2 * Copyright (c) 2020 Ori Bernstein
3 * Copyright (c) 2021 Stefan Sperling <stsp@openbsd.org>
5 * Permission to use, copy, modify, and distribute this software for any
6 * purpose with or without fee is hereby granted, provided that the above
7 * copyright notice and this permission notice appear in all copies.
9 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18 #include <sys/types.h>
28 #include "got_error.h"
30 #include "got_lib_deltify.h"
31 #include "murmurhash2.h"
34 #define MIN(_a,_b) ((_a) < (_b) ? (_a) : (_b))
38 * The algorihm used here is FastCDC (Fast Content-Defined Chunking)
39 * https://www.usenix.org/conference/atc16/technical-sessions/presentation/xia
42 static uint32_t geartab
[256] = {
43 0x67ed26b7, 0x32da500c, 0x53d0fee0, 0xce387dc7, 0xcd406d90, 0x2e83a4d4,
44 0x9fc9a38d, 0xb67259dc, 0xca6b1722, 0x6d2ea08c, 0x235cea2e, 0x3149bb5f,
45 0x1beda787, 0x2a6b77d5, 0x2f22d9ac, 0x91fc0544, 0xe413acfa, 0x5a30ff7a,
46 0xad6fdde0, 0x444fd0f5, 0x7ad87864, 0x58c5ff05, 0x8d2ec336, 0x2371f853,
47 0x550f8572, 0x6aa448dd, 0x7c9ddbcf, 0x95221e14, 0x2a82ec33, 0xcbec5a78,
48 0xc6795a0d, 0x243995b7, 0x1c909a2f, 0x4fded51c, 0x635d334b, 0x0e2b9999,
49 0x2702968d, 0x856de1d5, 0x3325d60e, 0xeb6a7502, 0xec2a9844, 0x0905835a,
50 0xa1820375, 0xa4be5cab, 0x96a6c058, 0x2c2ccd70, 0xba40fce3, 0xd794c46b,
51 0x8fbae83e, 0xc3aa7899, 0x3d3ff8ed, 0xa0d42b5b, 0x571c0c97, 0xd2811516,
52 0xf7e7b96c, 0x4fd2fcbd, 0xe2fdec94, 0x282cc436, 0x78e8e95c, 0x80a3b613,
53 0xcfbee20c, 0xd4a32d1c, 0x2a12ff13, 0x6af82936, 0xe5630258, 0x8efa6a98,
54 0x294fb2d1, 0xdeb57086, 0x5f0fddb3, 0xeceda7ce, 0x4c87305f, 0x3a6d3307,
55 0xe22d2942, 0x9d060217, 0x1e42ed02, 0xb6f63b52, 0x4367f39f, 0x055cf262,
56 0x03a461b2, 0x5ef9e382, 0x386bc03a, 0x2a1e79c7, 0xf1a0058b, 0xd4d2dea9,
57 0x56baf37d, 0x5daff6cc, 0xf03a951d, 0xaef7de45, 0xa8f4581e, 0x3960b555,
58 0xffbfff6d, 0xbe702a23, 0x8f5b6d6f, 0x061739fb, 0x98696f47, 0x3fd596d4,
59 0x151eac6b, 0xa9fcc4f5, 0x69181a12, 0x3ac5a107, 0xb5198fe7, 0x96bcb1da,
60 0x1b5ddf8e, 0xc757d650, 0x65865c3a, 0x8fc0a41a, 0x87435536, 0x99eda6f2,
61 0x41874794, 0x29cff4e8, 0xb70efd9a, 0x3103f6e7, 0x84d2453b, 0x15a450bd,
62 0x74f49af1, 0x60f664b1, 0xa1c86935, 0xfdafbce1, 0xe36353e3, 0x5d9ba739,
