patch_parse.c 32.1 KB
Newer Older
1 2 3 4 5 6
/*
 * Copyright (C) the libgit2 contributors. All rights reserved.
 *
 * This file is part of libgit2, distributed under the GNU GPL v2 with
 * a Linking Exception. For full terms see the included COPYING file.
 */
7 8 9

#include "patch_parse.h"

10 11
#include "git2/patch.h"
#include "patch.h"
12
#include "diff_parse.h"
13 14 15
#include "path.h"

typedef struct {
16 17 18
	git_patch base;

	git_patch_parse_ctx *ctx;
19

20 21 22 23 24 25 26 27 28 29 30 31 32 33
	/* the paths from the `diff --git` header, these will be used if this is not
	 * a rename (and rename paths are specified) or if no `+++`/`---` line specify
	 * the paths.
	 */
	char *header_old_path, *header_new_path;

	/* renamed paths are precise and are not prefixed */
	char *rename_old_path, *rename_new_path;

	/* the paths given in `---` and `+++` lines */
	char *old_path, *new_path;

	/* the prefixes from the old/new paths */
	char *old_prefix, *new_prefix;
34 35
} git_patch_parsed;

36 37 38 39 40 41 42 43 44 45 46 47
static int git_parse_err(const char *fmt, ...) GIT_FORMAT_PRINTF(1, 2);
static int git_parse_err(const char *fmt, ...)
{
	va_list ap;

	va_start(ap, fmt);
	git_error_vset(GIT_ERROR_PATCH, fmt, ap);
	va_end(ap);

	return -1;
}

48
static size_t header_path_len(git_patch_parse_ctx *ctx)
49 50
{
	bool inquote = 0;
51
	bool quoted = git_parse_ctx_contains_s(&ctx->parse_ctx, "\"");
52 53
	size_t len;

54 55
	for (len = quoted; len < ctx->parse_ctx.line_len; len++) {
		if (!quoted && git__isspace(ctx->parse_ctx.line[len]))
56
			break;
57
		else if (quoted && !inquote && ctx->parse_ctx.line[len] == '"') {
58 59 60 61
			len++;
			break;
		}

62
		inquote = (!inquote && ctx->parse_ctx.line[len] == '\\');
63 64 65 66 67
	}

	return len;
}

68
static int parse_header_path_buf(git_buf *path, git_patch_parse_ctx *ctx, size_t path_len)
69
{
70
	int error;
71

72
	if ((error = git_buf_put(path, ctx->parse_ctx.line, path_len)) < 0)
73
		return error;
74

75
	git_parse_advance_chars(&ctx->parse_ctx, path_len);
76 77 78

	git_buf_rtrim(path);

79 80 81
	if (path->size > 0 && path->ptr[0] == '"' &&
	    (error = git_buf_unquote(path)) < 0)
		return error;
82 83 84

	git_path_squash_slashes(path);

85 86 87 88 89
	if (!path->size)
		return git_parse_err("patch contains empty path at line %"PRIuZ,
				     ctx->parse_ctx.line_num);

	return 0;
90 91
}

92
static int parse_header_path(char **out, git_patch_parse_ctx *ctx)
93 94
{
	git_buf path = GIT_BUF_INIT;
95
	int error;
96

97 98
	if ((error = parse_header_path_buf(&path, ctx, header_path_len(ctx))) < 0)
		goto out;
99 100
	*out = git_buf_detach(&path);

101 102
out:
	git_buf_dispose(&path);
103 104 105 106
	return error;
}

static int parse_header_git_oldpath(
107
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
108
{
109 110 111
	git_buf old_path = GIT_BUF_INIT;
	int error;

112 113 114 115 116 117
	if (patch->old_path) {
		error = git_parse_err("patch contains duplicate old path at line %"PRIuZ,
				      ctx->parse_ctx.line_num);
		goto out;
	}

118 119 120 121 122 123
	if ((error = parse_header_path_buf(&old_path, ctx, ctx->parse_ctx.line_len - 1)) <  0)
		goto out;

	patch->old_path = git_buf_detach(&old_path);

out:
124
	git_buf_dispose(&old_path);
125
	return error;
126 127 128
}

static int parse_header_git_newpath(
129
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
130
{
131 132 133
	git_buf new_path = GIT_BUF_INIT;
	int error;

134 135 136
	if (patch->new_path) {
		error = git_parse_err("patch contains duplicate new path at line %"PRIuZ,
				      ctx->parse_ctx.line_num);
137
		goto out;
138
	}
139

140 141
	if ((error = parse_header_path_buf(&new_path, ctx, ctx->parse_ctx.line_len - 1)) <  0)
		goto out;
142 143 144
	patch->new_path = git_buf_detach(&new_path);

out:
145
	git_buf_dispose(&new_path);
146
	return error;
147 148
}

149
static int parse_header_mode(uint16_t *mode, git_patch_parse_ctx *ctx)
150
{
151
	int64_t m;
152

153
	if ((git_parse_advance_digit(&m, &ctx->parse_ctx, 8)) < 0)
154
		return git_parse_err("invalid file mode at line %"PRIuZ, ctx->parse_ctx.line_num);
155 156 157 158 159 160

	if (m > UINT16_MAX)
		return -1;

	*mode = (uint16_t)m;

161
	return 0;
162 163 164 165
}

static int parse_header_oid(
	git_oid *oid,
166
	uint16_t *oid_len,
167
	git_patch_parse_ctx *ctx)
168 169 170
{
	size_t len;

171 172
	for (len = 0; len < ctx->parse_ctx.line_len && len < GIT_OID_HEXSZ; len++) {
		if (!git__isxdigit(ctx->parse_ctx.line[len]))
173 174 175
			break;
	}

176
	if (len < GIT_OID_MINPREFIXLEN || len > GIT_OID_HEXSZ ||
177 178 179
		git_oid_fromstrn(oid, ctx->parse_ctx.line, len) < 0)
		return git_parse_err("invalid hex formatted object id at line %"PRIuZ,
			ctx->parse_ctx.line_num);
180

