?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
gitweb: Fix handling of non-ASCII characters in inserted HTML files
[git/jnareb-git/bp-gitweb.git] / http-walker.c
blob7271c7d19d3b08ef76c9c868ae57f6872f6186d0
1 #include "cache.h"
2 #include "commit.h"
3 #include "pack.h"
4 #include "walker.h"
5 #include "http.h"
6
7 #define PREV_BUF_SIZE 4096
8 #define RANGE_HEADER_SIZE 30
9
10 struct alt_base
11 {
12 char *base;
13 int got_indices;
14 struct packed_git *packs;
15 struct alt_base *next;
16 };
17
18 enum object_request_state {
19 WAITING,
20 ABORTED,
21 ACTIVE,
22 COMPLETE,
23 };
24
25 struct object_request
26 {
27 struct walker *walker;
28 unsigned char sha1[20];
29 struct alt_base *repo;
30 char *url;
31 char filename[PATH_MAX];
32 char tmpfile[PATH_MAX];
33 int local;
34 enum object_request_state state;
35 CURLcode curl_result;
36 char errorstr[CURL_ERROR_SIZE];
37 long http_code;
38 unsigned char real_sha1[20];
39 git_SHA_CTX c;
40 z_stream stream;
41 int zret;
42 int rename;
43 struct active_request_slot *slot;
44 struct object_request *next;
45 };
46
47 struct alternates_request {
48 struct walker *walker;
49 const char *base;
50 char *url;
51 struct strbuf *buffer;
52 struct active_request_slot *slot;
53 int http_specific;
54 };
55
56 struct walker_data {
57 const char *url;
58 int got_alternates;
59 struct alt_base *alt;
60 struct curl_slist *no_pragma_header;
61 };
62
63 static struct object_request *object_queue_head;
64
65 static size_t fwrite_sha1_file(void *ptr, size_t eltsize, size_t nmemb,
66 void *data)
67 {
68 unsigned char expn[4096];
69 size_t size = eltsize * nmemb;
70 int posn = 0;
71 struct object_request *obj_req = (struct object_request *)data;
72 do {
73 ssize_t retval = xwrite(obj_req->local,
74 (char *) ptr + posn, size - posn);
75 if (retval < 0)
76 return posn;
77 posn += retval;
78 } while (posn < size);
79
80 obj_req->stream.avail_in = size;
81 obj_req->stream.next_in = ptr;
82 do {
83 obj_req->stream.next_out = expn;
84 obj_req->stream.avail_out = sizeof(expn);
85 obj_req->zret = inflate(&obj_req->stream, Z_SYNC_FLUSH);
86 git_SHA1_Update(&obj_req->c, expn,
87 sizeof(expn) - obj_req->stream.avail_out);
88 } while (obj_req->stream.avail_in && obj_req->zret == Z_OK);
89 data_received++;
90 return size;
91 }
92
93 static void fetch_alternates(struct walker *walker, const char *base);
94
95 static void process_object_response(void *callback_data);
96
97 static void start_object_request(struct walker *walker,
98 struct object_request *obj_req)
99 {
100 char *hex = sha1_to_hex(obj_req->sha1);
101 char prevfile[PATH_MAX];
102 char *url;
103 char *posn;
104 int prevlocal;
105 unsigned char prev_buf[PREV_BUF_SIZE];
106 ssize_t prev_read = 0;
107 long prev_posn = 0;
108 char range[RANGE_HEADER_SIZE];
109 struct curl_slist *range_header = NULL;
110 struct active_request_slot *slot;
111 struct walker_data *data = walker->data;
112
113 snprintf(prevfile, sizeof(prevfile), "%s.prev", obj_req->filename);
114 unlink(prevfile);
115 rename(obj_req->tmpfile, prevfile);
116 unlink(obj_req->tmpfile);
117
118 if (obj_req->local != -1)
119 error("fd leakage in start: %d", obj_req->local);
120 obj_req->local = open(obj_req->tmpfile,
121 O_WRONLY | O_CREAT | O_EXCL, 0666);
122 /* This could have failed due to the "lazy directory creation";
123 * try to mkdir the last path component.
