2 * diff-delta.c: generate a delta between two buffers
4 * Many parts of this file have been lifted from LibXDiff version 0.10.
5 * http://www.xmailserver.org/xdiff-lib.html
7 * LibXDiff was written by Davide Libenzi <davidel@xmailserver.org>
8 * Copyright (C) 2003 Davide Libenzi
10 * Many mods for GIT usage by Nicolas Pitre <nico@cam.org>, (C) 2005.
12 * This file is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * Use of this within git automatically means that the LGPL
18 * licensing gets turned into GPLv2 within this project.
27 const unsigned char *ptr
;
31 static struct index
** delta_index(const unsigned char *buf
,
32 unsigned long bufsize
,
33 unsigned long trg_bufsize
)
36 unsigned int i
, hshift
, hlimit
, *hash_count
;
37 const unsigned char *data
;
38 struct index
*entry
, **hash
;
41 /* determine index hash size */
43 for (i
= 8; (1 << i
) < hsize
&& i
< 24; i
+= 2);
48 * Allocate lookup index. Note the first hash pointer
49 * is used to store the hash shift value.
51 mem
= malloc((1 + hsize
) * sizeof(*hash
) + bufsize
* sizeof(*entry
));
55 *hash
++ = (void *)hshift
;
56 entry
= mem
+ (1 + hsize
) * sizeof(*hash
);
57 memset(hash
, 0, hsize
* sizeof(*hash
));
59 /* allocate an array to count hash entries */
60 hash_count
= calloc(hsize
, sizeof(*hash_count
));
66 /* then populate the index */
67 data
= buf
+ bufsize
- 2;
70 i
= data
[0] ^ ((data
[1] ^ (data
[2] << hshift
)) << hshift
);
71 entry
->next
= hash
[i
];
77 * Determine a limit on the number of entries in the same hash
78 * bucket. This guard us against patological data sets causing
79 * really bad hash distribution with most entries in the same hash
80 * bucket that would bring us to O(m*n) computing costs (m and n
81 * corresponding to reference and target buffer sizes).
83 * The more the target buffer is large, the more it is important to
84 * have small entry lists for each hash buckets. With such a limit
85 * the cost is bounded to something more like O(m+n).
87 hlimit
= (1 << 26) / trg_bufsize
;
92 * Now make sure none of the hash buckets has more entries than
93 * we're willing to test. Otherwise we short-circuit the entry
94 * list uniformly to still preserve a good repartition across
95 * the reference buffer.
97 for (i
= 0; i
< hsize
; i
++) {
98 if (hash_count
[i
] < hlimit
)
102 struct index
*keep
= entry
;
103 int skip
= hash_count
[i
] / hlimit
/ 2;
106 } while(--skip
&& entry
);
115 /* provide the size of the copy opcode given the block offset and size */
116 #define COPYOP_SIZE(o, s) \
117 (!!(o & 0xff) + !!(o & 0xff00) + !!(o & 0xff0000) + !!(o & 0xff000000) + \
118 !!(s & 0xff) + !!(s & 0xff00) + 1)
120 /* the maximum size for any opcode */
121 #define MAX_OP_SIZE COPYOP_SIZE(0xffffffff, 0xffffffff)
123 void *diff_delta(void *from_buf
, unsigned long from_size
,
124 void *to_buf
, unsigned long to_size
,
125 unsigned long *delta_size
,
126 unsigned long max_size
,
129 unsigned int i
, outpos
, outsize
, inscnt
, hash_shift
;
130 const unsigned char *ref_data
, *ref_top
, *data
, *top
;
132 struct index
*entry
, **hash
;
134 if (!from_size
|| !to_size
)
136 if (from_index
&& *from_index
) {
139 hash
= delta_index(from_buf
, from_size
, to_size
);
145 hash_shift
= (unsigned int)(*hash
++);
149 if (max_size
&& outsize
>= max_size
)
150 outsize
= max_size
+ MAX_OP_SIZE
+ 1;
151 out
= malloc(outsize
);
159 ref_top
= from_buf
+ from_size
;
161 top
= to_buf
+ to_size
;
163 /* store reference buffer size */
164 out
[outpos
++] = from_size
;
167 out
[outpos
- 1] |= 0x80;
168 out
[outpos
++] = from_size
;
172 /* store target buffer size */
173 out
[outpos
++] = to_size
;
176 out
[outpos
- 1] |= 0x80;
177 out
[outpos
++] = to_size
;
184 unsigned int moff
= 0, msize
= 0;
185 if (data
+ 3 <= top
) {
186 i
= data
[0] ^ ((data
[1] ^ (data
[2] << hash_shift
)) << hash_shift
);
187 for (entry
= hash
[i
]; entry
; entry
= entry
->next
) {
188 const unsigned char *ref
= entry
->ptr
;
189 const unsigned char *src
= data
;
190 unsigned int ref_size
= ref_top
- ref
;
191 if (ref_size
> top
- src
)
192 ref_size
= top
- src
;
193 if (ref_size
> 0x10000)
195 if (ref_size
<= msize
)
199 while (ref_size
-- && *++src
== *++ref
);
200 if (msize
< ref
- entry
->ptr
) {
201 /* this is our best match so far */
202 msize
= ref
- entry
->ptr
;
203 moff
= entry
->ptr
- ref_data
;
208 if (!msize
|| msize
< COPYOP_SIZE(moff
, msize
)) {
211 out
[outpos
++] = *data
++;
213 if (inscnt
== 0x7f) {
214 out
[outpos
- inscnt
- 1] = inscnt
;
221 out
[outpos
- inscnt
- 1] = inscnt
;
229 if (moff
& 0xff) { out
[outpos
++] = moff
; i
|= 0x01; }
231 if (moff
& 0xff) { out
[outpos
++] = moff
; i
|= 0x02; }
233 if (moff
& 0xff) { out
[outpos
++] = moff
; i
|= 0x04; }
235 if (moff
& 0xff) { out
[outpos
++] = moff
; i
|= 0x08; }
237 if (msize
& 0xff) { out
[outpos
++] = msize
; i
|= 0x10; }
239 if (msize
& 0xff) { out
[outpos
++] = msize
; i
|= 0x20; }
244 if (outpos
>= outsize
- MAX_OP_SIZE
) {
246 outsize
= outsize
* 3 / 2;
247 if (max_size
&& outsize
>= max_size
)
248 outsize
= max_size
+ MAX_OP_SIZE
+ 1;
249 if (max_size
&& outpos
> max_size
)
252 out
= realloc(out
, outsize
);
263 out
[outpos
- inscnt
- 1] = inscnt
;
267 *delta_size
= outpos
;