annotate lib/bunzip.c @ 387:67b0ded3c56c

Add dirname, and typo fixes to wc header.
author Rob Landley <rob@landley.net>
date Fri, 18 Nov 2011 06:21:29 -0600
parents aaafa1ceaa91
children 786841fdb1e0
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
1 /* vi: set sw=4 ts=4: */
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
2 /* micro-bunzip, a small, simple bzip2 decompression implementation.
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
3
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
4 Copyright 2003, 2006 by Rob Landley (rob@landley.net).
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
5
54
60a7fb8ddeb4 Add a quick and dirt bzcat (stdin to stdout only for the moment) to test
Rob Landley <rob@landley.net>
parents: 46
diff changeset
6 Based on a close reading (but not the actual code) of the original bzip2
60a7fb8ddeb4 Add a quick and dirt bzcat (stdin to stdout only for the moment) to test
Rob Landley <rob@landley.net>
parents: 46
diff changeset
7 decompression code by Julian R Seward (jseward@acm.org), which also
60a7fb8ddeb4 Add a quick and dirt bzcat (stdin to stdout only for the moment) to test
Rob Landley <rob@landley.net>
parents: 46
diff changeset
8 acknowledges contributions by Mike Burrows, David Wheeler, Peter Fenwick,
60a7fb8ddeb4 Add a quick and dirt bzcat (stdin to stdout only for the moment) to test
Rob Landley <rob@landley.net>
parents: 46
diff changeset
9 Alistair Moffat, Radford Neal, Ian H. Witten, Robert Sedgewick, and
60a7fb8ddeb4 Add a quick and dirt bzcat (stdin to stdout only for the moment) to test
Rob Landley <rob@landley.net>
parents: 46
diff changeset
10 Jon L. Bentley.
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
11 */
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
12
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
13 #include "toys.h"
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
14
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
15 #define THREADS 1
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
16
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
17 // Constants for huffman coding
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
18 #define MAX_GROUPS 6
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
19 #define GROUP_SIZE 50 /* 64 would have been more efficient */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
20 #define MAX_HUFCODE_BITS 20 /* Longest huffman code allowed */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
21 #define MAX_SYMBOLS 258 /* 256 literals + RUNA + RUNB */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
22 #define SYMBOL_RUNA 0
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
23 #define SYMBOL_RUNB 1
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
24
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
25 // Other housekeeping constants
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
26 #define IOBUF_SIZE 4096
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
27
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
28 // Status return values
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
29 #define RETVAL_LAST_BLOCK (-100)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
30 #define RETVAL_NOT_BZIP_DATA (-1)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
31 #define RETVAL_DATA_ERROR (-2)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
32 #define RETVAL_OBSOLETE_INPUT (-3)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
33
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
34 char *bunzip_errors[]={
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
35 NULL,
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
36 "Not bzip data",
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
37 "Data error",
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
38 "Obsolete (pre 0.9.5) bzip format not supported."
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
39 };
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
40
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
41 // This is what we know about each huffman coding group
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
42 struct group_data {
62
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
43 int limit[MAX_HUFCODE_BITS+1], base[MAX_HUFCODE_BITS], permute[MAX_SYMBOLS];
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
44 char minLen, maxLen;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
45 };
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
46
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
47 // Data for burrows wheeler transform
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
48
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
49 struct bwdata {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
50 unsigned int origPtr;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
51 int byteCount[256];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
52 // State saved when interrupting output
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
53 int writePos, writeRun, writeCount, writeCurrent;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
54 unsigned int dataCRC, headerCRC;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
55 unsigned int *dbuf;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
56 };
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
57
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
58 // Structure holding all the housekeeping data, including IO buffers and
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
59 // memory that persists between calls to bunzip
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
60 struct bunzip_data {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
61
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
62 // Input stream, input buffer, input bit buffer
46
5a6241f16e95 Remove more warnings.
Rob Landley <rob@landley.net>
parents: 40
diff changeset
63 int in_fd, inbufCount, inbufPos;
5a6241f16e95 Remove more warnings.
Rob Landley <rob@landley.net>
parents: 40
diff changeset
64 char *inbuf;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
65 unsigned int inbufBitCount, inbufBits;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
66
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
67 // Output buffer
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
68 char outbuf[IOBUF_SIZE];
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
69 int outbufPos;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
70
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
71 unsigned int totalCRC;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
72
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
73 // First pass decompression data (Huffman and MTF decoding)
46
5a6241f16e95 Remove more warnings.