63 0xbc0559ba, 0x708b0054, 0xd41d808c, 0xb2f31723, 0x9027c41f, 0xf136d165,
64 0xb5374b12, 0x9420a6ac, 0x273958b6, 0xe6c2fad0, 0xebdc1f21, 0xfb33af8b,
65 0xc71c25cd, 0xe9a2d8e5, 0xbeb38a50, 0xbceb7cc2, 0x4e4e73f0, 0xcd6c251d,
66 0xde4c032c, 0x4b04ac30, 0x725b8b21, 0x4eb8c33b, 0x20d07b75, 0x0567aa63,
67 0xb56b2bb7, 0xc1f5fd3a, 0xcafd35ca, 0x470dd4da, 0xfe4f94cd, 0xfb8de424,
68 0xe8dbcf40, 0xfe50a37a, 0x62db5b5d, 0xf32f4ab6, 0x2c4a8a51, 0x18473dc0,
69 0xfe0cbb6e, 0xfe399efd, 0xdf34ecc9, 0x6ccd5055, 0x46097073, 0x139135c2,
70 0x721c76f6, 0x1c6a94b4, 0x6eee014d, 0x8a508e02, 0x3da538f5, 0x280d394f,
71 0x5248a0c4, 0x3ce94c6c, 0x9a71ad3a, 0x8493dd05, 0xe43f0ab6, 0x18e4ed42,
72 0x6c5c0e09, 0x42b06ec9, 0x8d330343, 0xa45b6f59, 0x2a573c0c, 0xd7fd3de6,
73 0xeedeab68, 0x5c84dafc, 0xbbd1b1a8, 0xa3ce1ad1, 0x85b70bed, 0xb6add07f,
74 0xa531309c, 0x8f8ab852, 0x564de332, 0xeac9ed0c, 0x73da402c, 0x3ec52761,
75 0x43af2f4d, 0xd6ff45c8, 0x4c367462, 0xd553bd6a, 0x44724855, 0x3b2aa728,
76 0x56e5eb65, 0xeaf16173, 0x33fa42ff, 0xd714bb5d, 0xfbd0a3b9, 0xaf517134,
77 0x9416c8cd, 0x534cf94f, 0x548947c2, 0x34193569, 0x32f4389a, 0xfe7028bc,
78 0xed73b1ed, 0x9db95770, 0x468e3922, 0x0440c3cd, 0x60059a62, 0x33504562,
79 0x2b229fbd, 0x5174dca5, 0xf7028752, 0xd63c6aa8, 0x31276f38, 0x0646721c,
80 0xb0191da8, 0xe00e6de0, 0x9eac1a6e, 0x9f7628a5, 0xed6c06ea, 0x0bb8af15,
81 0xf119fb12, 0x38693c1c, 0x732bc0fe, 0x84953275, 0xb82ec888, 0x33a4f1b3,
82 0x3099835e, 0x028a8782, 0x5fdd51d7, 0xc6c717b3, 0xb06caf71, 0x17c8c111,
83 0x61bad754, 0x9fd03061, 0xe09df1af, 0x3bc9eb73, 0x85878413, 0x9889aaf2,
84 0x3f5a9e46, 0x42c9f01f, 0x9984a4f4, 0xd5de43cc, 0xd294daed, 0xbecba2d2,
85 0xf1f6e72c, 0x5551128a, 0x83af87e2, 0x6f0342ba,
89 hashblk(const unsigned char *p
, off_t n
)
91 return murmurhash2(p
, n
, 0x1d7c5ac3);
94 static const struct got_error
*
95 addblk(struct got_delta_table
*dt
, FILE *f
, off_t file_offset0
, off_t len
,
96 off_t offset
, uint64_t h
)
98 const struct got_error
*err
= NULL
;
100 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
101 uint8_t buf2
[GOT_DELTIFY_MAXCHUNK
];
108 while (dt
->blocks
[i
].len
!= 0) {
110 * Avoid adding duplicate blocks.
111 * NB: A matching hash is insufficient for detecting equality.
112 * The hash can only detect inequality.