181
	git_parse_advance_chars(&ctx->parse_ctx, len);
182

183
	*oid_len = (uint16_t)len;
184 185 186 187 188

	return 0;
}

static int parse_header_git_index(
189
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
190
{
191 192
	char c;

193 194
	if (parse_header_oid(&patch->base.delta->old_file.id,
			&patch->base.delta->old_file.id_abbrev, ctx) < 0 ||
195
		git_parse_advance_expected_str(&ctx->parse_ctx, "..") < 0 ||
196 197
		parse_header_oid(&patch->base.delta->new_file.id,
			&patch->base.delta->new_file.id_abbrev, ctx) < 0)
198 199
		return -1;

200
	if (git_parse_peek(&c, &ctx->parse_ctx, 0) == 0 && c == ' ') {
201
		uint16_t mode = 0;
202

203
		git_parse_advance_chars(&ctx->parse_ctx, 1);
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218

		if (parse_header_mode(&mode, ctx) < 0)
			return -1;

		if (!patch->base.delta->new_file.mode)
			patch->base.delta->new_file.mode = mode;

		if (!patch->base.delta->old_file.mode)
			patch->base.delta->old_file.mode = mode;
	}

	return 0;
}

static int parse_header_git_oldmode(
219
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
220
{
221
	return parse_header_mode(&patch->base.delta->old_file.mode, ctx);
222 223 224
}

static int parse_header_git_newmode(
225
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
226
{
227
	return parse_header_mode(&patch->base.delta->new_file.mode, ctx);
228 229 230 231
}

static int parse_header_git_deletedfilemode(
	git_patch_parsed *patch,
232
	git_patch_parse_ctx *ctx)
233
{
234
	git__free((char *)patch->base.delta->new_file.path);
235

236
	patch->base.delta->new_file.path = NULL;
237
	patch->base.delta->status = GIT_DELTA_DELETED;
238
	patch->base.delta->nfiles = 1;
239

240
	return parse_header_mode(&patch->base.delta->old_file.mode, ctx);
241 242 243 244
}

static int parse_header_git_newfilemode(
	git_patch_parsed *patch,
245
	git_patch_parse_ctx *ctx)
246
{
247
	git__free((char *)patch->base.delta->old_file.path);
248

249
	patch->base.delta->old_file.path = NULL;
250
	patch->base.delta->status = GIT_DELTA_ADDED;
251
	patch->base.delta->nfiles = 1;
252

253
	return parse_header_mode(&patch->base.delta->new_file.mode, ctx);
254 255 256 257
}

static int parse_header_rename(
	char **out,
258
	git_patch_parse_ctx *ctx)
259 260 261
{
	git_buf path = GIT_BUF_INIT;

262
	if (parse_header_path_buf(&path, ctx, header_path_len(ctx)) < 0)
263 264
		return -1;

265 266 267 268
	/* Note: the `rename from` and `rename to` lines include the literal
	 * filename.  They do *not* include the prefix.  (Who needs consistency?)
	 */
	*out = git_buf_detach(&path);
269 270 271 272
	return 0;
}

static int parse_header_renamefrom(
273
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
274
{
275
	patch->base.delta->status = GIT_DELTA_RENAMED;
276
	return parse_header_rename(&patch->rename_old_path, ctx);
277 278 279
}

static int parse_header_renameto(
280
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
281
{
282
	patch->base.delta->status = GIT_DELTA_RENAMED;
283
	return parse_header_rename(&patch->rename_new_path, ctx);
284 285
}

286 287 288 289 290 291 292 293 294 295 296 297 298 299
static int parse_header_copyfrom(
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
{
	patch->base.delta->status = GIT_DELTA_COPIED;
	return parse_header_rename(&patch->rename_old_path, ctx);
}

static int parse_header_copyto(
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
{
	patch->base.delta->status = GIT_DELTA_COPIED;
	return parse_header_rename(&patch->rename_new_path, ctx);
}

300
static int parse_header_percent(uint16_t *out, git_patch_parse_ctx *ctx)
301
{
302
	int64_t val;
303

304
	if (git_parse_advance_digit(&val, &ctx->parse_ctx, 10) < 0)
305 306
		return -1;

307
	if (git_parse_advance_expected_str(&ctx->parse_ctx, "%") < 0)
308 309
		return -1;

310
	if (val < 0 || val > 100)
311 312
		return -1;

313
	*out = (uint16_t)val;
314 315 316 317
	return 0;
}

static int parse_header_similarity(
318
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
319 320
{
	if (parse_header_percent(&patch->base.delta->similarity, ctx) < 0)
321 322
		return git_parse_err("invalid similarity percentage at line %"PRIuZ,
			ctx->parse_ctx.line_num);
323 324 325 326 327

	return 0;
}

static int parse_header_dissimilarity(
328
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
329 330 331 332
{
	uint16_t dissimilarity;

	if (parse_header_percent(&dissimilarity, ctx) < 0)
333 334
		return git_parse_err("invalid similarity percentage at line %"PRIuZ,
			ctx->parse_ctx.line_num);
335 336 337 338 339 340

	patch->base.delta->similarity = 100 - dissimilarity;

	return 0;
}

341 342 343
static int parse_header_start(git_patch_parsed *patch, git_patch_parse_ctx *ctx)
{
	if (parse_header_path(&patch->header_old_path, ctx) < 0)
344 345
		return git_parse_err("corrupt old path in git diff header at line %"PRIuZ,
			ctx->parse_ctx.line_num);
346

347
	if (git_parse_advance_ws(&ctx->parse_ctx) < 0 ||
348
		parse_header_path(&patch->header_new_path, ctx) < 0)
349 350
		return git_parse_err("corrupt new path in git diff header at line %"PRIuZ,
			ctx->parse_ctx.line_num);
351