Rob Landley <rob@landley.net>
parents: 40
diff changeset
74 char selectors[32768]; // nSelectors=15 bits
5a6241f16e95 Remove more warnings.
Rob Landley <rob@landley.net>
parents: 40
diff changeset
75 struct group_data groups[MAX_GROUPS]; // huffman coding tables
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
76 int symTotal, groupCount, nSelectors;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
77 unsigned char symToByte[256], mtfSymbol[256];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
78
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
79 // The CRC values stored in the block header and calculated from the data
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
80 unsigned int crc32Table[256];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
81
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
82 // Second pass decompression data (burrows-wheeler transform)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
83 unsigned int dbufSize;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
84 struct bwdata bwdata[THREADS];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
85 };
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
86
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
87 // Return the next nnn bits of input. All reads from the compressed input
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
88 // are done through this function. All reads are big endian.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
89 static unsigned int get_bits(struct bunzip_data *bd, char bits_wanted)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
90 {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
91 unsigned int bits = 0;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
92
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
93 // If we need to get more data from the byte buffer, do so. (Loop getting
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
94 // one byte at a time to enforce endianness and avoid unaligned access.)
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
95 while (bd->inbufBitCount < bits_wanted) {
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
96
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
97 // If we need to read more data from file into byte buffer, do so
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
98 if (bd->inbufPos == bd->inbufCount) {
40
d046f8d1401e Trawling through the archives: a broken test, an unnecessary memset, and a
Rob Landley <rob@landley.net>
parents: 39
diff changeset
99 if (0 >= (bd->inbufCount = read(bd->in_fd, bd->inbuf, IOBUF_SIZE)))
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
100 error_exit("Unexpected input EOF");
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
101 bd->inbufPos = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
102 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
103
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
104 // Avoid 32-bit overflow (dump bit buffer to top of output)
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
105 if (bd->inbufBitCount>=24) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
106 bits = bd->inbufBits&((1<<bd->inbufBitCount)-1);
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
107 bits_wanted -= bd->inbufBitCount;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
108 bits <<= bits_wanted;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
109 bd->inbufBitCount = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
110 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
111
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
112 // Grab next 8 bits of input from buffer.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
113 bd->inbufBits = (bd->inbufBits<<8) | bd->inbuf[bd->inbufPos++];
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
114 bd->inbufBitCount += 8;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
115 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
116
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
117 // Calculate result
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
118 bd->inbufBitCount -= bits_wanted;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
119 bits |= (bd->inbufBits>>bd->inbufBitCount) & ((1<<bits_wanted)-1);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
120
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
121 return bits;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
122 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
123
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
124 /* Read block header at start of a new compressed data block. Consists of:
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
125 *
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
126 * 48 bits : Block signature, either pi (data block) or e (EOF block).