114 if (len
== dt
->blocks
[i
].len
&& h
== dt
->blocks
[i
].hash
) {
116 if (fseeko(f
, file_offset0
+ offset
, SEEK_SET
) == -1)
117 return got_error_from_errno("fseeko");
118 r
= fread(buf
, 1, len
, f
);
122 return got_ferror(f
, GOT_ERR_IO
);
125 if (fseeko(f
, file_offset0
+ dt
->blocks
[i
].offset
,
127 return got_error_from_errno("fseeko");
128 if (fread(buf2
, 1, len
, f
) != len
)
129 return got_ferror(f
, GOT_ERR_IO
);
130 if (memcmp(buf
, buf2
, len
) == 0)
134 i
= (i
+ 1) % dt
->nalloc
;
136 assert(dt
->blocks
[i
].len
== 0);
137 dt
->blocks
[i
].len
= len
;
138 dt
->blocks
[i
].offset
= offset
;
139 dt
->blocks
[i
].hash
= h
;
141 if (dt
->nalloc
< dt
->nblocks
+ 64) {
142 struct got_delta_block
*db
;
143 size_t old_size
= dt
->nalloc
;
145 dt
->blocks
= calloc(dt
->nalloc
+ 64,
146 sizeof(struct got_delta_block
));
147 if (dt
->blocks
== NULL
) {
148 err
= got_error_from_errno("calloc");
154 * Recompute all block positions. Hash-based indices of blocks
155 * in the array depend on the allocated length of the array.
158 for (i
= 0; i
< old_size
; i
++) {
161 err
= addblk(dt
, f
, file_offset0
, db
[i
].len
,
162 db
[i
].offset
, db
[i
].hash
);
172 static const struct got_error
*
173 addblk_mem(struct got_delta_table
*dt
, uint8_t *data
, off_t file_offset0
,
174 off_t len
, off_t offset
, uint64_t h
)
176 const struct got_error
*err
= NULL
;
185 while (dt
->blocks
[i
].len
!= 0) {
187 * Avoid adding duplicate blocks.
188 * NB: A matching hash is insufficient for detecting equality.
189 * The hash can only detect inequality.
191 if (len
== dt
->blocks
[i
].len
&& h
== dt
->blocks
[i
].hash
) {
192 block1
= data
+ file_offset0
+ dt
->blocks
[i
].offset
;
193 block2
= data
+ file_offset0
+ offset
;
194 if (memcmp(block1
, block2
, len
) == 0)
198 i
= (i
+ 1) % dt
->nalloc
;
200 assert(dt
->blocks
[i
].len
== 0);
201 dt
->blocks
[i
].len
= len
;
202 dt
->blocks
[i
].offset
= offset
;
203 dt
->blocks
[i
].hash
= h
;
205 if (dt
->nalloc
< dt
->nblocks
+ 64) {
206 struct got_delta_block
*db
;
207 size_t old_size
= dt
->nalloc
;
209 dt
->blocks
= calloc(dt
->nalloc
+ 64,
210 sizeof(struct got_delta_block
));
211 if (dt
->blocks
== NULL
) {
212 err
= got_error_from_errno("calloc");
218 * Recompute all block positions. Hash-based indices of blocks
219 * in the array depend on the allocated length of the array.
222 for (i
= 0; i
< old_size
; i
++) {
225 err
= addblk_mem(dt
, data
, file_offset0
, db
[i
].len
,
226 db
[i
].offset
, db
[i
].hash
);
236 static const struct got_error
*
237 lookupblk(struct got_delta_block
**block
, struct got_delta_table
*dt
,
238 unsigned char *p
, off_t len
, FILE *basefile
, off_t basefile_offset0
)
242 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
248 for (i
= h
% dt
->nalloc
; dt
->blocks
[i
].len
!= 0;
249 i
= (i
+ 1) % dt
->nalloc
) {
250 if (dt
->blocks
[i
].hash
!= h
||
251 dt
->blocks
[i
].len
!= len
)
253 if (fseeko(basefile
, basefile_offset0
+ dt
->blocks
[i
].offset
,
255 return got_error_from_errno("fseeko");
256 r
= fread(buf
, 1, len
, basefile
);
258 return got_ferror(basefile
, GOT_ERR_IO
);
259 if (memcmp(p
, buf
, len
) == 0) {
260 *block
= &dt
->blocks
[i
];
267 static const struct got_error
*
268 lookupblk_mem(struct got_delta_block
**block
, struct got_delta_table
*dt
,
269 unsigned char *p
, off_t len
, uint8_t *basedata
, off_t basefile_offset0
)
278 for (i
= h
% dt
->nalloc
; dt
->blocks
[i
].len
!= 0;
279 i
= (i
+ 1) % dt
->nalloc
) {
280 if (dt
->blocks
[i
].hash
!= h
||
281 dt
->blocks
[i
].len
!= len
)
283 b
= basedata
+ basefile_offset0
