Home | History | Annotate | Download | only in puff
      1 /*
      2  * puff.c
      3  * Copyright (C) 2002-2013 Mark Adler
      4  * For conditions of distribution and use, see copyright notice in puff.h
      5  * version 2.3, 21 Jan 2013
      6  *
      7  * puff.c is a simple inflate written to be an unambiguous way to specify the
      8  * deflate format.  It is not written for speed but rather simplicity.  As a
      9  * side benefit, this code might actually be useful when small code is more
     10  * important than speed, such as bootstrap applications.  For typical deflate
     11  * data, zlib's inflate() is about four times as fast as puff().  zlib's
     12  * inflate compiles to around 20K on my machine, whereas puff.c compiles to
     13  * around 4K on my machine (a PowerPC using GNU cc).  If the faster decode()
     14  * function here is used, then puff() is only twice as slow as zlib's
     15  * inflate().
     16  *
     17  * All dynamically allocated memory comes from the stack.  The stack required
     18  * is less than 2K bytes.  This code is compatible with 16-bit int's and
     19  * assumes that long's are at least 32 bits.  puff.c uses the short data type,
     20  * assumed to be 16 bits, for arrays in order to to conserve memory.  The code
     21  * works whether integers are stored big endian or little endian.
     22  *
     23  * In the comments below are "Format notes" that describe the inflate process
     24  * and document some of the less obvious aspects of the format.  This source
     25  * code is meant to supplement RFC 1951, which formally describes the deflate
     26  * format:
     27  *
     28  *    http://www.zlib.org/rfc-deflate.html
     29  */
     30 
     31 /*
     32  * Change history:
     33  *
     34  * 1.0  10 Feb 2002     - First version
     35  * 1.1  17 Feb 2002     - Clarifications of some comments and notes
     36  *                      - Update puff() dest and source pointers on negative
     37  *                        errors to facilitate debugging deflators
     38  *                      - Remove longest from struct huffman -- not needed
     39  *                      - Simplify offs[] index in construct()
     40  *                      - Add input size and checking, using longjmp() to
     41  *                        maintain easy readability
     42  *                      - Use short data type for large arrays
     43  *                      - Use pointers instead of long to specify source and
     44  *                        destination sizes to avoid arbitrary 4 GB limits
     45  * 1.2  17 Mar 2002     - Add faster version of decode(), doubles speed (!),
     46  *                        but leave simple version for readabilty
     47  *                      - Make sure invalid distances detected if pointers
     48  *                        are 16 bits
     49  *                      - Fix fixed codes table error
     50  *                      - Provide a scanning mode for determining size of
     51  *                        uncompressed data
     52  * 1.3  20 Mar 2002     - Go back to lengths for puff() parameters [Gailly]
     53  *                      - Add a puff.h file for the interface
     54  *                      - Add braces in puff() for else do [Gailly]
     55  *                      - Use indexes instead of pointers for readability
     56  * 1.4  31 Mar 2002     - Simplify construct() code set check
     57  *                      - Fix some comments
     58  *                      - Add FIXLCODES #define
     59  * 1.5   6 Apr 2002     - Minor comment fixes
     60  * 1.6   7 Aug 2002     - Minor format changes
     61  * 1.7   3 Mar 2003     - Added test code for distribution
     62  *                      - Added zlib-like license
     63  * 1.8   9 Jan 2004     - Added some comments on no distance codes case
     64  * 1.9  21 Feb 2008     - Fix bug on 16-bit integer architectures [Pohland]
     65  *                      - Catch missing end-of-block symbol error
     66  * 2.0  25 Jul 2008     - Add #define to permit distance too far back
     67  *                      - Add option in TEST code for puff to write the data
     68  *                      - Add option in TEST code to skip input bytes
     69  *                      - Allow TEST code to read from piped stdin
     70  * 2.1   4 Apr 2010     - Avoid variable initialization for happier compilers
     71  *                      - Avoid unsigned comparisons for even happier compilers
     72  * 2.2  25 Apr 2010     - Fix bug in variable initializations [Oberhumer]
     73  *                      - Add const where appropriate [Oberhumer]
     74  *                      - Split if's and ?'s for coverage testing
     75  *                      - Break out test code to separate file
     76  *                      - Move NIL to puff.h
     77  *                      - Allow incomplete code only if single code length is 1
     78  *                      - Add full code coverage test to Makefile
     79  * 2.3  21 Jan 2013     - Check for invalid code length codes in dynamic blocks
     80  */
     81 
     82 #include <setjmp.h>             /* for setjmp(), longjmp(), and jmp_buf */
     83 #include "puff.h"               /* prototype for puff() */
     84 
     85 #define local static            /* for local function definitions */
     86 
     87 /*
     88  * Maximums for allocations and loops.  It is not useful to change these --
     89  * they are fixed by the deflate format.