352 353 354 355 356 357 358
	/*
	 * We cannot expect to be able to always parse paths correctly at this
	 * point. Due to the possibility of unquoted names, whitespaces in
	 * filenames and custom prefixes we have to allow that, though, and just
	 * proceeed here. We then hope for the "---" and "+++" lines to fix that
	 * for us.
	 */
359 360
	if (!git_parse_ctx_contains(&ctx->parse_ctx, "\n", 1) &&
	    !git_parse_ctx_contains(&ctx->parse_ctx, "\r\n", 2)) {
361 362 363 364 365 366 367 368
		git_parse_advance_chars(&ctx->parse_ctx, ctx->parse_ctx.line_len - 1);

		git__free(patch->header_old_path);
		patch->header_old_path = NULL;
		git__free(patch->header_new_path);
		patch->header_new_path = NULL;
	}

369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
	return 0;
}

typedef enum {
	STATE_START,

	STATE_DIFF,
	STATE_FILEMODE,
	STATE_MODE,
	STATE_INDEX,
	STATE_PATH,

	STATE_SIMILARITY,
	STATE_RENAME,
	STATE_COPY,

	STATE_END,
} parse_header_state;

388 389
typedef struct {
	const char *str;
390 391
	parse_header_state expected_state;
	parse_header_state next_state;
392
	int(*fn)(git_patch_parsed *, git_patch_parse_ctx *);
393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
} parse_header_transition;

static const parse_header_transition transitions[] = {
	/* Start */
	{ "diff --git "         , STATE_START,      STATE_DIFF,       parse_header_start },

	{ "deleted file mode "  , STATE_DIFF,       STATE_FILEMODE,   parse_header_git_deletedfilemode },
	{ "new file mode "      , STATE_DIFF,       STATE_FILEMODE,   parse_header_git_newfilemode },
	{ "old mode "           , STATE_DIFF,       STATE_MODE,       parse_header_git_oldmode },
	{ "new mode "           , STATE_MODE,       STATE_END,        parse_header_git_newmode },

	{ "index "              , STATE_FILEMODE,   STATE_INDEX,      parse_header_git_index },
	{ "index "              , STATE_DIFF,       STATE_INDEX,      parse_header_git_index },
	{ "index "              , STATE_END,        STATE_INDEX,      parse_header_git_index },

408
	{ "--- "                , STATE_DIFF,       STATE_PATH,       parse_header_git_oldpath },
409
	{ "--- "                , STATE_INDEX,      STATE_PATH,       parse_header_git_oldpath },
410
	{ "--- "                , STATE_FILEMODE,   STATE_PATH,       parse_header_git_oldpath },
411 412 413 414
	{ "+++ "                , STATE_PATH,       STATE_END,        parse_header_git_newpath },
	{ "GIT binary patch"    , STATE_INDEX,      STATE_END,        NULL },
	{ "Binary files "       , STATE_INDEX,      STATE_END,        NULL },

415
	{ "similarity index "   , STATE_END,        STATE_SIMILARITY, parse_header_similarity },
416 417 418 419 420 421 422 423 424 425 426 427
	{ "similarity index "   , STATE_DIFF,       STATE_SIMILARITY, parse_header_similarity },
	{ "dissimilarity index ", STATE_DIFF,       STATE_SIMILARITY, parse_header_dissimilarity },
	{ "rename from "        , STATE_SIMILARITY, STATE_RENAME,     parse_header_renamefrom },
	{ "rename old "         , STATE_SIMILARITY, STATE_RENAME,     parse_header_renamefrom },
	{ "copy from "          , STATE_SIMILARITY, STATE_COPY,       parse_header_copyfrom },
	{ "rename to "          , STATE_RENAME,     STATE_END,        parse_header_renameto },
	{ "rename new "         , STATE_RENAME,     STATE_END,        parse_header_renameto },
	{ "copy to "            , STATE_COPY,       STATE_END,        parse_header_copyto },

	/* Next patch */
	{ "diff --git "         , STATE_END,        0,                NULL },
	{ "@@ -"                , STATE_END,        0,                NULL },
428
	{ "-- "                 , STATE_INDEX,      0,                NULL },
429
	{ "-- "                 , STATE_END,        0,                NULL },
430 431 432 433
};

static int parse_header_git(
	git_patch_parsed *patch,
434
	git_patch_parse_ctx *ctx)
435 436 437
{
	size_t i;
	int error = 0;
438
	parse_header_state state = STATE_START;
439 440

	/* Parse remaining header lines */
441
	for (; ctx->parse_ctx.remain_len > 0; git_parse_advance_line(&ctx->parse_ctx)) {
442 443
		bool found = false;

444
		if (ctx->parse_ctx.line_len == 0 || ctx->parse_ctx.line[ctx->parse_ctx.line_len - 1] != '\n')
445 446
			break;

447 448 449
		for (i = 0; i < ARRAY_SIZE(transitions); i++) {
			const parse_header_transition *transition = &transitions[i];
			size_t op_len = strlen(transition->str);
450

451
			if (transition->expected_state != state ||
452
			    git__prefixcmp(ctx->parse_ctx.line, transition->str) != 0)
453 454
				continue;

455 456
			state = transition->next_state;

457
			/* Do not advance if this is the patch separator */
458
			if (transition->fn == NULL)
459 460
				goto done;

461
			git_parse_advance_chars(&ctx->parse_ctx, op_len);
462

463
			if ((error = transition->fn(patch, ctx)) < 0)
464 465
				goto done;

466
			git_parse_advance_ws(&ctx->parse_ctx);
467

468 469 470
			if (git_parse_advance_expected_str(&ctx->parse_ctx, "\n") < 0 ||
			    ctx->parse_ctx.line_len > 0) {
				error = git_parse_err("trailing data at line %"PRIuZ, ctx->parse_ctx.line_num);
471 472 473
				goto done;
			}