124 */
125 if (obj_req->local < 0 && errno == ENOENT) {
126 char *dir = strrchr(obj_req->tmpfile, '/');
127 if (dir) {
128 *dir = 0;
129 mkdir(obj_req->tmpfile, 0777);
130 *dir = '/';
131 }
132 obj_req->local = open(obj_req->tmpfile,
133 O_WRONLY | O_CREAT | O_EXCL, 0666);
134 }
135
136 if (obj_req->local < 0) {
137 obj_req->state = ABORTED;
138 error("Couldn't create temporary file %s for %s: %s",
139 obj_req->tmpfile, obj_req->filename, strerror(errno));
140 return;
141 }
142
143 memset(&obj_req->stream, 0, sizeof(obj_req->stream));
144
145 inflateInit(&obj_req->stream);
146
147 git_SHA1_Init(&obj_req->c);
148
149 url = xmalloc(strlen(obj_req->repo->base) + 51);
150 obj_req->url = xmalloc(strlen(obj_req->repo->base) + 51);
151 strcpy(url, obj_req->repo->base);
152 posn = url + strlen(obj_req->repo->base);
153 strcpy(posn, "/objects/");
154 posn += 9;
155 memcpy(posn, hex, 2);
156 posn += 2;
157 *(posn++) = '/';
158 strcpy(posn, hex + 2);
159 strcpy(obj_req->url, url);
160
161 /* If a previous temp file is present, process what was already
162 fetched. */
163 prevlocal = open(prevfile, O_RDONLY);
164 if (prevlocal != -1) {
165 do {
166 prev_read = xread(prevlocal, prev_buf, PREV_BUF_SIZE);
167 if (prev_read>0) {
168 if (fwrite_sha1_file(prev_buf,
169 1,
170 prev_read,
171 obj_req) == prev_read) {
172 prev_posn += prev_read;
173 } else {
174 prev_read = -1;
175 }
176 }
177 } while (prev_read > 0);
178 close(prevlocal);
179 }
180 unlink(prevfile);
181
182 /* Reset inflate/SHA1 if there was an error reading the previous temp
183 file; also rewind to the beginning of the local file. */
184 if (prev_read == -1) {
185 memset(&obj_req->stream, 0, sizeof(obj_req->stream));
186 inflateInit(&obj_req->stream);
187 git_SHA1_Init(&obj_req->c);
188 if (prev_posn>0) {
189 prev_posn = 0;
190 lseek(obj_req->local, 0, SEEK_SET);
191 ftruncate(obj_req->local, 0);
192 }
193 }
194
195 slot = get_active_slot();
196 slot->callback_func = process_object_response;
197 slot->callback_data = obj_req;
198 obj_req->slot = slot;
199
200 curl_easy_setopt(slot->curl, CURLOPT_FILE, obj_req);
201 curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite_sha1_file);
202 curl_easy_setopt(slot->curl, CURLOPT_ERRORBUFFER, obj_req->errorstr);
203 curl_easy_setopt(slot->curl, CURLOPT_URL, url);
204 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, data->no_pragma_header);
205
206 /* If we have successfully processed data from a previous fetch
207 attempt, only fetch the data we don't already have. */
208 if (prev_posn>0) {
209 if (walker->get_verbosely)
210 fprintf(stderr,
211 "Resuming fetch of object %s at byte %ld\n",
212 hex, prev_posn);
213 sprintf(range, "Range: bytes=%ld-", prev_posn);
214 range_header = curl_slist_append(range_header, range);
215 curl_easy_setopt(slot->curl,
216 CURLOPT_HTTPHEADER, range_header);
217 }
218
219 /* Try to get the request started, abort the request on error */
220 obj_req->state = ACTIVE;
221 if (!start_active_slot(slot)) {
222 obj_req->state = ABORTED;
223 obj_req->slot = NULL;
224 close(obj_req->local); obj_req->local = -1;
225 free(obj_req->url);
226 return;
227 }
228 }
229
230 static void finish_object_request(struct object_request *obj_req)
231 {
232 struct stat st;
233
234 fchmod(obj_req->local, 0444);
235 close(obj_req->local); obj_req->local = -1;