     90  */
     91 #define MAXBITS 15              /* maximum bits in a code */
     92 #define MAXLCODES 286           /* maximum number of literal/length codes */
     93 #define MAXDCODES 30            /* maximum number of distance codes */
     94 #define MAXCODES (MAXLCODES+MAXDCODES)  /* maximum codes lengths to read */
     95 #define FIXLCODES 288           /* number of fixed literal/length codes */
     96 
     97 /* input and output state */
     98 struct state {
     99     /* output state */
    100     unsigned char *out;         /* output buffer */
    101     unsigned long outlen;       /* available space at out */
    102     unsigned long outcnt;       /* bytes written to out so far */
    103 
    104     /* input state */
    105     const unsigned char *in;    /* input buffer */
    106     unsigned long inlen;        /* available input at in */
    107     unsigned long incnt;        /* bytes read so far */
    108     int bitbuf;                 /* bit buffer */
    109     int bitcnt;                 /* number of bits in bit buffer */
    110 
    111     /* input limit error return state for bits() and decode() */
    112     jmp_buf env;
    113 };
    114 
    115 /*
    116  * Return need bits from the input stream.  This always leaves less than
    117  * eight bits in the buffer.  bits() works properly for need == 0.
    118  *
    119  * Format notes:
    120  *
    121  * - Bits are stored in bytes from the least significant bit to the most
    122  *   significant bit.  Therefore bits are dropped from the bottom of the bit
    123  *   buffer, using shift right, and new bytes are appended to the top of the
    124  *   bit buffer, using shift left.
    125  */
    126 local int bits(struct state *s, int need)
    127 {
    128     long val;           /* bit accumulator (can use up to 20 bits) */
    129 
    130     /* load at least need bits into val */
    131     val = s->bitbuf;
    132     while (s->bitcnt < need) {
    133         if (s->incnt == s->inlen)
    134             longjmp(s->env, 1);         /* out of input */
    135         val |= (long)(s->in[s->incnt++]) << s->bitcnt;  /* load eight bits */
    136         s->bitcnt += 8;
    137     }
    138 
    139     /* drop need bits and update buffer, always zero to seven bits left */
    140     s->bitbuf = (int)(val >> need);
    141     s->bitcnt -= need;
    142 
    143     /* return need bits, zeroing the bits above that */
    144     return (int)(val & ((1L << need) - 1));
    145 }
    146 
    147 /*
    148  * Process a stored block.
    149  *
    150  * Format notes:
    151  *
    152  * - After the two-bit stored block type (00), the stored block length and
    153  *   stored bytes are byte-aligned for fast copying.  Therefore any leftover
    154  *   bits in the byte that has the last bit of the type, as many as seven, are
    155  *   discarded.  The value of the discarded bits are not defined and should not
    156  *   be checked against any expectation.
    157  *
    158  * - The second inverted copy of the stored block length does not have to be
    159  *   checked, but it's probably a good idea to do so anyway.
    160  *
    161  * - A stored block can have zero length.  This is sometimes used to byte-align
    162  *   subsets of the compressed data for random access or partial recovery.
    163  */
    164 local int stored(struct state *s)
    165 {
    166     unsigned len;       /* length of stored block */
    167 
    168     /* discard leftover bits from current byte (assumes s->bitcnt < 8) */
    169     s->bitbuf = 0;
    170     s->bitcnt = 0;
    171 
    172     /* get length and check against its one's complement */
    173     if (s->incnt + 4 > s->inlen)
    174         return 2;                               /* not enough input */
    175     len = s->in[s->incnt++];
    176     len |= s->in[s->incnt++] << 8;
    177     if (s->in[s->incnt++] != (~len & 0xff) ||
    178         s->in[s->incnt++] != ((~len >> 8) & 0xff))
    179         return -2;                              /* didn't match complement! */
    180 
    181     /* copy len bytes from in to out */
    182     if (s->incnt + len > s->inlen)
    183         return 2;                               /* not enough input */
    184     if (s->out != NIL) {
    185         if (s->outcnt + len > s->outlen)
    186             return 1;                           /* not enough output space */
    187         while (len--)
    188             s->out[s->outcnt++] = s->in[s->incnt++];
    189     }
    190     else {                                      /* just scanning */
    191         s->outcnt += len;
    192         s->incnt += len;
    193     }
    194 
    195     /* done with a valid stored block */
    196     return 0;
    197 }
    198 
    199 /*
    200  * Huffman code decoding tables.  count[1..MAXBITS] is the number of symbols of
    201  * each length, which for a canonical code are stepped through in order.
    202  * symbol[] are the symbol values in canonical order, where the number of
    203  * entries is the sum of the counts in count[].  The decoding process can be
    204  * seen in the function decode() below.