474
			found = true;
475 476
			break;
		}
477

478
		if (!found) {
479 480
			error = git_parse_err("invalid patch header at line %"PRIuZ,
				ctx->parse_ctx.line_num);
481 482
			goto done;
		}
483 484
	}

485
	if (state != STATE_END) {
486
		error = git_parse_err("unexpected header line %"PRIuZ, ctx->parse_ctx.line_num);
487 488 489
		goto done;
	}

490 491 492 493
done:
	return error;
}

494
static int parse_int(int *out, git_patch_parse_ctx *ctx)
495
{
496
	int64_t num;
497

498
	if (git_parse_advance_digit(&num, &ctx->parse_ctx, 10) < 0 || !git__is_int(num))
499 500 501 502 503 504 505 506
		return -1;

	*out = (int)num;
	return 0;
}

static int parse_hunk_header(
	git_patch_hunk *hunk,
507
	git_patch_parse_ctx *ctx)
508
{
509
	const char *header_start = ctx->parse_ctx.line;
510
	char c;
511 512 513 514

	hunk->hunk.old_lines = 1;
	hunk->hunk.new_lines = 1;

515
	if (git_parse_advance_expected_str(&ctx->parse_ctx, "@@ -") < 0 ||
516 517 518
		parse_int(&hunk->hunk.old_start, ctx) < 0)
		goto fail;

519
	if (git_parse_peek(&c, &ctx->parse_ctx, 0) == 0 && c == ',') {
520
		if (git_parse_advance_expected_str(&ctx->parse_ctx, ",") < 0 ||
521 522 523 524
			parse_int(&hunk->hunk.old_lines, ctx) < 0)
			goto fail;
	}

525
	if (git_parse_advance_expected_str(&ctx->parse_ctx, " +") < 0 ||
526 527 528
		parse_int(&hunk->hunk.new_start, ctx) < 0)
		goto fail;

529
	if (git_parse_peek(&c, &ctx->parse_ctx, 0) == 0 && c == ',') {
530
		if (git_parse_advance_expected_str(&ctx->parse_ctx, ",") < 0 ||
531 532 533 534
			parse_int(&hunk->hunk.new_lines, ctx) < 0)
			goto fail;
	}

535
	if (git_parse_advance_expected_str(&ctx->parse_ctx, " @@") < 0)
536 537
		goto fail;

538
	git_parse_advance_line(&ctx->parse_ctx);
539 540 541 542

	if (!hunk->hunk.old_lines && !hunk->hunk.new_lines)
		goto fail;

543
	hunk->hunk.header_len = ctx->parse_ctx.line - header_start;
544
	if (hunk->hunk.header_len > (GIT_DIFF_HUNK_HEADER_SIZE - 1))
545 546
		return git_parse_err("oversized patch hunk header at line %"PRIuZ,
			ctx->parse_ctx.line_num);
547 548 549 550 551 552 553

	memcpy(hunk->hunk.header, header_start, hunk->hunk.header_len);
	hunk->hunk.header[hunk->hunk.header_len] = '\0';

	return 0;

fail:
554
	git_error_set(GIT_ERROR_PATCH, "invalid patch hunk header at line %"PRIuZ,
555
		ctx->parse_ctx.line_num);
556 557 558
	return -1;
}

559 560 561 562 563 564 565 566
static int eof_for_origin(int origin) {
	if (origin == GIT_DIFF_LINE_ADDITION)
		return GIT_DIFF_LINE_ADD_EOFNL;
	if (origin == GIT_DIFF_LINE_DELETION)
		return GIT_DIFF_LINE_DEL_EOFNL;
	return GIT_DIFF_LINE_CONTEXT_EOFNL;
}

567 568 569
static int parse_hunk_body(
	git_patch_parsed *patch,
	git_patch_hunk *hunk,
570
	git_patch_parse_ctx *ctx)
571 572 573 574 575 576
{
	git_diff_line *line;
	int error = 0;

	int oldlines = hunk->hunk.old_lines;
	int newlines = hunk->hunk.new_lines;
577
	int last_origin = 0;
578 579

	for (;
580
		ctx->parse_ctx.remain_len > 1 &&
581
		(oldlines || newlines) &&
582
		!git_parse_ctx_contains_s(&ctx->parse_ctx, "@@ -");
583
		git_parse_advance_line(&ctx->parse_ctx)) {
584

585
		int old_lineno, new_lineno, origin, prefix = 1;
586
		char c;
587 588 589 590 591 592 593 594 595

		if (git__add_int_overflow(&old_lineno, hunk->hunk.old_start, hunk->hunk.old_lines) ||
		    git__sub_int_overflow(&old_lineno, old_lineno, oldlines) ||
		    git__add_int_overflow(&new_lineno, hunk->hunk.new_start, hunk->hunk.new_lines) ||
		    git__sub_int_overflow(&new_lineno, new_lineno, newlines)) {
			error = git_parse_err("unrepresentable line count at line %"PRIuZ,
					      ctx->parse_ctx.line_num);
			goto done;
		}
596

597 598 599
		if (ctx->parse_ctx.line_len == 0 || ctx->parse_ctx.line[ctx->parse_ctx.line_len - 1] != '\n') {
			error = git_parse_err("invalid patch instruction at line %"PRIuZ,
				ctx->parse_ctx.line_num);
600 601 602
			goto done;
		}

603 604 605
		git_parse_peek(&c, &ctx->parse_ctx, 0);

		switch (c) {
606 607
		case '\n':
			prefix = 0;
608
			/* fall through */
609 610 611 612 613 614 615 616 617 618

		case ' ':
			origin = GIT_DIFF_LINE_CONTEXT;
			oldlines--;
			newlines--;
			break;

		case '-':
			origin = GIT_DIFF_LINE_DELETION;
			oldlines--;
619
			new_lineno = -1;
620 621 622 623 624
			break;

		case '+':
			origin = GIT_DIFF_LINE_ADDITION;
			newlines--;
625
			old_lineno = -1;
626 627
			break;