+ dt
->blocks
[i
].offset
;
284 if (memcmp(p
, b
, len
) == 0) {
285 *block
= &dt
->blocks
[i
];
292 static const struct got_error
*
293 nextblk(uint8_t *buf
, off_t
*blocklen
, FILE *f
)
296 const unsigned char *p
;
298 off_t pos
= ftello(f
);
302 r
= fread(buf
, 1, GOT_DELTIFY_MAXCHUNK
, f
);
303 if (r
== 0 && ferror(f
))
304 return got_ferror(f
, GOT_ERR_IO
);
305 if (r
< GOT_DELTIFY_MINCHUNK
)
306 return NULL
; /* no more delta-worthy blocks left */
308 /* Got a deltifiable block. Find the split-point where it ends. */
309 p
= buf
+ GOT_DELTIFY_MINCHUNK
;
311 while (p
!= buf
+ r
) {
312 gh
= (gh
<< 1) + geartab
[*p
++];
313 if ((gh
& GOT_DELTIFY_SPLITMASK
) == 0)
317 *blocklen
= (p
- buf
);
318 if (fseeko(f
, pos
+ *blocklen
, SEEK_SET
) == -1)
319 return got_error_from_errno("fseeko");
324 static const struct got_error
*
325 nextblk_mem(off_t
*blocklen
, uint8_t *data
, off_t fileoffset
, off_t filesize
)
328 const unsigned char *p
;
332 if (fileoffset
>= filesize
||
333 filesize
- fileoffset
< GOT_DELTIFY_MINCHUNK
)
334 return NULL
; /* no more delta-worthy blocks left */
336 /* Got a deltifiable block. Find the split-point where it ends. */
337 p
= data
+ fileoffset
+ GOT_DELTIFY_MINCHUNK
;
339 while (p
!= data
+ MIN(fileoffset
+ GOT_DELTIFY_MAXCHUNK
, filesize
)) {
340 gh
= (gh
<< 1) + geartab
[*p
++];
341 if ((gh
& GOT_DELTIFY_SPLITMASK
) == 0)
345 *blocklen
= (p
- (data
+ fileoffset
));
349 const struct got_error
*
350 got_deltify_init(struct got_delta_table
**dt
, FILE *f
, off_t fileoffset
,
353 const struct got_error
*err
= NULL
;
355 const off_t offset0
= fileoffset
;
357 *dt
= calloc(1, sizeof(**dt
));
359 return got_error_from_errno("calloc");
363 (*dt
)->blocks
= calloc((*dt
)->nalloc
, sizeof(struct got_delta_block
));
364 if ((*dt
)->blocks
== NULL
) {
365 err
= got_error_from_errno("calloc");
369 if (fseeko(f
, fileoffset
, SEEK_SET
) == -1)
370 return got_error_from_errno("fseeko");
372 while (fileoffset
< filesize
) {
373 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
375 err
= nextblk(buf
, &blocklen
, f
);
380 h
= hashblk(buf
, blocklen
);
381 err
= addblk(*dt
, f
, offset0
, blocklen
,
382 fileoffset
- offset0
, h
);
385 fileoffset
+= blocklen
;
386 if (fseeko(f
, fileoffset
, SEEK_SET
) == -1)
387 return got_error_from_errno("fseeko");
399 const struct got_error
*
400 got_deltify_init_mem(struct got_delta_table
**dt
, uint8_t *data
,
401 off_t fileoffset
, off_t filesize
)
403 const struct got_error
*err
= NULL
;
405 const off_t offset0
= fileoffset
;
407 *dt
= calloc(1, sizeof(**dt
));
409 return got_error_from_errno("calloc");
413 (*dt
)->blocks
= calloc((*dt
)->nalloc
, sizeof(struct got_delta_block
));
414 if ((*dt
)->blocks
== NULL
) {
415 err
= got_error_from_errno("calloc");
419 while (fileoffset
< filesize
) {
421 err
= nextblk_mem(&blocklen
, data
, fileoffset
, filesize
);
426 h
= hashblk(data
+ fileoffset
, blocklen
);
427 err
= addblk_mem(*dt
, data
, offset0
, blocklen
,
428 fileoffset
- offset0
, h
);
431 fileoffset
+= blocklen
;
444 got_deltify_free(struct got_delta_table
*dt
)
452 static const struct got_error
*
453 emitdelta(struct got_delta_instruction
**deltas
, size_t *nalloc
, int *ndeltas
,
454 const size_t alloc_chunk_size
, int copy
, off_t offset
, off_t len
)
456 struct got_delta_instruction
*d
, *p
;
458 if (*nalloc
< *ndeltas
+ alloc_chunk_size
) {
459 p
= reallocarray(*deltas
, *nalloc
+ alloc_chunk_size
,
460 sizeof(struct got_delta_instruction
));
462 return got_error_from_errno("reallocarray");
464 *nalloc
+= alloc_chunk_size
;
467 d