    205  */
    206 struct huffman {
    207     short *count;       /* number of symbols of each length */
    208     short *symbol;      /* canonically ordered symbols */
    209 };
    210 
    211 /*
    212  * Decode a code from the stream s using huffman table h.  Return the symbol or
    213  * a negative value if there is an error.  If all of the lengths are zero, i.e.
    214  * an empty code, or if the code is incomplete and an invalid code is received,
    215  * then -10 is returned after reading MAXBITS bits.
    216  *
    217  * Format notes:
    218  *
    219  * - The codes as stored in the compressed data are bit-reversed relative to
    220  *   a simple integer ordering of codes of the same lengths.  Hence below the
    221  *   bits are pulled from the compressed data one at a time and used to
    222  *   build the code value reversed from what is in the stream in order to
    223  *   permit simple integer comparisons for decoding.  A table-based decoding
    224  *   scheme (as used in zlib) does not need to do this reversal.
    225  *
    226  * - The first code for the shortest length is all zeros.  Subsequent codes of
    227  *   the same length are simply integer increments of the previous code.  When
    228  *   moving up a length, a zero bit is appended to the code.  For a complete
    229  *   code, the last code of the longest length will be all ones.
    230  *
    231  * - Incomplete codes are handled by this decoder, since they are permitted
    232  *   in the deflate format.  See the format notes for fixed() and dynamic().
    233  */
    234 #ifdef SLOW
    235 local int decode(struct state *s, const struct huffman *h)
    236 {
    237     int len;            /* current number of bits in code */
    238     int code;           /* len bits being decoded */
    239     int first;          /* first code of length len */
    240     int count;          /* number of codes of length len */
    241     int index;          /* index of first code of length len in symbol table */
    242 
    243     code = first = index = 0;
    244     for (len = 1; len <= MAXBITS; len++) {
    245         code |= bits(s, 1);             /* get next bit */
    246         count = h->count[len];
    247         if (code - count < first)       /* if length len, return symbol */
    248             return h->symbol[index + (code - first)];
    249         index += count;                 /* else update for next length */
    250         first += count;
    251         first <<= 1;
    252         code <<= 1;
    253     }
    254     return -10;                         /* ran out of codes */
    255 }
    256 
    257 /*
    258  * A faster version of decode() for real applications of this code.   It's not
    259  * as readable, but it makes puff() twice as fast.  And it only makes the code
    260  * a few percent larger.
    261  */
    262 #else /* !SLOW */
    263 local int decode(struct state *s, const struct huffman *h)
    264 {
    265     int len;            /* current number of bits in code */
    266     int code;           /* len bits being decoded */
    267     int first;          /* first code of length len */
    268     int count;          /* number of codes of length len */
    269     int index;          /* index of first code of length len in symbol table */
    270     int bitbuf;         /* bits from stream */
    271     int left;           /* bits left in next or left to process */
    272     short *next;        /* next number of codes */
    273 
    274     bitbuf = s->bitbuf;
    275     left = s->bitcnt;
    276     code = first = index = 0;
    277     len = 1;
    278     next = h->count + 1;
    279     while (1) {
    280         while (left--) {
    281             code |= bitbuf & 1;
    282             bitbuf >>= 1;
    283             count = *next++;
    284             if (code - count < first) { /* if length len, return symbol */
    285                 s->bitbuf = bitbuf;
    286                 s->bitcnt = (s->bitcnt - len) & 7;
    287                 return h->symbol[index + (code - first)];
    288             }
    289             index += count;             /* else update for next length */
    290             first += count;
    291             first <<= 1;
    292             code <<= 1;
    293             len++;
    294         }
    295         left = (MAXBITS+1) - len;
    296         if (left == 0)
    297             break;
    298         if (s->incnt == s->inlen)
    299             longjmp(s->env, 1);         /* out of input */
    300         bitbuf = s->in[s->incnt++];
    301         if (left > 8)
    302             left = 8;
    303     }
    304     return -10;                         /* ran out of codes */
    305 }
    306 #endif /* SLOW */
    307 
    308 /*
    309  * Given the list of code lengths length[0..n-1] representing a canonical
    310  * Huffman code for n symbols, construct the tables required to decode those
    311  * codes.  Those tables are the number of codes of each length, and the symbols
    312  * sorted by length, retaining their original order within each length.  The
    313  * return value is zero for a complete code set, negative for an over-
    314  * subscribed code set, and positive for an incomplete code set.  The tables
    315  * can be used if the return value is zero or positive, but they cannot be used
    316  * if the return value is negative.  If the return value is zero, it is not
    317  * possible for decode() using that table to return an error--any stream of
    318  * enough bits will resolve to a symbol.  If the return value is positive, then
    319  * it is possible for decode() using that table to return an error for received
    320  * codes past the end of the incomplete lengths.