628 629 630 631 632 633
		case '\\':
			/*
			 * If there are no oldlines left, then this is probably
			 * the "\ No newline at end of file" marker. Do not
			 * verify its format, as it may be localized.
			 */
634 635 636 637 638 639 640
			if (!oldlines) {
				prefix = 0;
				origin = eof_for_origin(last_origin);
				old_lineno = -1;
				new_lineno = -1;
				break;
			}
641 642
			/* fall through */

643
		default:
644
			error = git_parse_err("invalid patch hunk at line %"PRIuZ, ctx->parse_ctx.line_num);
645 646 647 648
			goto done;
		}

		line = git_array_alloc(patch->base.lines);
649
		GIT_ERROR_CHECK_ALLOC(line);
650 651 652

		memset(line, 0x0, sizeof(git_diff_line));

653
		line->content_len = ctx->parse_ctx.line_len - prefix;
654
		line->content = git__strndup(ctx->parse_ctx.line + prefix, line->content_len);
655
		GIT_ERROR_CHECK_ALLOC(line->content);
656
		line->content_offset = ctx->parse_ctx.content_len - ctx->parse_ctx.remain_len;
657
		line->origin = origin;
658 659 660
		line->num_lines = 1;
		line->old_lineno = old_lineno;
		line->new_lineno = new_lineno;
661 662

		hunk->line_count++;
663 664

		last_origin = origin;
665 666 667
	}

	if (oldlines || newlines) {
668
		error = git_parse_err(
669 670 671 672 673
			"invalid patch hunk, expected %d old lines and %d new lines",
			hunk->hunk.old_lines, hunk->hunk.new_lines);
		goto done;
	}

674
	/*
675
	 * Handle "\ No newline at end of file". Only expect the leading
676 677 678 679
	 * backslash, though, because the rest of the string could be
	 * localized.  Because `diff` optimizes for the case where you
	 * want to apply the patch by hand.
	 */
680
	if (git_parse_ctx_contains_s(&ctx->parse_ctx, "\\ ") &&
681 682 683 684 685
		git_array_size(patch->base.lines) > 0) {

		line = git_array_get(patch->base.lines, git_array_size(patch->base.lines) - 1);

		if (line->content_len < 1) {
686
			error = git_parse_err("last line has no trailing newline");
687 688 689
			goto done;
		}

690 691 692 693 694 695
		line = git_array_alloc(patch->base.lines);
		GIT_ERROR_CHECK_ALLOC(line);

		memset(line, 0x0, sizeof(git_diff_line));

		line->content_len = ctx->parse_ctx.line_len;
696 697
		line->content = git__strndup(ctx->parse_ctx.line, line->content_len);
		GIT_ERROR_CHECK_ALLOC(line->content);
698 699 700 701 702 703 704
		line->content_offset = ctx->parse_ctx.content_len - ctx->parse_ctx.remain_len;
		line->origin = eof_for_origin(last_origin);
		line->num_lines = 1;
		line->old_lineno = -1;
		line->new_lineno = -1;

		hunk->line_count++;
705

706
		git_parse_advance_line(&ctx->parse_ctx);
707 708 709 710 711 712
	}

done:
	return error;
}

713
static int parse_patch_header(
714
	git_patch_parsed *patch,
715
	git_patch_parse_ctx *ctx)
716 717 718
{
	int error = 0;

719
	for (; ctx->parse_ctx.remain_len > 0; git_parse_advance_line(&ctx->parse_ctx)) {
720
		/* This line is too short to be a patch header. */
721
		if (ctx->parse_ctx.line_len < 6)
722 723 724
			continue;

		/* This might be a hunk header without a patch header, provide a
725
		 * sensible error message. */
726 727
		if (git_parse_ctx_contains_s(&ctx->parse_ctx, "@@ -")) {
			size_t line_num = ctx->parse_ctx.line_num;
728 729 730 731 732 733
			git_patch_hunk hunk;

			/* If this cannot be parsed as a hunk header, it's just leading
			* noise, continue.
			*/
			if (parse_hunk_header(&hunk, ctx) < 0) {
734
				git_error_clear();
735 736 737
				continue;
			}

738
			error = git_parse_err("invalid hunk header outside patch at line %"PRIuZ,
739 740 741 742 743
				line_num);
			goto done;
		}

		/* This buffer is too short to contain a patch. */
744
		if (ctx->parse_ctx.remain_len < ctx->parse_ctx.line_len + 6)
745 746 747
			break;

		/* A proper git patch */
748
		if (git_parse_ctx_contains_s(&ctx->parse_ctx, "diff --git ")) {
749
			error = parse_header_git(patch, ctx);
750 751 752 753 754 755 756
			goto done;
		}

		error = 0;
		continue;
	}

757
	git_error_set(GIT_ERROR_PATCH, "no patch found");
758
	error = GIT_ENOTFOUND;
759 760 761 762 763

done:
	return error;
}

764
static int parse_patch_binary_side(
765
	git_diff_binary_file *binary,
766
	git_patch_parse_ctx *ctx)
767 768 769
{
	git_diff_binary_t type = GIT_DIFF_BINARY_NONE;
	git_buf base85 = GIT_BUF_INIT, decoded = GIT_BUF_INIT;
770
	int64_t len;
771 772
	int error = 0;

773
	if (git_parse_ctx_contains_s(&ctx->parse_ctx, "literal ")) {
774
		type = GIT_DIFF_BINARY_LITERAL;
775 776
		git_parse_advance_chars(&ctx->parse_ctx, 8);
	} else if (git_parse_ctx_contains_s(&ctx->parse_ctx, "delta ")) {
777
		type = GIT_DIFF_BINARY_DELTA;
778
		git_parse_advance_chars(&ctx->parse_ctx, 6);
779
	} else {
780 781
		error = git_parse_err(
			"unknown binary delta type at line %"PRIuZ, ctx->parse_ctx.line_num);
782 783 784
		goto done;
	}