236
237 if (obj_req->http_code == 416) {
238 fprintf(stderr, "Warning: requested range invalid; we may already have all the data.\n");
239 } else if (obj_req->curl_result != CURLE_OK) {
240 if (stat(obj_req->tmpfile, &st) == 0)
241 if (st.st_size == 0)
242 unlink(obj_req->tmpfile);
243 return;
244 }
245
246 inflateEnd(&obj_req->stream);
247 git_SHA1_Final(obj_req->real_sha1, &obj_req->c);
248 if (obj_req->zret != Z_STREAM_END) {
249 unlink(obj_req->tmpfile);
250 return;
251 }
252 if (hashcmp(obj_req->sha1, obj_req->real_sha1)) {
253 unlink(obj_req->tmpfile);
254 return;
255 }
256 obj_req->rename =
257 move_temp_to_file(obj_req->tmpfile, obj_req->filename);
258
259 if (obj_req->rename == 0)
260 walker_say(obj_req->walker, "got %s\n", sha1_to_hex(obj_req->sha1));
261 }
262
263 static void process_object_response(void *callback_data)
264 {
265 struct object_request *obj_req =
266 (struct object_request *)callback_data;
267 struct walker *walker = obj_req->walker;
268 struct walker_data *data = walker->data;
269 struct alt_base *alt = data->alt;
270
271 obj_req->curl_result = obj_req->slot->curl_result;
272 obj_req->http_code = obj_req->slot->http_code;
273 obj_req->slot = NULL;
274 obj_req->state = COMPLETE;
275
276 /* Use alternates if necessary */
277 if (missing_target(obj_req)) {
278 fetch_alternates(walker, alt->base);
279 if (obj_req->repo->next != NULL) {
280 obj_req->repo =
281 obj_req->repo->next;
282 close(obj_req->local);
283 obj_req->local = -1;
284 start_object_request(walker, obj_req);
285 return;
286 }
287 }
288
289 finish_object_request(obj_req);
290 }
291
292 static void release_object_request(struct object_request *obj_req)
293 {
294 struct object_request *entry = object_queue_head;
295
296 if (obj_req->local != -1)
297 error("fd leakage in release: %d", obj_req->local);
298 if (obj_req == object_queue_head) {
299 object_queue_head = obj_req->next;
300 } else {
301 while (entry->next != NULL && entry->next != obj_req)
302 entry = entry->next;
303 if (entry->next == obj_req)
304 entry->next = entry->next->next;
305 }
306
307 free(obj_req->url);
308 free(obj_req);
309 }
310
311 #ifdef USE_CURL_MULTI
312 static int fill_active_slot(struct walker *walker)
313 {
314 struct object_request *obj_req;
315
316 for (obj_req = object_queue_head; obj_req; obj_req = obj_req->next) {
317 if (obj_req->state == WAITING) {
318 if (has_sha1_file(obj_req->sha1))
319 obj_req->state = COMPLETE;
320 else {
321 start_object_request(walker, obj_req);
322 return 1;
323 }
324 }
325 }
326 return 0;
327 }
328 #endif
329
330 static void prefetch(struct walker *walker, unsigned char *sha1)
331 {
332 struct object_request *newreq;
333 struct object_request *tail;
334 struct walker_data *data = walker->data;
335 char *filename = sha1_file_name(sha1);
336
337 newreq = xmalloc(sizeof(*newreq));
338 newreq->walker = walker;
339 hashcpy(newreq->sha1, sha1);
340 newreq->repo = data->alt;
341 newreq->url = NULL;
342 newreq->local = -1;
343 newreq->state = WAITING;
344 snprintf(newreq->filename, sizeof(newreq->filename), "%s", filename);
345 snprintf(newreq->tmpfile, sizeof(newreq->tmpfile),
346 "%s.temp", filename);
347 newreq->slot = NULL;
348 newreq->next = NULL;
349
350 if (object_queue_head == NULL) {
351 object_queue_head = newreq;
352 } else {
353 tail = object_queue_head;