    321  *
    322  * Not used by decode(), but used for error checking, h->count[0] is the number
    323  * of the n symbols not in the code.  So n - h->count[0] is the number of
    324  * codes.  This is useful for checking for incomplete codes that have more than
    325  * one symbol, which is an error in a dynamic block.
    326  *
    327  * Assumption: for all i in 0..n-1, 0 <= length[i] <= MAXBITS
    328  * This is assured by the construction of the length arrays in dynamic() and
    329  * fixed() and is not verified by construct().
    330  *
    331  * Format notes:
    332  *
    333  * - Permitted and expected examples of incomplete codes are one of the fixed
    334  *   codes and any code with a single symbol which in deflate is coded as one
    335  *   bit instead of zero bits.  See the format notes for fixed() and dynamic().
    336  *
    337  * - Within a given code length, the symbols are kept in ascending order for
    338  *   the code bits definition.
    339  */
    340 local int construct(struct huffman *h, const short *length, int n)
    341 {
    342     int symbol;         /* current symbol when stepping through length[] */
    343     int len;            /* current length when stepping through h->count[] */
    344     int left;           /* number of possible codes left of current length */
    345     short offs[MAXBITS+1];      /* offsets in symbol table for each length */
    346 
    347     /* count number of codes of each length */
    348     for (len = 0; len <= MAXBITS; len++)
    349         h->count[len] = 0;
    350     for (symbol = 0; symbol < n; symbol++)
    351         (h->count[length[symbol]])++;   /* assumes lengths are within bounds */
    352     if (h->count[0] == n)               /* no codes! */
    353         return 0;                       /* complete, but decode() will fail */
    354 
    355     /* check for an over-subscribed or incomplete set of lengths */
    356     left = 1;                           /* one possible code of zero length */
    357     for (len = 1; len <= MAXBITS; len++) {
    358         left <<= 1;                     /* one more bit, double codes left */
    359         left -= h->count[len];          /* deduct count from possible codes */
    360         if (left < 0)
    361             return left;                /* over-subscribed--return negative */
    362     }                                   /* left > 0 means incomplete */
    363 
    364     /* generate offsets into symbol table for each length for sorting */
    365     offs[1] = 0;
    366     for (len = 1; len < MAXBITS; len++)
    367         offs[len + 1] = offs[len] + h->count[len];
    368 
    369     /*
    370      * put symbols in table sorted by length, by symbol order within each
    371      * length
    372      */
    373     for (symbol = 0; symbol < n; symbol++)
    374         if (length[symbol] != 0)
    375             h->symbol[offs[length[symbol]]++] = symbol;
    376 
    377     /* return zero for complete set, positive for incomplete set */
    378     return left;
    379 }
    380 
    381 /*
    382  * Decode literal/length and distance codes until an end-of-block code.
    383  *
    384  * Format notes:
    385  *
    386  * - Compressed data that is after the block type if fixed or after the code
    387  *   description if dynamic is a combination of literals and length/distance
    388  *   pairs terminated by and end-of-block code.  Literals are simply Huffman
    389  *   coded bytes.  A length/distance pair is a coded length followed by a
    390  *   coded distance to represent a string that occurs earlier in the
    391  *   uncompressed data that occurs again at the current location.
    392  *
    393  * - Literals, lengths, and the end-of-block code are combined into a single
    394  *   code of up to 286 symbols.  They are 256 literals (0..255), 29 length
    395  *   symbols (257..285), and the end-of-block symbol (256).
    396  *
    397  * - There are 256 possible lengths (3..258), and so 29 symbols are not enough
    398  *   to represent all of those.  Lengths 3..10 and 258 are in fact represented
    399  *   by just a length symbol.  Lengths 11..257 are represented as a symbol and
    400  *   some number of extra bits that are added as an integer to the base length
    401  *   of the length symbol.  The number of extra bits is determined by the base
    402  *   length symbol.  These are in the static arrays below, lens[] for the base
    403  *   lengths and lext[] for the corresponding number of extra bits.
    404  *
    405  * - The reason that 258 gets its own symbol is that the longest length is used
    406  *   often in highly redundant files.  Note that 258 can also be coded as the
    407  *   base value 227 plus the maximum extra value of 31.  While a good deflate
    408  *   should never do this, it is not an error, and should be decoded properly.
    409  *
    410  * - If a length is decoded, including its extra bits if any, then it is
    411  *   followed a distance code.  There are up to 30 distance symbols.  Again
    412  *   there are many more possible distances (1..32768), so extra bits are added
    413  *   to a base value represented by the symbol.  The distances 1..4 get their
    414  *   own symbol, but the rest require extra bits.  The base distances and
    415  *   corresponding number of extra bits are below in the static arrays dist[]
    416  *   and dext[].