785 786
	if (git_parse_advance_digit(&len, &ctx->parse_ctx, 10) < 0 ||
	    git_parse_advance_nl(&ctx->parse_ctx) < 0 || len < 0) {
787
		error = git_parse_err("invalid binary size at line %"PRIuZ, ctx->parse_ctx.line_num);
788 789 790
		goto done;
	}

791
	while (ctx->parse_ctx.line_len) {
792
		char c;
793 794
		size_t encoded_len, decoded_len = 0, decoded_orig = decoded.size;

795 796
		git_parse_peek(&c, &ctx->parse_ctx, 0);

797 798 799 800 801 802 803 804
		if (c == '\n')
			break;
		else if (c >= 'A' && c <= 'Z')
			decoded_len = c - 'A' + 1;
		else if (c >= 'a' && c <= 'z')
			decoded_len = c - 'a' + (('z' - 'a') + 1) + 1;

		if (!decoded_len) {
805
			error = git_parse_err("invalid binary length at line %"PRIuZ, ctx->parse_ctx.line_num);
806 807 808
			goto done;
		}

809
		git_parse_advance_chars(&ctx->parse_ctx, 1);
810 811 812

		encoded_len = ((decoded_len / 4) + !!(decoded_len % 4)) * 5;

813
		if (!encoded_len || !ctx->parse_ctx.line_len || encoded_len > ctx->parse_ctx.line_len - 1) {
814
			error = git_parse_err("truncated binary data at line %"PRIuZ, ctx->parse_ctx.line_num);
815 816 817 818
			goto done;
		}

		if ((error = git_buf_decode_base85(
819
			&decoded, ctx->parse_ctx.line, encoded_len, decoded_len)) < 0)
820 821 822
			goto done;

		if (decoded.size - decoded_orig != decoded_len) {
823
			error = git_parse_err("truncated binary data at line %"PRIuZ, ctx->parse_ctx.line_num);
824 825 826
			goto done;
		}

827
		git_parse_advance_chars(&ctx->parse_ctx, encoded_len);
828

829 830
		if (git_parse_advance_nl(&ctx->parse_ctx) < 0) {
			error = git_parse_err("trailing data at line %"PRIuZ, ctx->parse_ctx.line_num);
831 832 833 834 835 836 837 838 839 840
			goto done;
		}
	}

	binary->type = type;
	binary->inflatedlen = (size_t)len;
	binary->datalen = decoded.size;
	binary->data = git_buf_detach(&decoded);

done:
841 842
	git_buf_dispose(&base85);
	git_buf_dispose(&decoded);
843 844 845
	return error;
}

846
static int parse_patch_binary(
847
	git_patch_parsed *patch,
848
	git_patch_parse_ctx *ctx)
849 850 851
{
	int error;

852 853 854
	if (git_parse_advance_expected_str(&ctx->parse_ctx, "GIT binary patch") < 0 ||
		git_parse_advance_nl(&ctx->parse_ctx) < 0)
		return git_parse_err("corrupt git binary header at line %"PRIuZ, ctx->parse_ctx.line_num);
855 856

	/* parse old->new binary diff */
857
	if ((error = parse_patch_binary_side(
858 859 860
			&patch->base.binary.new_file, ctx)) < 0)
		return error;

861 862 863
	if (git_parse_advance_nl(&ctx->parse_ctx) < 0)
		return git_parse_err("corrupt git binary separator at line %"PRIuZ,
			ctx->parse_ctx.line_num);
864 865

	/* parse new->old binary diff */
866
	if ((error = parse_patch_binary_side(
867 868 869
			&patch->base.binary.old_file, ctx)) < 0)
		return error;

870 871 872
	if (git_parse_advance_nl(&ctx->parse_ctx) < 0)
		return git_parse_err("corrupt git binary patch separator at line %"PRIuZ,
			ctx->parse_ctx.line_num);
873

874 875 876 877 878 879 880 881 882
	patch->base.binary.contains_data = 1;
	patch->base.delta->flags |= GIT_DIFF_FLAG_BINARY;
	return 0;
}

static int parse_patch_binary_nodata(
	git_patch_parsed *patch,
	git_patch_parse_ctx *ctx)
{
883 884 885 886 887 888
	const char *old = patch->old_path ? patch->old_path : patch->header_old_path;
	const char *new = patch->new_path ? patch->new_path : patch->header_new_path;

	if (!old || !new)
		return git_parse_err("corrupt binary data without paths at line %"PRIuZ, ctx->parse_ctx.line_num);

889 890 891 892 893
	if (patch->base.delta->status == GIT_DELTA_ADDED)
		old = "/dev/null";
	else if (patch->base.delta->status == GIT_DELTA_DELETED)
		new = "/dev/null";

894
	if (git_parse_advance_expected_str(&ctx->parse_ctx, "Binary files ") < 0 ||
895 896 897 898 899
	    git_parse_advance_expected_str(&ctx->parse_ctx, old) < 0 ||
	    git_parse_advance_expected_str(&ctx->parse_ctx, " and ") < 0 ||
	    git_parse_advance_expected_str(&ctx->parse_ctx, new) < 0 ||
	    git_parse_advance_expected_str(&ctx->parse_ctx, " differ") < 0 ||
	    git_parse_advance_nl(&ctx->parse_ctx) < 0)
900
		return git_parse_err("corrupt git binary header at line %"PRIuZ, ctx->parse_ctx.line_num);
901 902

	patch->base.binary.contains_data = 0;
903 904 905 906
	patch->base.delta->flags |= GIT_DIFF_FLAG_BINARY;
	return 0;
}

907
static int parse_patch_hunks(
908
	git_patch_parsed *patch,
909
	git_patch_parse_ctx *ctx)
910 911 912 913
{
	git_patch_hunk *hunk;
	int error = 0;