354 while (tail->next != NULL) {
355 tail = tail->next;
356 }
357 tail->next = newreq;
358 }
359
360 #ifdef USE_CURL_MULTI
361 fill_active_slots();
362 step_active_slots();
363 #endif
364 }
365
366 static int fetch_index(struct walker *walker, struct alt_base *repo, unsigned char *sha1)
367 {
368 char *hex = sha1_to_hex(sha1);
369 char *filename;
370 char *url;
371 char tmpfile[PATH_MAX];
372 long prev_posn = 0;
373 char range[RANGE_HEADER_SIZE];
374 struct curl_slist *range_header = NULL;
375 struct walker_data *data = walker->data;
376
377 FILE *indexfile;
378 struct active_request_slot *slot;
379 struct slot_results results;
380
381 if (has_pack_index(sha1))
382 return 0;
383
384 if (walker->get_verbosely)
385 fprintf(stderr, "Getting index for pack %s\n", hex);
386
387 url = xmalloc(strlen(repo->base) + 64);
388 sprintf(url, "%s/objects/pack/pack-%s.idx", repo->base, hex);
389
390 filename = sha1_pack_index_name(sha1);
391 snprintf(tmpfile, sizeof(tmpfile), "%s.temp", filename);
392 indexfile = fopen(tmpfile, "a");
393 if (!indexfile)
394 return error("Unable to open local file %s for pack index",
395 tmpfile);
396
397 slot = get_active_slot();
398 slot->results = &results;
399 curl_easy_setopt(slot->curl, CURLOPT_FILE, indexfile);
400 curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite);
401 curl_easy_setopt(slot->curl, CURLOPT_URL, url);
402 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, data->no_pragma_header);
403 slot->local = indexfile;
404
405 /* If there is data present from a previous transfer attempt,
406 resume where it left off */
407 prev_posn = ftell(indexfile);
408 if (prev_posn>0) {
409 if (walker->get_verbosely)
410 fprintf(stderr,
411 "Resuming fetch of index for pack %s at byte %ld\n",
412 hex, prev_posn);
413 sprintf(range, "Range: bytes=%ld-", prev_posn);
414 range_header = curl_slist_append(range_header, range);
415 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, range_header);
416 }
417
418 if (start_active_slot(slot)) {
419 run_active_slot(slot);
420 if (results.curl_result != CURLE_OK) {
421 fclose(indexfile);
422 return error("Unable to get pack index %s\n%s", url,
423 curl_errorstr);
424 }
425 } else {
426 fclose(indexfile);
427 return error("Unable to start request");
428 }
429
430 fclose(indexfile);
431
432 return move_temp_to_file(tmpfile, filename);
433 }
434
435 static int setup_index(struct walker *walker, struct alt_base *repo, unsigned char *sha1)
436 {
437 struct packed_git *new_pack;
438 if (has_pack_file(sha1))
439 return 0; /* don't list this as something we can get */
440
441 if (fetch_index(walker, repo, sha1))
442 return -1;
443
444 new_pack = parse_pack_index(sha1);
445 if (!new_pack)
446 return -1; /* parse_pack_index() already issued error message */
447 new_pack->next = repo->packs;
448 repo->packs = new_pack;
449 return 0;
450 }
451
452 static void process_alternates_response(void *callback_data)
453 {
454 struct alternates_request *alt_req =
455 (struct alternates_request *)callback_data;
456 struct walker *walker = alt_req->walker;
457 struct walker_data *cdata = walker->data;
458 struct active_request_slot *slot = alt_req->slot;
459 struct alt_base *tail = cdata->alt;
460 const char *base = alt_req->base;
461 static const char null_byte = '\0';
462 char *data;
463 int i = 0;