    417  *
    418  * - Literal bytes are simply written to the output.  A length/distance pair is
    419  *   an instruction to copy previously uncompressed bytes to the output.  The
    420  *   copy is from distance bytes back in the output stream, copying for length
    421  *   bytes.
    422  *
    423  * - Distances pointing before the beginning of the output data are not
    424  *   permitted.
    425  *
    426  * - Overlapped copies, where the length is greater than the distance, are
    427  *   allowed and common.  For example, a distance of one and a length of 258
    428  *   simply copies the last byte 258 times.  A distance of four and a length of
    429  *   twelve copies the last four bytes three times.  A simple forward copy
    430  *   ignoring whether the length is greater than the distance or not implements
    431  *   this correctly.  You should not use memcpy() since its behavior is not
    432  *   defined for overlapped arrays.  You should not use memmove() or bcopy()
    433  *   since though their behavior -is- defined for overlapping arrays, it is
    434  *   defined to do the wrong thing in this case.
    435  */
    436 local int codes(struct state *s,
    437                 const struct huffman *lencode,
    438                 const struct huffman *distcode)
    439 {
    440     int symbol;         /* decoded symbol */
    441     int len;            /* length for copy */
    442     unsigned dist;      /* distance for copy */
    443     static const short lens[29] = { /* Size base for length codes 257..285 */
    444         3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
    445         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258};
    446     static const short lext[29] = { /* Extra bits for length codes 257..285 */
    447         0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
    448         3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0};
    449     static const short dists[30] = { /* Offset base for distance codes 0..29 */
    450         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
    451         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
    452         8193, 12289, 16385, 24577};
    453     static const short dext[30] = { /* Extra bits for distance codes 0..29 */
    454         0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
    455         7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
    456         12, 12, 13, 13};
    457 
    458     /* decode literals and length/distance pairs */
    459     do {
    460         symbol = decode(s, lencode);
    461         if (symbol < 0)
    462             return symbol;              /* invalid symbol */
    463         if (symbol < 256) {             /* literal: symbol is the byte */
    464             /* write out the literal */
    465             if (s->out != NIL) {
    466                 if (s->outcnt == s->outlen)
    467                     return 1;
    468                 s->out[s->outcnt] = symbol;
    469             }
    470             s->outcnt++;
    471         }
    472         else if (symbol > 256) {        /* length */
    473             /* get and compute length */
    474             symbol -= 257;
    475             if (symbol >= 29)
    476                 return -10;             /* invalid fixed code */
    477             len = lens[symbol] + bits(s, lext[symbol]);
    478 
    479             /* get and check distance */
    480             symbol = decode(s, distcode);
    481             if (symbol < 0)
    482                 return symbol;          /* invalid symbol */
    483             dist = dists[symbol] + bits(s, dext[symbol]);
    484 #ifndef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
    485             if (dist > s->outcnt)
    486                 return -11;     /* distance too far back */
    487 #endif
    488 
    489             /* copy length bytes from distance bytes back */
    490             if (s->out != NIL) {
    491                 if (s->outcnt + len > s->outlen)
    492                     return 1;
    493                 while (len--) {
    494                     s->out[s->outcnt] =
    495 #ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
    496                         dist > s->outcnt ?
    497                             0 :
    498 #endif
    499                             s->out[s->outcnt - dist];
    500                     s->outcnt++;
    501                 }
    502             }
    503             else
    504                 s->outcnt += len;
    505         }
    506     } while (symbol != 256);            /* end of block symbol */
    507 
    508     /* done with a valid fixed or dynamic block */
    509     return 0;
    510 }
    511 
    512 /*
    513  * Process a fixed codes block.
    514  *
    515  * Format notes:
    516  *
    517  * - This block type can be useful for compressing small amounts of data for
    518  *   which the size of the code descriptions in a dynamic block exceeds the
    519  *   benefit of custom codes for that block.  For fixed codes, no bits are
    520  *   spent on code descriptions.  Instead the code lengths for literal/length
    521  *   codes and distance codes are fixed.  The specific lengths for each symbol
    522  *   can be seen in the "for" loops below.
    523  *
    524  * - The literal/length code is complete, but has two symbols that are invalid
    525  *   and should result in an error if received.  This cannot be implemented
    526  *   simply as an incomplete code since those two symbols are in the "middle"
    527  *   of the code.  They are eight bits long and the longest literal/length\
    528  *   code is nine bits.  Therefore the code must be constructed with those
    529  *   symbols, and the invalid symbols must be detected after decoding.
    530  *
    531  * - The fixed distance codes also have two invalid symbols that should result
    532  *   in an error if received.  Since all of the distance codes are the same
    533  *   length, this can be implemented as an incomplete code.  Then the invalid
    534  *   codes are detected while decoding.