= &(*deltas
)[*ndeltas
- 1];
474 static const struct got_error
*
475 stretchblk(FILE *basefile
, off_t base_offset0
, struct got_delta_block
*block
,
476 FILE *f
, off_t filesize
, off_t
*blocklen
)
478 uint8_t basebuf
[GOT_DELTIFY_MAXCHUNK
], buf
[GOT_DELTIFY_MAXCHUNK
];
482 if (fseeko(basefile
, base_offset0
+ block
->offset
+ *blocklen
,
484 return got_error_from_errno("fseeko");
486 while (buf_equal
&& *blocklen
< (1 << 24) - 1) {
487 base_r
= fread(basebuf
, 1, sizeof(basebuf
), basefile
);
489 if (ferror(basefile
))
490 return got_ferror(basefile
, GOT_ERR_IO
);
493 r
= fread(buf
, 1, sizeof(buf
), f
);
496 return got_ferror(f
, GOT_ERR_IO
);
499 for (i
= 0; i
< MIN(base_r
, r
); i
++) {
500 if (buf
[i
] != basebuf
[i
]) {
511 static const struct got_error
*
512 stretchblk_file_mem(uint8_t *basedata
, off_t base_offset0
, off_t basefile_size
,
513 struct got_delta_block
*block
, FILE *f
, off_t filesize
, off_t
*blocklen
)
515 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
518 off_t base_offset
= base_offset0
+ block
->offset
+ *blocklen
;
520 if (base_offset
> basefile_size
) {
521 return got_error_fmt(GOT_ERR_RANGE
,
522 "read beyond the size of delta base at offset %llu",
526 while (buf_equal
&& *blocklen
< (1 << 24) - 1) {
527 if (base_offset
+ *blocklen
>= basefile_size
)
529 r
= fread(buf
, 1, sizeof(buf
), f
);
532 return got_ferror(f
, GOT_ERR_IO
);
535 for (i
= 0; i
< MIN(basefile_size
- base_offset
, r
); i
++) {
536 if (buf
[i
] != *(basedata
+ base_offset
+ i
)) {
547 static const struct got_error
*
548 stretchblk_mem_file(FILE *basefile
, off_t base_offset0
,
549 struct got_delta_block
*block
, uint8_t *data
, off_t fileoffset
,
550 off_t filesize
, off_t
*blocklen
)
552 uint8_t basebuf
[GOT_DELTIFY_MAXCHUNK
];
556 if (fileoffset
> filesize
) {
557 return got_error_fmt(GOT_ERR_RANGE
,
558 "read beyond the size of deltify file at offset %llu",
562 if (fseeko(basefile
, base_offset0
+ block
->offset
+ *blocklen
,
564 return got_error_from_errno("fseeko");
566 while (buf_equal
&& *blocklen
< (1 << 24) - 1) {
567 if (fileoffset
+ *blocklen
>= filesize
)
569 base_r
= fread(basebuf
, 1, sizeof(basebuf
), basefile
);
571 if (ferror(basefile
))
572 return got_ferror(basefile
, GOT_ERR_IO
);
575 for (i
= 0; i
< MIN(base_r
, filesize
- fileoffset
); i
++) {
576 if (*(data
+ fileoffset
+ i
) != basebuf
[i
]) {
587 static const struct got_error
*
588 stretchblk_mem_mem(uint8_t *basedata
, off_t base_offset0
, off_t basefile_size
,
589 struct got_delta_block
*block
, uint8_t *data
, off_t fileoffset
,
590 off_t filesize
, off_t
*blocklen
)
593 off_t base_offset
= base_offset0
+ block
->offset
+ *blocklen
;
596 if (base_offset
> basefile_size
) {
597 return got_error_fmt(GOT_ERR_RANGE
,
598 "read beyond the size of delta base at offset %llu",
602 if (fileoffset
> filesize
) {
603 return got_error_fmt(GOT_ERR_RANGE
,
604 "read beyond the size of deltify file at offset %llu",
608 p
= data
+ fileoffset
;
609 q
= basedata
+ base_offset
;
610 maxlen
= MIN(basefile_size
- base_offset
, filesize
- fileoffset
);
611 for (i
= 0; i
< maxlen
&& *blocklen
< (1 << 24) - 1; i
++) {
620 const struct got_error
*
621 got_deltify(struct got_delta_instruction
**deltas
, int *ndeltas
,
622 FILE *f
, off_t fileoffset
, off_t filesize
,
623 struct got_delta_table
*dt
, FILE *basefile
,
624 off_t basefile_offset0
, off_t basefile_size
)
626 const struct got_error
*err
= NULL
;
627 const off_t offset0
= fileoffset
;
629 const size_t alloc_chunk_size
= 64;
635 * offset0 indicates where data to be deltified begins.