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
127 * 32 bits : bw->headerCRC
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
128 * 1 bit : obsolete feature flag.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
129 * 24 bits : origPtr (Burrows-wheeler unwind index, only 20 bits ever used)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
130 * 16 bits : Mapping table index.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
131 *[16 bits]: symToByte[symTotal] (Mapping table. For each bit set in mapping
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
132 * table index above, read another 16 bits of mapping table data.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
133 * If correspondig bit is unset, all bits in that mapping table
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
134 * section are 0.)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
135 * 3 bits : groupCount (how many huffman tables used to encode, anywhere
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
136 * from 2 to MAX_GROUPS)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
137 * variable: hufGroup[groupCount] (MTF encoded huffman table data.)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
138 */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
139
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
140 static int read_block_header(struct bunzip_data *bd, struct bwdata *bw)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
141 {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
142 struct group_data *hufGroup;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
143 int hh, ii, jj, kk, symCount, *base, *limit;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
144 unsigned char uc;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
145
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
146 // Read in header signature and CRC (which is stored big endian)
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
147 ii = get_bits(bd, 24);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
148 jj = get_bits(bd, 24);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
149 bw->headerCRC = get_bits(bd,32);
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
150
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
151 // Is this the EOF block with CRC for whole file? (Constant is "e")
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
152 if (ii==0x177245 && jj==0x385090) return RETVAL_LAST_BLOCK;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
153
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
154 // Is this a valid data block? (Constant is "pi".)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
155 if (ii!=0x314159 || jj!=0x265359) return RETVAL_NOT_BZIP_DATA;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
156
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
157 // We can add support for blockRandomised if anybody complains.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
158 if (get_bits(bd,1)) return RETVAL_OBSOLETE_INPUT;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
159 if ((bw->origPtr = get_bits(bd,24)) > bd->dbufSize)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
160 return RETVAL_DATA_ERROR;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
161
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
162 // mapping table: if some byte values are never used (encoding things
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
163 // like ascii text), the compression code removes the gaps to have fewer
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
164 // symbols to deal with, and writes a sparse bitfield indicating which
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
165 // values were present. We make a translation table to convert the symbols
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
166 // back to the corresponding bytes.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
167 hh = get_bits(bd, 16);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
168 bd->symTotal = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
169 for (ii=0; ii<16; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
170 if (hh & (1 << (15 - ii))) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
171 kk = get_bits(bd, 16);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
172 for (jj=0; jj<16; jj++)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
173 if (kk & (1 << (15 - jj)))
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
174 bd->symToByte[bd->symTotal++] = (16 * ii) + jj;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
175 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
176 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
177
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
178 // How many different huffman coding groups does this block use?
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
179 bd->groupCount = get_bits(bd,3);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
180 if (bd->groupCount<2 || bd->groupCount>MAX_GROUPS) return RETVAL_DATA_ERROR;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
181
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
182 // nSelectors: Every GROUP_SIZE many symbols we switch huffman coding
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
183 // tables. Each group has a selector, which is an index into the huffman
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
184 // coding table arrays.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
185 //
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
186 // Read in the group selector array, which is stored as MTF encoded
56
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
187 // bit runs. (MTF = Move To Front. Every time a symbol occurs it's moved
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
188 // to the front of the table, so it has a shorter encoding next time.)
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
189 if (!(bd->nSelectors = get_bits(bd, 15))) return RETVAL_DATA_ERROR;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
190 for (ii=0; ii<bd->groupCount; ii++) bd->mtfSymbol[ii] = ii;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
191 for (ii=0; ii<bd->nSelectors; ii++) {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
192
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
193 // Get next value
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
194 for(jj=0;get_bits(bd,1);jj++)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
195 if (jj>=bd->groupCount) return RETVAL_DATA_ERROR;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
196
56
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
197 // Decode MTF to get the next selector, and move it to the front.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
198 uc = bd->mtfSymbol[jj];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
199 memmove(bd->mtfSymbol+1, bd->mtfSymbol, jj);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
200 bd->mtfSymbol[0] = bd->selectors[ii] = uc;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
201 }
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
202
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
203 // Read the huffman coding tables for each group, which code for symTotal
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
204 // literal symbols, plus two run symbols (RUNA, RUNB)
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
205 symCount = bd->symTotal+2;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
206 for (jj=0; jj<bd->groupCount; jj++) {
305
b094b81830b0 A pathological case of huffman coding that uses 8 bits to code each of 256
Rob Landley <rob@landley.net>
parents: 233
diff changeset
207 unsigned char length[MAX_SYMBOLS];
b094b81830b0 A pathological case of huffman coding that uses 8 bits to code each of 256
Rob Landley <rob@landley.net>
parents: 233
diff changeset
208 unsigned temp[MAX_HUFCODE_BITS+1];
b094b81830b0 A pathological case of huffman coding that uses 8 bits to code each of 256
Rob Landley <rob@landley.net>
parents: 233
diff changeset
209 int minLen, maxLen, pp;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
210
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
211 // Read lengths
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
212 hh = get_bits(bd, 5);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
213 for (ii = 0; ii < symCount; ii++) {
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
214 for(;;) {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
215 // !hh || hh > MAX_HUFCODE_BITS in one test.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
216 if (MAX_HUFCODE_BITS-1 < (unsigned)hh-1)
57
034de958ee65 Fix off by one error in a test.