    535  */
    536 local int fixed(struct state *s)
    537 {
    538     static int virgin = 1;
    539     static short lencnt[MAXBITS+1], lensym[FIXLCODES];
    540     static short distcnt[MAXBITS+1], distsym[MAXDCODES];
    541     static struct huffman lencode, distcode;
    542 
    543     /* build fixed huffman tables if first call (may not be thread safe) */
    544     if (virgin) {
    545         int symbol;
    546         short lengths[FIXLCODES];
    547 
    548         /* construct lencode and distcode */
    549         lencode.count = lencnt;
    550         lencode.symbol = lensym;
    551         distcode.count = distcnt;
    552         distcode.symbol = distsym;
    553 
    554         /* literal/length table */
    555         for (symbol = 0; symbol < 144; symbol++)
    556             lengths[symbol] = 8;
    557         for (; symbol < 256; symbol++)
    558             lengths[symbol] = 9;
    559         for (; symbol < 280; symbol++)
    560             lengths[symbol] = 7;
    561         for (; symbol < FIXLCODES; symbol++)
    562             lengths[symbol] = 8;
    563         construct(&lencode, lengths, FIXLCODES);
    564 
    565         /* distance table */
    566         for (symbol = 0; symbol < MAXDCODES; symbol++)
    567             lengths[symbol] = 5;
    568         construct(&distcode, lengths, MAXDCODES);
    569 
    570         /* do this just once */
    571         virgin = 0;
    572     }
    573 
    574     /* decode data until end-of-block code */
    575     return codes(s, &lencode, &distcode);
    576 }
    577 
    578 /*
    579  * Process a dynamic codes block.
    580  *
    581  * Format notes:
    582  *
    583  * - A dynamic block starts with a description of the literal/length and
    584  *   distance codes for that block.  New dynamic blocks allow the compressor to
    585  *   rapidly adapt to changing data with new codes optimized for that data.
    586  *
    587  * - The codes used by the deflate format are "canonical", which means that
    588  *   the actual bits of the codes are generated in an unambiguous way simply
    589  *   from the number of bits in each code.  Therefore the code descriptions
    590  *   are simply a list of code lengths for each symbol.
    591  *
    592  * - The code lengths are stored in order for the symbols, so lengths are
    593  *   provided for each of the literal/length symbols, and for each of the
    594  *   distance symbols.
    595  *
    596  * - If a symbol is not used in the block, this is represented by a zero as
    597  *   as the code length.  This does not mean a zero-length code, but rather
    598  *   that no code should be created for this symbol.  There is no way in the
    599  *   deflate format to represent a zero-length code.
    600  *
    601  * - The maximum number of bits in a code is 15, so the possible lengths for
    602  *   any code are 1..15.
    603  *
    604  * - The fact that a length of zero is not permitted for a code has an
    605  *   interesting consequence.  Normally if only one symbol is used for a given
    606  *   code, then in fact that code could be represented with zero bits.  However
    607  *   in deflate, that code has to be at least one bit.  So for example, if
    608  *   only a single distance base symbol appears in a block, then it will be
    609  *   represented by a single code of length one, in particular one 0 bit.  This
    610  *   is an incomplete code, since if a 1 bit is received, it has no meaning,
    611  *   and should result in an error.  So incomplete distance codes of one symbol
    612  *   should be permitted, and the receipt of invalid codes should be handled.
    613  *
    614  * - It is also possible to have a single literal/length code, but that code
    615  *   must be the end-of-block code, since every dynamic block has one.  This
    616  *   is not the most efficient way to create an empty block (an empty fixed
    617  *   block is fewer bits), but it is allowed by the format.  So incomplete
    618  *   literal/length codes of one symbol should also be permitted.
    619  *
    620  * - If there are only literal codes and no lengths, then there are no distance
    621  *   codes.  This is represented by one distance code with zero bits.
    622  *
    623  * - The list of up to 286 length/literal lengths and up to 30 distance lengths
    624  *   are themselves compressed using Huffman codes and run-length encoding.  In
    625  *   the list of code lengths, a 0 symbol means no code, a 1..15 symbol means
    626  *   that length, and the symbols 16, 17, and 18 are run-length instructions.
    627  *   Each of 16, 17, and 18 are follwed by extra bits to define the length of
    628  *   the run.  16 copies the last length 3 to 6 times.  17 represents 3 to 10
    629  *   zero lengths, and 18 represents 11 to 138 zero lengths.  Unused symbols
    630  *   are common, hence the special coding for zero lengths.
    631  *
    632  * - The symbols for 0..18 are Huffman coded, and so that code must be
    633  *   described first.  This is simply a sequence of up to 19 three-bit values
    634  *   representing no code (0) or the code length for that symbol (1..7).