914
	while (git_parse_ctx_contains_s(&ctx->parse_ctx, "@@ -")) {
915
		hunk = git_array_alloc(patch->base.hunks);
916
		GIT_ERROR_CHECK_ALLOC(hunk);
917 918 919 920 921 922 923 924 925 926 927

		memset(hunk, 0, sizeof(git_patch_hunk));

		hunk->line_start = git_array_size(patch->base.lines);
		hunk->line_count = 0;

		if ((error = parse_hunk_header(hunk, ctx)) < 0 ||
			(error = parse_hunk_body(patch, hunk, ctx)) < 0)
			goto done;
	}

928 929
	patch->base.delta->flags |= GIT_DIFF_FLAG_NOT_BINARY;

930 931 932 933
done:
	return error;
}

934 935
static int parse_patch_body(
	git_patch_parsed *patch, git_patch_parse_ctx *ctx)
936
{
937
	if (git_parse_ctx_contains_s(&ctx->parse_ctx, "GIT binary patch"))
938
		return parse_patch_binary(patch, ctx);
939
	else if (git_parse_ctx_contains_s(&ctx->parse_ctx, "Binary files "))
940
		return parse_patch_binary_nodata(patch, ctx);
941
	else
942
		return parse_patch_hunks(patch, ctx);
943 944
}

945
static int check_header_names(
946 947 948 949 950 951 952 953 954
	const char *one,
	const char *two,
	const char *old_or_new,
	bool two_null)
{
	if (!one || !two)
		return 0;

	if (two_null && strcmp(two, "/dev/null") != 0)
955
		return git_parse_err("expected %s path of '/dev/null'", old_or_new);
956 957

	else if (!two_null && strcmp(one, two) != 0)
958
		return git_parse_err("mismatched %s path names", old_or_new);
959 960 961 962 963 964 965 966 967 968 969

	return 0;
}

static int check_prefix(
	char **out,
	size_t *out_len,
	git_patch_parsed *patch,
	const char *path_start)
{
	const char *path = path_start;
970
	size_t prefix_len = patch->ctx->opts.prefix_len;
971
	size_t remain_len = prefix_len;
972 973 974 975

	*out = NULL;
	*out_len = 0;

976
	if (prefix_len == 0)
977 978 979 980 981 982
		goto done;

	/* leading slashes do not count as part of the prefix in git apply */
	while (*path == '/')
		path++;

983
	while (*path && remain_len) {
984
		if (*path == '/')
985
			remain_len--;
986 987 988 989

		path++;
	}

990
	if (remain_len || !*path)
991
		return git_parse_err(
992
			"header filename does not contain %"PRIuZ" path components",
993
			prefix_len);
994 995 996 997 998

done:
	*out_len = (path - path_start);
	*out = git__strndup(path_start, *out_len);

999
	return (*out == NULL) ? -1 : 0;
1000 1001 1002
}

static int check_filenames(git_patch_parsed *patch)
1003
{
1004 1005 1006 1007 1008 1009
	const char *prefixed_new, *prefixed_old;
	size_t old_prefixlen = 0, new_prefixlen = 0;
	bool added = (patch->base.delta->status == GIT_DELTA_ADDED);
	bool deleted = (patch->base.delta->status == GIT_DELTA_DELETED);

	if (patch->old_path && !patch->new_path)
1010
		return git_parse_err("missing new path");
1011 1012

	if (!patch->old_path && patch->new_path)
1013
		return git_parse_err("missing old path");
1014 1015

	/* Ensure (non-renamed) paths match */
1016 1017
	if (check_header_names(patch->header_old_path, patch->old_path, "old", added) < 0 ||
	    check_header_names(patch->header_new_path, patch->new_path, "new", deleted) < 0)
1018 1019
		return -1;

1020 1021
	prefixed_old = (!added && patch->old_path) ? patch->old_path : patch->header_old_path;
	prefixed_new = (!deleted && patch->new_path) ? patch->new_path : patch->header_new_path;
1022

1023 1024
	if ((prefixed_old && check_prefix(&patch->old_prefix, &old_prefixlen, patch, prefixed_old) < 0) ||
	    (prefixed_new && check_prefix(&patch->new_prefix, &new_prefixlen, patch, prefixed_new) < 0))
1025 1026 1027 1028 1029
		return -1;

	/* Prefer the rename filenames as they are unambiguous and unprefixed */
	if (patch->rename_old_path)
		patch->base.delta->old_file.path = patch->rename_old_path;
1030
	else if (prefixed_old)
1031
		patch->base.delta->old_file.path = prefixed_old + old_prefixlen;
1032 1033
	else
		patch->base.delta->old_file.path = NULL;
1034 1035 1036

	if (patch->rename_new_path)
		patch->base.delta->new_file.path = patch->rename_new_path;
1037
	else if (prefixed_new)
1038
		patch->base.delta->new_file.path = prefixed_new + new_prefixlen;
1039 1040
	else
		patch->base.delta->new_file.path = NULL;
1041

1042
	if (!patch->base.delta->old_file.path &&
1043
	    !patch->base.delta->new_file.path)
1044
		return git_parse_err("git diff header lacks old / new paths");
1045

1046 1047 1048 1049 1050
	return 0;
}

static int check_patch(git_patch_parsed *patch)
{
1051 1052
	git_diff_delta *delta = patch->base.delta;

1053 1054
	if (check_filenames(patch) < 0)
		return -1;
1055

1056
	if (delta->old_file.path &&
1057 1058
	    delta->status != GIT_DELTA_DELETED &&
	    !delta->new_file.mode)
1059
		delta->new_file.mode = delta->old_file.mode;
1060

1061
	if (delta->status == GIT_DELTA_MODIFIED &&
1062 1063 1064
	    !(delta->flags & GIT_DIFF_FLAG_BINARY) &&
	    delta->new_file.mode == delta->old_file.mode &&
	    git_array_size(patch->base.hunks) == 0)
1065
		return git_parse_err("patch with no hunks");
1066