Rob Landley <rob@landley.net>
parents: 56
diff changeset
217 return RETVAL_DATA_ERROR;
67
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
218 // Grab 2 bits instead of 1 (slightly smaller/faster). Stop if
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
219 // first bit is 0, otherwise second bit says whether to
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
220 // increment or decrement.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
221 kk = get_bits(bd, 2);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
222 if (kk & 2) hh += 1 - ((kk&1)<<1);
67
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
223 else {
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
224 bd->inbufBitCount++;
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
225 break;
b6646155e924 Another suggestion from Manuel: Grab 2 bits instead of 1 inside a loop. Saves
Rob Landley <rob@landley.net>
parents: 64
diff changeset
226 }
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
227 }
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
228 length[ii] = hh;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
229 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
230
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
231 // Find largest and smallest lengths in this group
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
232 minLen = maxLen = length[0];
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
233 for (ii = 1; ii < symCount; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
234 if(length[ii] > maxLen) maxLen = length[ii];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
235 else if(length[ii] < minLen) minLen = length[ii];
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
236 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
237
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
238 /* Calculate permute[], base[], and limit[] tables from length[].
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
239 *
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
240 * permute[] is the lookup table for converting huffman coded symbols
68
7b89650905c0 Merge two similar loops and expand a comment.
Rob Landley <rob@landley.net>
parents: 67
diff changeset
241 * into decoded symbols. It contains symbol values sorted by length.
156
1e8f4b05cb65 Remove trailing whitespace (thanks to Charlie Shepherd), and a couple comment
Rob Landley <rob@landley.net>
parents: 94
diff changeset
242 *
68
7b89650905c0 Merge two similar loops and expand a comment.
Rob Landley <rob@landley.net>
parents: 67
diff changeset
243 * base[] is the amount to subtract from the value of a huffman symbol
7b89650905c0 Merge two similar loops and expand a comment.
Rob Landley <rob@landley.net>
parents: 67
diff changeset
244 * of a given length when using permute[].
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
245 *
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
246 * limit[] indicates the largest numerical value a symbol with a given
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
247 * number of bits can have. It lets us know when to stop reading.
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
248 *
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
249 * To use these, keep reading bits until value <= limit[bitcount] or
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
250 * you've read over 20 bits (error). Then the decoded symbol
56
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
251 * equals permute[hufcode_value - base[hufcode_bitcount]].
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
252 */
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
253 hufGroup = bd->groups+jj;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
254 hufGroup->minLen = minLen;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
255 hufGroup->maxLen = maxLen;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
256
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
257 // Note that minLen can't be smaller than 1, so we adjust the base
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
258 // and limit array pointers so we're not always wasting the first
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
259 // entry. We do this again when using them (during symbol decoding).
56
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
260 base = hufGroup->base-1;
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
261 limit = hufGroup->limit-1;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
262
233
d4176f3f3835 Zap toys/Config.in and instead create generated/Config.in from contents of
Rob Landley <rob@landley.net>
parents: 215
diff changeset
263 // zero temp[] and limit[], and calculate permute[]
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
264 pp = 0;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
265 for (ii = minLen; ii <= maxLen; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
266 temp[ii] = limit[ii] = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
267 for (hh = 0; hh < symCount; hh++)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
268 if (length[hh] == ii)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
269 hufGroup->permute[pp++] = hh;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
270 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
271
68
7b89650905c0 Merge two similar loops and expand a comment.
Rob Landley <rob@landley.net>
parents: 67
diff changeset
272 // Count symbols coded for at each bit length
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
273 for (ii = 0; ii < symCount; ii++) temp[length[ii]]++;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
274
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
275 /* Calculate limit[] (the largest symbol-coding value at each bit
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
276 * length, which is (previous limit<<1)+symbols at this level), and
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
277 * base[] (number of symbols to ignore at each bit length, which is
68
7b89650905c0 Merge two similar loops and expand a comment.
Rob Landley <rob@landley.net>
parents: 67
diff changeset
278 * limit minus the cumulative count of symbols coded for already). */
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
279 pp = hh = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
280 for (ii = minLen; ii < maxLen; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
281 pp += temp[ii];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
282 limit[ii] = pp-1;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
283 pp <<= 1;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
284 base[ii+1] = pp-(hh+=temp[ii]);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
285 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
286 limit[maxLen] = pp+temp[maxLen]-1;
62
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
287 limit[maxLen+1] = INT_MAX;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
288 base[minLen] = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
289 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
290
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
291 return 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
292 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
293
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
294 /* First pass, read block's symbols into dbuf[dbufCount].