    635  *
    636  * - A dynamic block starts with three fixed-size counts from which is computed
    637  *   the number of literal/length code lengths, the number of distance code
    638  *   lengths, and the number of code length code lengths (ok, you come up with
    639  *   a better name!) in the code descriptions.  For the literal/length and
    640  *   distance codes, lengths after those provided are considered zero, i.e. no
    641  *   code.  The code length code lengths are received in a permuted order (see
    642  *   the order[] array below) to make a short code length code length list more
    643  *   likely.  As it turns out, very short and very long codes are less likely
    644  *   to be seen in a dynamic code description, hence what may appear initially
    645  *   to be a peculiar ordering.
    646  *
    647  * - Given the number of literal/length code lengths (nlen) and distance code
    648  *   lengths (ndist), then they are treated as one long list of nlen + ndist
    649  *   code lengths.  Therefore run-length coding can and often does cross the
    650  *   boundary between the two sets of lengths.
    651  *
    652  * - So to summarize, the code description at the start of a dynamic block is
    653  *   three counts for the number of code lengths for the literal/length codes,
    654  *   the distance codes, and the code length codes.  This is followed by the
    655  *   code length code lengths, three bits each.  This is used to construct the
    656  *   code length code which is used to read the remainder of the lengths.  Then
    657  *   the literal/length code lengths and distance lengths are read as a single
    658  *   set of lengths using the code length codes.  Codes are constructed from
    659  *   the resulting two sets of lengths, and then finally you can start
    660  *   decoding actual compressed data in the block.
    661  *
    662  * - For reference, a "typical" size for the code description in a dynamic
    663  *   block is around 80 bytes.
    664  */
    665 local int dynamic(struct state *s)
    666 {
    667     int nlen, ndist, ncode;             /* number of lengths in descriptor */
    668     int index;                          /* index of lengths[] */
    669     int err;                            /* construct() return value */
    670     short lengths[MAXCODES];            /* descriptor code lengths */
    671     short lencnt[MAXBITS+1], lensym[MAXLCODES];         /* lencode memory */
    672     short distcnt[MAXBITS+1], distsym[MAXDCODES];       /* distcode memory */
    673     struct huffman lencode, distcode;   /* length and distance codes */
    674     static const short order[19] =      /* permutation of code length codes */
    675         {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
    676 
    677     /* construct lencode and distcode */
    678     lencode.count = lencnt;
    679     lencode.symbol = lensym;
    680     distcode.count = distcnt;
    681     distcode.symbol = distsym;
    682 
    683     /* get number of lengths in each table, check lengths */
    684     nlen = bits(s, 5) + 257;
    685     ndist = bits(s, 5) + 1;
    686     ncode = bits(s, 4) + 4;
    687     if (nlen > MAXLCODES || ndist > MAXDCODES)
    688         return -3;                      /* bad counts */
    689 
    690     /* read code length code lengths (really), missing lengths are zero */
    691     for (index = 0; index < ncode; index++)
    692         lengths[order[index]] = bits(s, 3);
    693     for (; index < 19; index++)
    694         lengths[order[index]] = 0;
    695 
    696     /* build huffman table for code lengths codes (use lencode temporarily) */
    697     err = construct(&lencode, lengths, 19);
    698     if (err != 0)               /* require complete code set here */
    699         return -4;
    700 
    701     /* read length/literal and distance code length tables */
    702     index = 0;
    703     while (index < nlen + ndist) {
    704         int symbol;             /* decoded value */
    705         int len;                /* last length to repeat */
    706 
    707         symbol = decode(s, &lencode);
    708         if (symbol < 0)
    709             return symbol;          /* invalid symbol */
    710         if (symbol < 16)                /* length in 0..15 */
    711             lengths[index++] = symbol;
    712         else {                          /* repeat instruction */
    713             len = 0;                    /* assume repeating zeros */
    714             if (symbol == 16) {         /* repeat last length 3..6 times */
    715                 if (index == 0)
    716                     return -5;          /* no last length! */
    717                 len = lengths[index - 1];       /* last length */
    718                 symbol = 3 + bits(s, 2);
    719             }
    720             else if (symbol == 17)      /* repeat zero 3..10 times */
    721                 symbol = 3 + bits(s, 3);
    722             else                        /* == 18, repeat zero 11..138 times */
    723                 symbol = 11 + bits(s, 7);
    724             if (index + symbol > nlen + ndist)
    725                 return -6;              /* too many lengths! */
    726             while (symbol--)            /* repeat last or zero symbol times */
    727                 lengths[index++] = len;
    728         }
    729     }
    730 
    731     /* check for end-of-block code -- there better be one! */
    732     if (lengths[256] == 0)
    733         return -9;
    734 
    735     /* build huffman table for literal/length codes */
    736     err = construct(&lencode, lengths, nlen);
    737     if (err && (err < 0 || nlen != lencode.count[0] + lencode.count[1]))
    738         return -7;      /* incomplete code ok only for single length 1 code */
    739 
    740     /* build huffman table for distance codes */
    741     err = construct(&distcode, lengths + nlen, ndist);
    742     if (err && (err < 0 || ndist != distcode.count[0] + distcode.count[1]))
    743         return -8;      /* incomplete code ok only for single length 1 code */
    744 
    745     /* decode data until end-of-block code */
    746     return codes(s, &lencode, &distcode);
    747 }
    748 
    749 /*
    750  * Inflate source to dest.  On return, destlen and sourcelen are updated to the
    751  * size of the uncompressed data and the size of the deflate data respectively.