464
465 if (alt_req->http_specific) {
466 if (slot->curl_result != CURLE_OK ||
467 !alt_req->buffer->len) {
468
469 /* Try reusing the slot to get non-http alternates */
470 alt_req->http_specific = 0;
471 sprintf(alt_req->url, "%s/objects/info/alternates",
472 base);
473 curl_easy_setopt(slot->curl, CURLOPT_URL,
474 alt_req->url);
475 active_requests++;
476 slot->in_use = 1;
477 if (slot->finished != NULL)
478 (*slot->finished) = 0;
479 if (!start_active_slot(slot)) {
480 cdata->got_alternates = -1;
481 slot->in_use = 0;
482 if (slot->finished != NULL)
483 (*slot->finished) = 1;
484 }
485 return;
486 }
487 } else if (slot->curl_result != CURLE_OK) {
488 if (!missing_target(slot)) {
489 cdata->got_alternates = -1;
490 return;
491 }
492 }
493
494 fwrite_buffer(&null_byte, 1, 1, alt_req->buffer);
495 alt_req->buffer->len--;
496 data = alt_req->buffer->buf;
497
498 while (i < alt_req->buffer->len) {
499 int posn = i;
500 while (posn < alt_req->buffer->len && data[posn] != '\n')
501 posn++;
502 if (data[posn] == '\n') {
503 int okay = 0;
504 int serverlen = 0;
505 struct alt_base *newalt;
506 char *target = NULL;
507 if (data[i] == '/') {
508 /* This counts
509 * http://git.host/pub/scm/linux.git/
510 * -----------here^
511 * so memcpy(dst, base, serverlen) will
512 * copy up to "...git.host".
513 */
514 const char *colon_ss = strstr(base,"://");
515 if (colon_ss) {
516 serverlen = (strchr(colon_ss + 3, '/')
517 - base);
518 okay = 1;
519 }
520 } else if (!memcmp(data + i, "../", 3)) {
521 /* Relative URL; chop the corresponding
522 * number of subpath from base (and ../
523 * from data), and concatenate the result.
524 *
525 * The code first drops ../ from data, and
526 * then drops one ../ from data and one path
527 * from base. IOW, one extra ../ is dropped
528 * from data than path is dropped from base.
529 *
530 * This is not wrong. The alternate in
531 * http://git.host/pub/scm/linux.git/
532 * to borrow from
533 * http://git.host/pub/scm/linus.git/
534 * is ../../linus.git/objects/. You need
535 * two ../../ to borrow from your direct
536 * neighbour.
537 */
538 i += 3;
539 serverlen = strlen(base);
540 while (i + 2 < posn &&
541 !memcmp(data + i, "../", 3)) {
542 do {
543 serverlen--;
544 } while (serverlen &&
545 base[serverlen - 1] != '/');
546 i += 3;
547 }
548 /* If the server got removed, give up. */
549 okay = strchr(base, ':') - base + 3 <
550 serverlen;
551 } else if (alt_req->http_specific) {
552 char *colon = strchr(data + i, ':');
553 char *slash = strchr(data + i, '/');
554 if (colon && slash && colon < data + posn &&
555 slash < data + posn && colon < slash) {
556 okay = 1;
557 }
558 }
559 /* skip "objects\n" at end */
560 if (okay) {
561 target = xmalloc(serverlen + posn - i - 6);
562 memcpy(target, base, serverlen);
563 memcpy(target + serverlen, data + i,
564 posn - i - 7);
565 target[serverlen + posn - i - 7] = 0;
566 if (walker->get_verbosely)
567 fprintf(stderr,
568 "Also look at %s\n", target);
569 newalt = xmalloc(sizeof(*newalt));
570 newalt->next = NULL;
571 newalt->base = target;
572 newalt->got_indices = 0;
573 newalt->packs = NULL;
574
575 while (tail->next != NULL)
576 tail = tail->next;
577 tail->next = newalt;
578 }
579 }
580 i = posn + 1;
581 }
582
583 cdata->got_alternates = 1;
584 }
585