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
295 *
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
296 * This undoes three types of compression: huffman coding, run length encoding,
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
297 * and move to front encoding. We have to undo all those to know when we've
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
298 * read enough input.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
299 */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
300
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
301 static int read_huffman_data(struct bunzip_data *bd, struct bwdata *bw)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
302 {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
303 struct group_data *hufGroup;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
304 int hh, ii, jj, kk, runPos, dbufCount, symCount, selector, nextSym,
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
305 *byteCount, *base, *limit;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
306 unsigned int *dbuf = bw->dbuf;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
307 unsigned char uc;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
308
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
309 // We've finished reading and digesting the block header. Now read this
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
310 // block's huffman coded symbols from the file and undo the huffman coding
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
311 // and run length encoding, saving the result into dbuf[dbufCount++] = uc
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
312
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
313 // Initialize symbol occurrence counters and symbol mtf table
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
314 byteCount = bw->byteCount;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
315 for(ii=0; ii<256; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
316 byteCount[ii] = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
317 bd->mtfSymbol[ii] = ii;
69
530168fc253f Merge a memset with an existing loop, tweak comments.
Rob Landley <rob@landley.net>
parents: 68
diff changeset
318 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
319
56
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
320 // Loop through compressed symbols. This is the first "tight inner loop"
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
321 // that needs to be micro-optimized for speed. (This one fills out dbuf[]
7794562e2e64 Comment and whitespace changes.
Rob Landley <rob@landley.net>
parents: 55
diff changeset
322 // linearly, staying in cache more, so isn't as limited by DRAM access.)
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
323 runPos = dbufCount = symCount = selector = 0;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
324 // Some unnecessary initializations to shut gcc up.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
325 base = limit = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
326 hufGroup = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
327 hh = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
328
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
329 for (;;) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
330
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
331 // Have we reached the end of this huffman group?
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
332 if (!(symCount--)) {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
333 // Determine which huffman coding group to use.
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
334 symCount = GROUP_SIZE-1;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
335 if (selector >= bd->nSelectors) return RETVAL_DATA_ERROR;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
336 hufGroup = bd->groups + bd->selectors[selector++];
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
337 base = hufGroup->base-1;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
338 limit = hufGroup->limit-1;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
339 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
340
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
341 // Read next huffman-coded symbol (into jj).
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
342 ii = hufGroup->minLen;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
343 jj = get_bits(bd, ii);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
344 while (jj > limit[ii]) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
345 // if (ii > hufGroup->maxLen) return RETVAL_DATA_ERROR;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
346 ii++;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
347
62
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
348 // Unroll get_bits() to avoid a function call when the data's in
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
349 // the buffer already.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
350 kk = bd->inbufBitCount
62
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
351 ? (bd->inbufBits >> --(bd->inbufBitCount)) & 1
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
352 : get_bits(bd, 1);
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
353 jj = (jj << 1) | kk;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
354 }
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
355 // Huffman decode jj into nextSym (with bounds checking)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
356 jj-=base[ii];
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
357
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
358 if (ii > hufGroup->maxLen || (unsigned)jj >= MAX_SYMBOLS)
62
f41f997c1e73 More optimizations originally suggested by Manuel Nova: Use a sentinel value
Rob Landley <rob@landley.net>
parents: 61
diff changeset
359 return RETVAL_DATA_ERROR;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
360 nextSym = hufGroup->permute[jj];
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
361
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
362 // If this is a repeated run, loop collecting data
58
b060142aee01 Minor optimization the compiler already does for us (according to bloatcheck),
Rob Landley <rob@landley.net>
parents: 57
diff changeset
363 if ((unsigned)nextSym <= SYMBOL_RUNB) {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
364
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
365 // If this is the start of a new run, zero out counter
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
366 if(!runPos) {
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
367 runPos = 1;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
368 hh = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
369 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
370
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
371 /* Neat trick that saves 1 symbol: instead of or-ing 0 or 1 at
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
372 each bit position, add 1 or 2 instead. For example,
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
373 1011 is 1<<0 + 1<<1 + 2<<2. 1010 is 2<<0 + 2<<1 + 1<<2.