    752  * On success, the return value of puff() is zero.  If there is an error in the
    753  * source data, i.e. it is not in the deflate format, then a negative value is
    754  * returned.  If there is not enough input available or there is not enough
    755  * output space, then a positive error is returned.  In that case, destlen and
    756  * sourcelen are not updated to facilitate retrying from the beginning with the
    757  * provision of more input data or more output space.  In the case of invalid
    758  * inflate data (a negative error), the dest and source pointers are updated to
    759  * facilitate the debugging of deflators.
    760  *
    761  * puff() also has a mode to determine the size of the uncompressed output with
    762  * no output written.  For this dest must be (unsigned char *)0.  In this case,
    763  * the input value of *destlen is ignored, and on return *destlen is set to the
    764  * size of the uncompressed output.
    765  *
    766  * The return codes are:
    767  *
    768  *   2:  available inflate data did not terminate
    769  *   1:  output space exhausted before completing inflate
    770  *   0:  successful inflate
    771  *  -1:  invalid block type (type == 3)
    772  *  -2:  stored block length did not match one's complement
    773  *  -3:  dynamic block code description: too many length or distance codes
    774  *  -4:  dynamic block code description: code lengths codes incomplete
    775  *  -5:  dynamic block code description: repeat lengths with no first length
    776  *  -6:  dynamic block code description: repeat more than specified lengths
    777  *  -7:  dynamic block code description: invalid literal/length code lengths
    778  *  -8:  dynamic block code description: invalid distance code lengths
    779  *  -9:  dynamic block code description: missing end-of-block code
    780  * -10:  invalid literal/length or distance code in fixed or dynamic block
    781  * -11:  distance is too far back in fixed or dynamic block
    782  *
    783  * Format notes:
    784  *
    785  * - Three bits are read for each block to determine the kind of block and
    786  *   whether or not it is the last block.  Then the block is decoded and the
    787  *   process repeated if it was not the last block.
    788  *
    789  * - The leftover bits in the last byte of the deflate data after the last
    790  *   block (if it was a fixed or dynamic block) are undefined and have no
    791  *   expected values to check.
    792  */
    793 int puff(unsigned char *dest,           /* pointer to destination pointer */
    794          unsigned long *destlen,        /* amount of output space */
    795          const unsigned char *source,   /* pointer to source data pointer */
    796          unsigned long *sourcelen)      /* amount of input available */
    797 {
    798     struct state s;             /* input/output state */
    799     int last, type;             /* block information */
    800     int err;                    /* return value */
    801 
    802     /* initialize output state */
    803     s.out = dest;
    804     s.outlen = *destlen;                /* ignored if dest is NIL */
    805     s.outcnt = 0;
    806 
    807     /* initialize input state */
    808     s.in = source;
    809     s.inlen = *sourcelen;
    810     s.incnt = 0;
    811     s.bitbuf = 0;
    812     s.bitcnt = 0;
    813 
    814     /* return if bits() or decode() tries to read past available input */
    815     if (setjmp(s.env) != 0)             /* if came back here via longjmp() */
    816         err = 2;                        /* then skip do-loop, return error */
    817     else {
    818         /* process blocks until last block or error */
    819         do {
    820             last = bits(&s, 1);         /* one if last block */
    821             type = bits(&s, 2);         /* block type 0..3 */
    822             err = type == 0 ?
    823                     stored(&s) :
    824                     (type == 1 ?
    825                         fixed(&s) :
    826                         (type == 2 ?
    827                             dynamic(&s) :
    828                             -1));       /* type == 3, invalid */
    829             if (err != 0)
    830                 break;                  /* return with error */
    831         } while (!last);
    832     }
    833 
    834     /* update the lengths and return */
    835     if (err <= 0) {
    836         *destlen = s.outcnt;
    837         *sourcelen = s.incnt;
    838     }
    839     return err;
    840 }
    841