1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
	if (delta->status == GIT_DELTA_ADDED) {
		memset(&delta->old_file.id, 0x0, sizeof(git_oid));
		delta->old_file.id_abbrev = 0;
	}

	if (delta->status == GIT_DELTA_DELETED) {
		memset(&delta->new_file.id, 0x0, sizeof(git_oid));
		delta->new_file.id_abbrev = 0;
	}

1077 1078 1079
	return 0;
}

1080
git_patch_parse_ctx *git_patch_parse_ctx_init(
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090
	const char *content,
	size_t content_len,
	const git_patch_options *opts)
{
	git_patch_parse_ctx *ctx;
	git_patch_options default_opts = GIT_PATCH_OPTIONS_INIT;

	if ((ctx = git__calloc(1, sizeof(git_patch_parse_ctx))) == NULL)
		return NULL;

1091 1092 1093
	if ((git_parse_ctx_init(&ctx->parse_ctx, content, content_len)) < 0) {
		git__free(ctx);
		return NULL;
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
	}

	if (opts)
		memcpy(&ctx->opts, opts, sizeof(git_patch_options));
	else
		memcpy(&ctx->opts, &default_opts, sizeof(git_patch_options));

	GIT_REFCOUNT_INC(ctx);
	return ctx;
}

static void patch_parse_ctx_free(git_patch_parse_ctx *ctx)
{
	if (!ctx)
		return;

1110
	git_parse_ctx_clear(&ctx->parse_ctx);
1111 1112 1113
	git__free(ctx);
}

1114
void git_patch_parse_ctx_free(git_patch_parse_ctx *ctx)
1115 1116 1117 1118
{
	GIT_REFCOUNT_DEC(ctx, patch_parse_ctx_free);
}

1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
int git_patch_parsed_from_diff(git_patch **out, git_diff *d, size_t idx)
{
	git_diff_parsed *diff = (git_diff_parsed *)d;
	git_patch *p;

	if ((p = git_vector_get(&diff->patches, idx)) == NULL)
		return -1;

	GIT_REFCOUNT_INC(p);
	*out = p;

	return 0;
}

1133 1134 1135
static void patch_parsed__free(git_patch *p)
{
	git_patch_parsed *patch = (git_patch_parsed *)p;
1136 1137
	git_diff_line *line;
	size_t i;
1138 1139 1140 1141

	if (!patch)
		return;

1142 1143
	git_patch_parse_ctx_free(patch->ctx);

1144 1145 1146
	git__free((char *)patch->base.binary.old_file.data);
	git__free((char *)patch->base.binary.new_file.data);
	git_array_clear(patch->base.hunks);
1147 1148
	git_array_foreach(patch->base.lines, i, line)
		git__free((char *) line->content);
1149 1150 1151
	git_array_clear(patch->base.lines);
	git__free(patch->base.delta);

1152 1153 1154 1155 1156 1157 1158 1159
	git__free(patch->old_prefix);
	git__free(patch->new_prefix);
	git__free(patch->header_old_path);
	git__free(patch->header_new_path);
	git__free(patch->rename_old_path);
	git__free(patch->rename_new_path);
	git__free(patch->old_path);
	git__free(patch->new_path);
1160
	git__free(patch);
1161 1162
}

1163
int git_patch_parse(
1164
	git_patch **out,
1165
	git_patch_parse_ctx *ctx)
1166 1167
{
	git_patch_parsed *patch;
1168
	size_t start, used;
1169 1170
	int error = 0;

Edward Thomson committed
1171 1172
	GIT_ASSERT_ARG(out);
	GIT_ASSERT_ARG(ctx);
1173

1174 1175 1176
	*out = NULL;

	patch = git__calloc(1, sizeof(git_patch_parsed));
1177
	GIT_ERROR_CHECK_ALLOC(patch);
1178

1179 1180
	patch->ctx = ctx;
	GIT_REFCOUNT_INC(patch->ctx);
1181 1182

	patch->base.free_fn = patch_parsed__free;
1183

1184
	patch->base.delta = git__calloc(1, sizeof(git_diff_delta));
1185
	GIT_ERROR_CHECK_ALLOC(patch->base.delta);
1186

1187
	patch->base.delta->status = GIT_DELTA_MODIFIED;
1188
	patch->base.delta->nfiles = 2;
1189

1190
	start = ctx->parse_ctx.remain_len;
1191

1192 1193
	if ((error = parse_patch_header(patch, ctx)) < 0 ||
		(error = parse_patch_body(patch, ctx)) < 0 ||
1194 1195 1196
		(error = check_patch(patch)) < 0)
		goto done;

1197 1198
	used = start - ctx->parse_ctx.remain_len;
	ctx->parse_ctx.remain += used;
1199

1200 1201 1202 1203
	patch->base.diff_opts.old_prefix = patch->old_prefix;
	patch->base.diff_opts.new_prefix = patch->new_prefix;
	patch->base.diff_opts.flags |= GIT_DIFF_SHOW_BINARY;

1204
	GIT_REFCOUNT_INC(&patch->base);
1205 1206 1207
	*out = &patch->base;

done:
1208 1209 1210
	if (error < 0)
		patch_parsed__free(&patch->base);

1211 1212
	return error;
}
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223

int git_patch_from_buffer(
	git_patch **out,
	const char *content,
	size_t content_len,
	const git_patch_options *opts)
{
	git_patch_parse_ctx *ctx;
	int error;

	ctx = git_patch_parse_ctx_init(content, content_len, opts);
1224
	GIT_ERROR_CHECK_ALLOC(ctx);
1225 1226 1227 1228 1229 1230 1231

	error = git_patch_parse(out, ctx);

	git_patch_parse_ctx_free(ctx);
	return error;
}