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
374 You can make any bit pattern that way using 1 less symbol than
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
375 the basic or 0/1 method (except all bits 0, which would use no
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
376 symbols, but a run of length 0 doesn't mean anything in this
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
377 context). Thus space is saved. */
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
378 hh += (runPos << nextSym); // +runPos if RUNA; +2*runPos if RUNB
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
379 runPos <<= 1;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
380 continue;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
381 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
382
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
383 /* When we hit the first non-run symbol after a run, we now know
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
384 how many times to repeat the last literal, so append that many
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
385 copies to our buffer of decoded symbols (dbuf) now. (The last
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
386 literal used is the one at the head of the mtfSymbol array.) */
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
387 if (runPos) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
388 runPos = 0;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
389 if (dbufCount+hh >= bd->dbufSize) return RETVAL_DATA_ERROR;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
390
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
391 uc = bd->symToByte[bd->mtfSymbol[0]];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
392 byteCount[uc] += hh;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
393 while (hh--) dbuf[dbufCount++] = uc;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
394 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
395
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
396 // Is this the terminating symbol?
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
397 if (nextSym>bd->symTotal) break;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
398
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
399 /* At this point, the symbol we just decoded indicates a new literal
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
400 character. Subtract one to get the position in the MTF array
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
401 at which this literal is currently to be found. (Note that the
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
402 result can't be -1 or 0, because 0 and 1 are RUNA and RUNB.
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
403 Another instance of the first symbol in the mtf array, position 0,
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
404 would have been handled as part of a run.) */
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
405 if (dbufCount>=bd->dbufSize) return RETVAL_DATA_ERROR;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
406 ii = nextSym - 1;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
407 uc = bd->mtfSymbol[ii];
61
d6ece20e13ce Minor cleanups.
Rob Landley <rob@landley.net>
parents: 60
diff changeset
408 // On my laptop, unrolling this memmove() into a loop shaves 3.5% off
d6ece20e13ce Minor cleanups.
Rob Landley <rob@landley.net>
parents: 60
diff changeset
409 // the total running time.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
410 while(ii--) bd->mtfSymbol[ii+1] = bd->mtfSymbol[ii];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
411 bd->mtfSymbol[0] = uc;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
412 uc = bd->symToByte[uc];
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
413
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
414 // We have our literal byte. Save it into dbuf.
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
415 byteCount[uc]++;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
416 dbuf[dbufCount++] = (unsigned int)uc;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
417 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
418
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
419 // Now we know what dbufCount is, do a better sanity check on origPtr.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
420 if (bw->origPtr >= (bw->writeCount = dbufCount)) return RETVAL_DATA_ERROR;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
421
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
422 return 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
423 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
424
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
425 // Flush output buffer to disk
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
426 void flush_bunzip_outbuf(struct bunzip_data *bd, int out_fd)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
427 {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
428 if (bd->outbufPos) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
429 if (write(out_fd, bd->outbuf, bd->outbufPos) != bd->outbufPos)
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
430 error_exit("Unexpected output EOF");
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
431 bd->outbufPos = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
432 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
433 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
434
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
435 void burrows_wheeler_prep(struct bunzip_data *bd, struct bwdata *bw)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
436 {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
437 int ii, jj;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
438 unsigned int *dbuf = bw->dbuf;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
439 int *byteCount = bw->byteCount;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
440
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
441 // Technically this part is preparation for the burrows-wheeler
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
442 // transform, but it's quick and convenient to do here.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
443
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
444 // Turn byteCount into cumulative occurrence counts of 0 to n-1.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
445 jj = 0;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
446 for (ii=0; ii<256; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
447 int kk = jj + byteCount[ii];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
448 byteCount[ii] = jj;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
449 jj = kk;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
450 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
451
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
452 // Use occurrence counts to quickly figure out what order dbuf would be in
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
453 // if we sorted it.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
454 for (ii=0; ii < bw->writeCount; ii++) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
455 unsigned char uc = dbuf[ii];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
456 dbuf[byteCount[uc]] |= (ii << 8);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
457 byteCount[uc]++;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
458 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
459
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
460 // blockRandomised support would go here.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
461
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
462 // Using ii as position, jj as previous character, hh as current character,
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
463 // and uc as run count.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
464 bw->dataCRC = 0xffffffffL;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
465
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
466 /* Decode first byte by hand to initialize "previous" byte. Note that it