586 static void fetch_alternates(struct walker *walker, const char *base)
587 {
588 struct strbuf buffer = STRBUF_INIT;
589 char *url;
590 struct active_request_slot *slot;
591 struct alternates_request alt_req;
592 struct walker_data *cdata = walker->data;
593
594 /* If another request has already started fetching alternates,
595 wait for them to arrive and return to processing this request's
596 curl message */
597 #ifdef USE_CURL_MULTI
598 while (cdata->got_alternates == 0) {
599 step_active_slots();
600 }
601 #endif
602
603 /* Nothing to do if they've already been fetched */
604 if (cdata->got_alternates == 1)
605 return;
606
607 /* Start the fetch */
608 cdata->got_alternates = 0;
609
610 if (walker->get_verbosely)
611 fprintf(stderr, "Getting alternates list for %s\n", base);
612
613 url = xmalloc(strlen(base) + 31);
614 sprintf(url, "%s/objects/info/http-alternates", base);
615
616 /* Use a callback to process the result, since another request
617 may fail and need to have alternates loaded before continuing */
618 slot = get_active_slot();
619 slot->callback_func = process_alternates_response;
620 alt_req.walker = walker;
621 slot->callback_data = &alt_req;
622
623 curl_easy_setopt(slot->curl, CURLOPT_FILE, &buffer);
624 curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite_buffer);
625 curl_easy_setopt(slot->curl, CURLOPT_URL, url);
626
627 alt_req.base = base;
628 alt_req.url = url;
629 alt_req.buffer = &buffer;
630 alt_req.http_specific = 1;
631 alt_req.slot = slot;
632
633 if (start_active_slot(slot))
634 run_active_slot(slot);
635 else
636 cdata->got_alternates = -1;
637
638 strbuf_release(&buffer);
639 free(url);
640 }
641
642 static int fetch_indices(struct walker *walker, struct alt_base *repo)
643 {
644 unsigned char sha1[20];
645 char *url;
646 struct strbuf buffer = STRBUF_INIT;
647 char *data;
648 int i = 0;
649 int ret = 0;
650
651 struct active_request_slot *slot;
652 struct slot_results results;
653
654 if (repo->got_indices)
655 return 0;
656
657 if (walker->get_verbosely)
658 fprintf(stderr, "Getting pack list for %s\n", repo->base);
659
660 url = xmalloc(strlen(repo->base) + 21);
661 sprintf(url, "%s/objects/info/packs", repo->base);
662
663 slot = get_active_slot();
664 slot->results = &results;
665 curl_easy_setopt(slot->curl, CURLOPT_FILE, &buffer);
666 curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite_buffer);
667 curl_easy_setopt(slot->curl, CURLOPT_URL, url);
668 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, NULL);
669 if (start_active_slot(slot)) {
670 run_active_slot(slot);
671 if (results.curl_result != CURLE_OK) {
672 if (missing_target(&results)) {
673 repo->got_indices = 1;
674 goto cleanup;
675 } else {
676 repo->got_indices = 0;
677 ret = error("%s", curl_errorstr);
678 goto cleanup;
679 }
680 }
681 } else {
682 repo->got_indices = 0;
683 ret = error("Unable to start request");
684 goto cleanup;
685 }
686
687 data = buffer.buf;
688 while (i < buffer.len) {
689 switch (data[i]) {
690 case 'P':
691 i++;
692 if (i + 52 <= buffer.len &&
693 !prefixcmp(data + i, " pack-") &&
694 !prefixcmp(data + i + 46, ".pack\n")) {
695 get_sha1_hex(data + i + 6, sha1);
696 setup_index(walker, repo, sha1);
697 i += 51;
698 break;
699 }
700 default:
701 while (i < buffer.len && data[i] != '\n')
702 i++;
703 }
704 i++;
705 }
706
