version of parsers
2812 2812 * should not have this module constant. */
2813 2813 PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
2814 2814
2815 2815 dirs_module_init(mod);
2816 2816 manifest_module_init(mod);
2817 2817
2818 2818 indexType.tp_new = PyType_GenericNew;
2819 2819 if (PyType_Ready(&indexType) < 0 ||
2820 2820 PyType_Ready(&dirstateTupleType) < 0)
2821 2821 return;
2822 2822 Py_INCREF(&indexType);
2823 2823 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2824 2824 Py_INCREF(&dirstateTupleType);
2825 2825 PyModule_AddObject(mod, "dirstatetuple",
2826 2826 (PyObject *)&dirstateTupleType);
2827 2827
2828 2828 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2829 2829 -1, -1, -1, -1, nullid, 20);
2830 2830 if (nullentry)
2831 2831 PyObject_GC_UnTrack(nullentry);
2832 2832 }
2833 2833
2834 2834 static int check_python_version(void)
2835 2835 {
2836 2836 PyObject *sys = PyImport_ImportModule("sys"), *ver;
2837 2837 long hexversion;
2838 2838 if (!sys)
2839 2839 return -1;
2840 2840 ver = PyObject_GetAttrString(sys, "hexversion");
2841 2841 Py_DECREF(sys);
2842 2842 if (!ver)
2843 2843 return -1;
2844 2844 hexversion = PyInt_AsLong(ver);
2845 2845 Py_DECREF(ver);
2846 2846 /* sys.hexversion is a 32-bit number by default, so the -1 case
2847 2847 * should only occur in unusual circumstances (e.g. if sys.hexversion
2848 2848 * is manually set to an invalid value). */
2849 2849 if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
2850 2850 PyErr_Format(PyExc_ImportError, "%s: The Mercurial extension "
2851 2851 "modules were compiled with Python " PY_VERSION ", but "
2852 2852 "Mercurial is currently using Python with sys.hexversion=%ld: "
2853 2853 "Python %s\n at: %s", versionerrortext, hexversion,
2854 2854 Py_GetVersion(), Py_GetProgramFullPath());
2855 2855 return -1;
2856 2856 }
2857 2857 return 0;
2858 2858 }
2859 2859
2860 2860 #ifdef IS_PY3K
2861 2861 static struct PyModuleDef parsers_module = {
2862 2862 PyModuleDef_HEAD_INIT,
2863 2863 "parsers",
2864 2864 parsers_doc,
2865 2865 -1,
2866 2866 methods
2867 2867 };
2868 2868
2869 2869 PyMODINIT_FUNC PyInit_parsers(void)
2870 2870 {
2871 2871 PyObject *mod;
2872 2872
2873 2873 if (check_python_version() == -1)
2874 2874 return;
2875 2875 mod = PyModule_Create(&parsers_module);
2876 2876 module_init(mod);
2877 2877 return mod;
2878 2878 }
2879 2879 #else
2880 2880 PyMODINIT_FUNC initparsers(void)
2881 2881 {
2882 2882 PyObject *mod;
2883 2883
2884 2884 if (check_python_version() == -1)
2885 2885 return;
2886 2886 mod = Py_InitModule3("parsers", methods, parsers_doc);
2887 2887 module_init(mod);
2888 2888 }
2889 2889 #endif
General Comments 0
You need to be logged in to leave comments. Login now