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
467 doesn't get output, and if the first three characters are identical
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
468 it doesn't qualify as a run (hence uc=255, which will either wrap
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
469 to 1 or get reset). */
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
470 if (bw->writeCount) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
471 bw->writePos = dbuf[bw->origPtr];
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
472 bw->writeCurrent = (unsigned char)bw->writePos;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
473 bw->writePos >>= 8;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
474 bw->writeRun = -1;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
475 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
476 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
477
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
478 // Decompress a block of text to intermediate buffer
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
479 int read_bunzip_data(struct bunzip_data *bd)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
480 {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
481 int rc = read_block_header(bd, bd->bwdata);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
482 if (!rc) rc=read_huffman_data(bd, bd->bwdata);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
483
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
484 // First thing that can be done by a background thread.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
485 burrows_wheeler_prep(bd, bd->bwdata);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
486
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
487 return rc;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
488 }
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
489
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
490 // Undo burrows-wheeler transform on intermediate buffer to produce output.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
491 // If !len, write up to len bytes of data to buf. Otherwise write to out_fd.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
492 // Returns len ? bytes written : 0. Notice all errors are negative #'s.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
493 //
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
494 // Burrows-wheeler transform is described at:
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
495 // http://dogma.net/markn/articles/bwt/bwt.htm
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
496 // http://marknelson.us/1996/09/01/bwt/
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
497
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
498 int write_bunzip_data(struct bunzip_data *bd, struct bwdata *bw, int out_fd, char *outbuf, int len)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
499 {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
500 unsigned int *dbuf = bw->dbuf;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
501 int count, pos, current, run, copies, outbyte, previous, gotcount = 0;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
502
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
503 for (;;) {
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
504
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
505 // If last read was short due to end of file, return last block now
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
506 if (bw->writeCount < 0) return bw->writeCount;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
507
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
508 // If we need to refill dbuf, do it.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
509 if (!bw->writeCount) {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
510 int i = read_bunzip_data(bd);
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
511 if (i) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
512 if (i == RETVAL_LAST_BLOCK) {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
513 bw->writeCount = i;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
514 return gotcount;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
515 } else return i;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
516 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
517 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
518
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
519 // loop generating output
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
520 count = bw->writeCount;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
521 pos = bw->writePos;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
522 current = bw->writeCurrent;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
523 run = bw->writeRun;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
524 while (count) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
525
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
526 // If somebody (like tar) wants a certain number of bytes of
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
527 // data from memory instead of written to a file, humor them.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
528 if (len && bd->outbufPos>=len) goto dataus_interruptus;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
529 count--;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
530
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
531 // Follow sequence vector to undo Burrows-Wheeler transform.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
532 previous = current;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
533 pos = dbuf[pos];
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
534 current = pos&0xff;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
535 pos >>= 8;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
536
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
537 // Whenever we see 3 consecutive copies of the same byte,
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
538 // the 4th is a repeat count
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
539 if (run++ == 3) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
540 copies = current;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
541 outbyte = previous;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
542 current = -1;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
543 } else {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
544 copies = 1;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
545 outbyte = current;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
546 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
547
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
548 // Output bytes to buffer, flushing to file if necessary
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
549 while (copies--) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
550 if (bd->outbufPos == IOBUF_SIZE) flush_bunzip_outbuf(bd,out_fd);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
551 bd->outbuf[bd->outbufPos++] = outbyte;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
552 bw->dataCRC = (bw->dataCRC << 8)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
553 ^ bd->crc32Table[(bw->dataCRC >> 24) ^ outbyte];
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
554 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
555 if (current!=previous) run=0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
556 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
557
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
558 // decompression of this block completed successfully
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
559 bw->dataCRC = ~(bw->dataCRC);
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
560 bd->totalCRC = ((bd->totalCRC << 1) | (bd->totalCRC >> 31))
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
561 ^ bw->dataCRC;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
562
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
563 // if this block had a crc error, force file level crc error.