707 repo->got_indices = 1;
708 cleanup:
709 strbuf_release(&buffer);
710 free(url);
711 return ret;
712 }
713
714 static int fetch_pack(struct walker *walker, struct alt_base *repo, unsigned char *sha1)
715 {
716 char *url;
717 struct packed_git *target;
718 struct packed_git **lst;
719 FILE *packfile;
720 char *filename;
721 char tmpfile[PATH_MAX];
722 int ret;
723 long prev_posn = 0;
724 char range[RANGE_HEADER_SIZE];
725 struct curl_slist *range_header = NULL;
726 struct walker_data *data = walker->data;
727
728 struct active_request_slot *slot;
729 struct slot_results results;
730
731 if (fetch_indices(walker, repo))
732 return -1;
733 target = find_sha1_pack(sha1, repo->packs);
734 if (!target)
735 return -1;
736
737 if (walker->get_verbosely) {
738 fprintf(stderr, "Getting pack %s\n",
739 sha1_to_hex(target->sha1));
740 fprintf(stderr, " which contains %s\n",
741 sha1_to_hex(sha1));
742 }
743
744 url = xmalloc(strlen(repo->base) + 65);
745 sprintf(url, "%s/objects/pack/pack-%s.pack",
746 repo->base, sha1_to_hex(target->sha1));
747
748 filename = sha1_pack_name(target->sha1);
749 snprintf(tmpfile, sizeof(tmpfile), "%s.temp", filename);
750 packfile = fopen(tmpfile, "a");
751 if (!packfile)
752 return error("Unable to open local file %s for pack",
753 tmpfile);
754
755 slot = get_active_slot();
756 slot->results = &results;
757 curl_easy_setopt(slot->curl, CURLOPT_FILE, packfile);
758 curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite);
759 curl_easy_setopt(slot->curl, CURLOPT_URL, url);
760 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, data->no_pragma_header);
761 slot->local = packfile;
762
763 /* If there is data present from a previous transfer attempt,
764 resume where it left off */
765 prev_posn = ftell(packfile);
766 if (prev_posn>0) {
767 if (walker->get_verbosely)
768 fprintf(stderr,
769 "Resuming fetch of pack %s at byte %ld\n",
770 sha1_to_hex(target->sha1), prev_posn);
771 sprintf(range, "Range: bytes=%ld-", prev_posn);
772 range_header = curl_slist_append(range_header, range);
773 curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, range_header);
774 }
775
776 if (start_active_slot(slot)) {
777 run_active_slot(slot);
778 if (results.curl_result != CURLE_OK) {
779 fclose(packfile);
780 return error("Unable to get pack file %s\n%s", url,
781 curl_errorstr);
782 }
783 } else {
784 fclose(packfile);
785 return error("Unable to start request");
786 }
787
788 target->pack_size = ftell(packfile);
789 fclose(packfile);
790
791 ret = move_temp_to_file(tmpfile, filename);
792 if (ret)
793 return ret;
794
795 lst = &repo->packs;
796 while (*lst != target)
797 lst = &((*lst)->next);
798 *lst = (*lst)->next;
799
800 if (verify_pack(target))
801 return -1;
802 install_packed_git(target);
803
804 return 0;
805 }
806
807 static void abort_object_request(struct object_request *obj_req)
808 {
809 if (obj_req->local >= 0) {
810 close(obj_req->local);
811 obj_req->local = -1;
812 }
813 unlink(obj_req->tmpfile);
814 if (obj_req->slot) {
815 release_active_slot(obj_req->slot);
816 obj_req->slot = NULL;
817 }
818 release_object_request(obj_req);
819 }
820
821 static int fetch_object(struct walker *walker, struct alt_base *repo, unsigned char *sha1)
822 {
823 char *hex = sha1_to_hex(sha1);
824 int ret = 0;
825 struct object_request *obj_req = object_queue_head;