636 * For example, we want to avoid deltifying a Git object header at
637 * the beginning of the file.
639 if (fseeko(f
, offset0
, SEEK_SET
) == -1)
640 return got_error_from_errno("fseeko");
642 *deltas
= reallocarray(NULL
, alloc_chunk_size
,
643 sizeof(struct got_delta_instruction
));
645 return got_error_from_errno("reallocarray");
646 nalloc
= alloc_chunk_size
;
648 while (fileoffset
< filesize
) {
649 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
651 struct got_delta_block
*block
;
652 err
= nextblk(buf
, &blocklen
, f
);
656 /* Source remainder from the file itself. */
657 if (fileoffset
< filesize
) {
658 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
659 alloc_chunk_size
, 0, fileoffset
- offset0
,
660 filesize
- fileoffset
);
664 err
= lookupblk(&block
, dt
, buf
, blocklen
, basefile
,
670 * We have found a matching block in the delta base.
671 * Attempt to stretch the block as far as possible and
672 * generate a copy instruction.
674 err
= stretchblk(basefile
, basefile_offset0
, block
,
675 f
, filesize
, &blocklen
);
678 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
679 alloc_chunk_size
, 1, block
->offset
, blocklen
);
685 * This block needs to be sourced from the file itself.
687 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
688 alloc_chunk_size
, 0, fileoffset
- offset0
, blocklen
);
692 fileoffset
+= blocklen
;
693 if (fseeko(f
, fileoffset
, SEEK_SET
) == -1) {
694 err
= got_error_from_errno("fseeko");
707 const struct got_error
*
708 got_deltify_file_mem(struct got_delta_instruction
**deltas
, int *ndeltas
,
709 FILE *f
, off_t fileoffset
, off_t filesize
,
710 struct got_delta_table
*dt
, uint8_t *basedata
,
711 off_t basefile_offset0
, off_t basefile_size
)
713 const struct got_error
*err
= NULL
;
714 const off_t offset0
= fileoffset
;
716 const size_t alloc_chunk_size
= 64;
722 * offset0 indicates where data to be deltified begins.
723 * For example, we want to avoid deltifying a Git object header at
724 * the beginning of the file.
726 if (fseeko(f
, offset0
, SEEK_SET
) == -1)
727 return got_error_from_errno("fseeko");
729 *deltas
= reallocarray(NULL
, alloc_chunk_size
,
730 sizeof(struct got_delta_instruction
));
732 return got_error_from_errno("reallocarray");
733 nalloc
= alloc_chunk_size
;
735 while (fileoffset
< filesize
) {
736 uint8_t buf
[GOT_DELTIFY_MAXCHUNK
];
738 struct got_delta_block
*block
;
739 err
= nextblk(buf
, &blocklen
, f
);
743 /* Source remainder from the file itself. */
744 if (fileoffset
< filesize
) {
745 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
746 alloc_chunk_size
, 0, fileoffset
- offset0
,
747 filesize
- fileoffset
);
751 err
= lookupblk_mem(&block
, dt
, buf
, blocklen
, basedata
,
757 * We have found a matching block in the delta base.
758 * Attempt to stretch the block as far as possible and
759 * generate a copy instruction.