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
564 if (bw->dataCRC != bw->headerCRC) {
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
565 bd->totalCRC = bw->headerCRC+1;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
566
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
567 return RETVAL_LAST_BLOCK;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
568 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
569 dataus_interruptus:
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
570 bw->writeCount = count;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
571 if (len) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
572 gotcount += bd->outbufPos;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
573 memcpy(outbuf, bd->outbuf, len);
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
574
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
575 // If we got enough data, checkpoint loop state and return
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
576 if ((len-=bd->outbufPos)<1) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
577 bd->outbufPos -= len;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
578 if (bd->outbufPos)
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
579 memmove(bd->outbuf, bd->outbuf+len, bd->outbufPos);
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
580 bw->writePos = pos;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
581 bw->writeCurrent = current;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
582 bw->writeRun = run;
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
583
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
584 return gotcount;
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
585 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
586 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
587 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
588 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
589
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
590 // Allocate the structure, read file header. If !len, src_fd contains
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
591 // filehandle to read from. Else inbuf contains data.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
592 int start_bunzip(struct bunzip_data **bdp, int src_fd, char *inbuf, int len)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
593 {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
594 struct bunzip_data *bd;
334
83c461db9df7 Check in crc_init needed by cksum. (Oops.)
Rob Landley <rob@landley.net>
parents: 305
diff changeset
595 unsigned int i;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
596
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
597 // Figure out how much data to allocate.
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
598 i = sizeof(struct bunzip_data);
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
599 if (!len) i += IOBUF_SIZE;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
600
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
601 // Allocate bunzip_data. Most fields initialize to zero.
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
602 bd = *bdp = xzalloc(i);
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
603 if (len) {
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
604 bd->inbuf = inbuf;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
605 bd->inbufCount = len;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
606 bd->in_fd = -1;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
607 } else {
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
608 bd->inbuf = (char *)(bd+1);
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
609 bd->in_fd = src_fd;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
610 }
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
611
337
aaafa1ceaa91 Add -N, -I, -L, and -P options to cksum.
Rob Landley <rob@landley.net>
parents: 334
diff changeset
612 crc_init(bd->crc32Table, 0);
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
613
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
614 // Ensure that file starts with "BZh".
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
615 for (i=0;i<3;i++)
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
616 if (get_bits(bd,8)!="BZh"[i]) return RETVAL_NOT_BZIP_DATA;
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
617
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
618 // Next byte ascii '1'-'9', indicates block size in units of 100k of
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
619 // uncompressed data. Allocate intermediate buffer for block.
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
620 i = get_bits(bd, 8);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
621 if (i<'1' || i>'9') return RETVAL_NOT_BZIP_DATA;
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
622 bd->dbufSize = 100000*(i-'0')*THREADS;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
623 for (i=0; i<THREADS; i++)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
624 bd->bwdata[i].dbuf = xmalloc(bd->dbufSize * sizeof(int));
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
625
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
626 return 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
627 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
628
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
629 // Example usage: decompress src_fd to dst_fd. (Stops at end of bzip data,
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
630 // not end of file.)
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
631 void bunzipStream(int src_fd, int dst_fd)
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
632 {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
633 struct bunzip_data *bd;
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
634 int i, j;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
635
39
00e73c4dea66 More whitespace/cowbell. (And change comment style to //.)
Rob Landley <rob@landley.net>
parents: 37
diff changeset
636 if (!(i = start_bunzip(&bd,src_fd,0,0))) {
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
637 i = write_bunzip_data(bd,bd->bwdata,dst_fd,0,0);
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
638 if (i==RETVAL_LAST_BLOCK && bd->bwdata[0].headerCRC==bd->totalCRC)
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
639 i = 0;
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
640 }
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
641 flush_bunzip_outbuf(bd,dst_fd);
215
34e36dffd47a Major refactoring of bunzip.c in preparation for doing a multi-threaded version.
Rob Landley <rob@landley.net>
parents: 156
diff changeset
642 for (j=0; j<THREADS; j++) free(bd->bwdata[j].dbuf);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
643 free(bd);
64
67ee3a0b76e1 In bunzip replace setjmp/longjmp handling with error_exit(), replace string
Rob Landley <rob@landley.net>
parents: 62
diff changeset
644 if (i) error_exit(bunzip_errors[-i]);
37
14d27eab3acf Add my old micro-bunzip library. Needs some cleanup...
Rob Landley <rob@landley.net>
parents:
diff changeset
645 }