826
827 while (obj_req != NULL && hashcmp(obj_req->sha1, sha1))
828 obj_req = obj_req->next;
829 if (obj_req == NULL)
830 return error("Couldn't find request for %s in the queue", hex);
831
832 if (has_sha1_file(obj_req->sha1)) {
833 abort_object_request(obj_req);
834 return 0;
835 }
836
837 #ifdef USE_CURL_MULTI
838 while (obj_req->state == WAITING) {
839 step_active_slots();
840 }
841 #else
842 start_object_request(walker, obj_req);
843 #endif
844
845 while (obj_req->state == ACTIVE) {
846 run_active_slot(obj_req->slot);
847 }
848 if (obj_req->local != -1) {
849 close(obj_req->local); obj_req->local = -1;
850 }
851
852 if (obj_req->state == ABORTED) {
853 ret = error("Request for %s aborted", hex);
854 } else if (obj_req->curl_result != CURLE_OK &&
855 obj_req->http_code != 416) {
856 if (missing_target(obj_req))
857 ret = -1; /* Be silent, it is probably in a pack. */
858 else
859 ret = error("%s (curl_result = %d, http_code = %ld, sha1 = %s)",
860 obj_req->errorstr, obj_req->curl_result,
861 obj_req->http_code, hex);
862 } else if (obj_req->zret != Z_STREAM_END) {
863 walker->corrupt_object_found++;
864 ret = error("File %s (%s) corrupt", hex, obj_req->url);
865 } else if (hashcmp(obj_req->sha1, obj_req->real_sha1)) {
866 ret = error("File %s has bad hash", hex);
867 } else if (obj_req->rename < 0) {
868 ret = error("unable to write sha1 filename %s",
869 obj_req->filename);
870 }
871
872 release_object_request(obj_req);
873 return ret;
874 }
875
876 static int fetch(struct walker *walker, unsigned char *sha1)
877 {
878 struct walker_data *data = walker->data;
879 struct alt_base *altbase = data->alt;
880
881 if (!fetch_object(walker, altbase, sha1))
882 return 0;
883 while (altbase) {
884 if (!fetch_pack(walker, altbase, sha1))
885 return 0;
886 fetch_alternates(walker, data->alt->base);
887 altbase = altbase->next;
888 }
889 return error("Unable to find %s under %s", sha1_to_hex(sha1),
890 data->alt->base);
891 }
892
893 static int fetch_ref(struct walker *walker, struct ref *ref)
894 {
895 struct walker_data *data = walker->data;
896 return http_fetch_ref(data->alt->base, ref);
897 }
898
899 static void cleanup(struct walker *walker)
900 {
901 struct walker_data *data = walker->data;
902 http_cleanup();
903
904 curl_slist_free_all(data->no_pragma_header);
905 }
906
907 struct walker *get_http_walker(const char *url, struct remote *remote)
908 {
909 char *s;
910 struct walker_data *data = xmalloc(sizeof(struct walker_data));
911 struct walker *walker = xmalloc(sizeof(struct walker));
912
913 http_init(remote);
914
915 data->no_pragma_header = curl_slist_append(NULL, "Pragma:");
916
917 data->alt = xmalloc(sizeof(*data->alt));
918 data->alt->base = xmalloc(strlen(url) + 1);
919 strcpy(data->alt->base, url);
920 for (s = data->alt->base + strlen(data->alt->base) - 1; *s == '/'; --s)
921 *s = 0;
922
923 data->alt->got_indices = 0;
924 data->alt->packs = NULL;
925 data->alt->next = NULL;
926 data->got_alternates = -1;
927
928 walker->corrupt_object_found = 0;
929 walker->fetch = fetch;
930 walker->fetch_ref = fetch_ref;
931 walker->prefetch = prefetch;
932 walker->cleanup = cleanup;
933 walker->data = data;
934
935 #ifdef USE_CURL_MULTI
936 add_fill_function(walker, (int (*)(void *)) fill_active_slot);
937 #endif
938
939 return walker;
940 }