761 err
= stretchblk_file_mem(basedata
, basefile_offset0
,
762 basefile_size
, block
, f
, filesize
, &blocklen
);
765 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
766 alloc_chunk_size
, 1, block
->offset
, blocklen
);
772 * This block needs to be sourced from the file itself.
774 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
775 alloc_chunk_size
, 0, fileoffset
- offset0
, blocklen
);
779 fileoffset
+= blocklen
;
780 if (fseeko(f
, fileoffset
, SEEK_SET
) == -1) {
781 err
= got_error_from_errno("fseeko");
794 const struct got_error
*
795 got_deltify_mem_file(struct got_delta_instruction
**deltas
, int *ndeltas
,
796 uint8_t *data
, off_t fileoffset
, off_t filesize
,
797 struct got_delta_table
*dt
, FILE *basefile
,
798 off_t basefile_offset0
, off_t basefile_size
)
800 const struct got_error
*err
= NULL
;
801 const off_t offset0
= fileoffset
;
803 const size_t alloc_chunk_size
= 64;
808 *deltas
= reallocarray(NULL
, alloc_chunk_size
,
809 sizeof(struct got_delta_instruction
));
811 return got_error_from_errno("reallocarray");
812 nalloc
= alloc_chunk_size
;
814 while (fileoffset
< filesize
) {
816 struct got_delta_block
*block
;
817 err
= nextblk_mem(&blocklen
, data
, fileoffset
, filesize
);
821 /* Source remainder from the file itself. */
822 if (fileoffset
< filesize
) {
823 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
824 alloc_chunk_size
, 0, fileoffset
- offset0
,
825 filesize
- fileoffset
);
829 err
= lookupblk(&block
, dt
, data
+ fileoffset
, blocklen
,
830 basefile
, basefile_offset0
);
835 * We have found a matching block in the delta base.
836 * Attempt to stretch the block as far as possible and
837 * generate a copy instruction.
839 err
= stretchblk_mem_file(basefile
, basefile_offset0
,
840 block
, data
, fileoffset
+ blocklen
, filesize
,
844 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
845 alloc_chunk_size
, 1, block
->offset
, blocklen
);
851 * This block needs to be sourced from the file itself.
853 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
854 alloc_chunk_size
, 0, fileoffset
- offset0
, blocklen
);
858 fileoffset
+= blocklen
;
869 const struct got_error
*
870 got_deltify_mem_mem(struct got_delta_instruction
**deltas
, int *ndeltas
,
871 uint8_t *data
, off_t fileoffset
, off_t filesize
,
872 struct got_delta_table
*dt
, uint8_t *basedata
,
873 off_t basefile_offset0
, off_t basefile_size
)
875 const struct got_error
*err
= NULL
;
876 const off_t offset0
= fileoffset
;
878 const size_t alloc_chunk_size
= 64;
883 *deltas
= reallocarray(NULL
, alloc_chunk_size
,
884 sizeof(struct got_delta_instruction
));
886 return got_error_from_errno("reallocarray");
887 nalloc
= alloc_chunk_size
;
889 while (fileoffset
< filesize
) {
891 struct got_delta_block
*block
;
892 err
= nextblk_mem(&blocklen
, data
, fileoffset
, filesize
);
896 /* Source remainder from the file itself. */
897 if (fileoffset
< filesize
) {
898 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
899 alloc_chunk_size
, 0, fileoffset
- offset0
,
900 filesize
- fileoffset
);
904 err
= lookupblk_mem(&block
, dt
, data
+ fileoffset
, blocklen
,
905 basedata
, basefile_offset0
);
910 * We have found a matching block in the delta base.
911 * Attempt to stretch the block as far as possible and
912 * generate a copy instruction.
914 err
= stretchblk_mem_mem(basedata
, basefile_offset0
,
915 basefile_size
, block
, data
, fileoffset
+ blocklen
,
916 filesize
, &blocklen
);
919 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
920 alloc_chunk_size
, 1, block
->offset
, blocklen
);
926 * This block needs to be sourced from the file itself.
928 err
= emitdelta(deltas
, &nalloc
, ndeltas
,
929 alloc_chunk_size
, 0, fileoffset
- offset0
, blocklen
);
933 fileoffset
+= blocklen
;