blast.c

开放源码的编译器open watcom 1.6.0版的源代码

        offs[len + 1] = offs[len] + h->count[len];

    /*
     * put symbols in table sorted by length, by symbol order within each
     * length
     */
    for (symbol = 0; symbol < n; symbol++)
        if (length[symbol] != 0)
            h->symbol[offs[length[symbol]]++] = symbol;

    /* return zero for complete set, positive for incomplete set */
    return left;
}

/*
 * Decode PKWare Compression Library stream.
 *
 * Format notes:
 *
 * - First byte is 0 if literals are uncoded or 1 if they are coded.  Second
 *   byte is 4, 5, or 6 for the number of extra bits in the distance code.
 *   This is the base-2 logarithm of the dictionary size minus six.
 *
 * - Compressed data is a combination of literals and length/distance pairs
 *   terminated by an end code.  Literals are either Huffman coded or
 *   uncoded bytes.  A length/distance pair is a coded length followed by a
 *   coded distance to represent a string that occurs earlier in the
 *   uncompressed data that occurs again at the current location.
 *
 * - A bit preceding a literal or length/distance pair indicates which comes
 *   next, 0 for literals, 1 for length/distance.
 *
 * - If literals are uncoded, then the next eight bits are the literal, in the
 *   normal bit order in th stream, i.e. no bit-reversal is needed. Similarly,
 *   no bit reversal is needed for either the length extra bits or the distance
 *   extra bits.
 *
 * - Literal bytes are simply written to the output.  A length/distance pair is
 *   an instruction to copy previously uncompressed bytes to the output.  The
 *   copy is from distance bytes back in the output stream, copying for length
 *   bytes.
 *
 * - Distances pointing before the beginning of the output data are not
 *   permitted.
 *
 * - Overlapped copies, where the length is greater than the distance, are
 *   allowed and common.  For example, a distance of one and a length of 518
 *   simply copies the last byte 518 times.  A distance of four and a length of
 *   twelve copies the last four bytes three times.  A simple forward copy
 *   ignoring whether the length is greater than the distance or not implements
 *   this correctly.
 */
local int decomp(struct state *s)
{
    int lit;            /* true if literals are coded */
    int dict;           /* log2(dictionary size) - 6 */
    int symbol;         /* decoded symbol, extra bits for distance */
    int len;            /* length for copy */
    int dist;           /* distance for copy */
    int copy;           /* copy counter */
    unsigned char *from, *to;   /* copy pointers */
    static int virgin = 1;                              /* build tables once */
    static short litcnt[MAXBITS+1], litsym[256];        /* litcode memory */
    static short lencnt[MAXBITS+1], lensym[16];         /* lencode memory */
    static short distcnt[MAXBITS+1], distsym[64];       /* distcode memory */
    static struct huffman litcode = {litcnt, litsym};   /* length code */
    static struct huffman lencode = {lencnt, lensym};   /* length code */
    static struct huffman distcode = {distcnt, distsym};/* distance code */
        /* bit lengths of literal codes */
    static const unsigned char litlen[] = {
        11, 124, 8, 7, 28, 7, 188, 13, 76, 4, 10, 8, 12, 10, 12, 10, 8, 23, 8,
        9, 7, 6, 7, 8, 7, 6, 55, 8, 23, 24, 12, 11, 7, 9, 11, 12, 6, 7, 22, 5,
        7, 24, 6, 11, 9, 6, 7, 22, 7, 11, 38, 7, 9, 8, 25, 11, 8, 11, 9, 12,
        8, 12, 5, 38, 5, 38, 5, 11, 7, 5, 6, 21, 6, 10, 53, 8, 7, 24, 10, 27,
        44, 253, 253, 253, 252, 252, 252, 13, 12, 45, 12, 45, 12, 61, 12, 45,
        44, 173};
        /* bit lengths of length codes 0..15 */
    static const unsigned char lenlen[] = {2, 35, 36, 53, 38, 23};
        /* bit lengths of distance codes 0..63 */
    static const unsigned char distlen[] = {2, 20, 53, 230, 247, 151, 248};
    static const short base[16] = {     /* base for length codes */
        3, 2, 4, 5, 6, 7, 8, 9, 10, 12, 16, 24, 40, 72, 136, 264};
    static const char extra[16] = {     /* extra bits for length codes */
        0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 3, 4, 5, 6, 7, 8};

    /* set up decoding tables (once--might not be thread-safe) */
    if (virgin) {
        construct(&litcode, litlen, sizeof(litlen));
        construct(&lencode, lenlen, sizeof(lenlen));
        construct(&distcode, distlen, sizeof(distlen));
        virgin = 0;
    }

    /* read header */
    lit = bits(s, 8);
    if (lit > 1) return -1;
    dict = bits(s, 8);
    if (dict < 4 || dict > 6) return -2;

    /* decode literals and length/distance pairs */
    do {
        if (bits(s, 1)) {
            /* get length */
            symbol = decode(s, &lencode);
            len = base[symbol] + bits(s, extra[symbol]);
            if (len == 519) break;              /* end code */

            /* get distance */
            symbol = len == 2 ? 2 : dict;
            dist = decode(s, &distcode) << symbol;
            dist += bits(s, symbol);
            dist++;
            if (s->first && dist > s->next)
                return -3;              /* distance too far back */

            /* copy length bytes from distance bytes back */
            do {
                to = s->out + s->next;
                from = to - dist;
                copy = MAXWIN;
                if (s->next < dist) {
                    from += copy;
                    copy = dist;
                }
                copy -= s->next;
                if (copy > len) copy = len;
                len -= copy;
                s->next += copy;
                do {
                    *to++ = *from++;
                } while (--copy);
                if (s->next == MAXWIN) {
                    if (s->outfun(s->outhow, s->out, s->next)) return 1;
                    s->next = 0;
                    s->first = 0;
                }
            } while (len != 0);
        }
        else {
            /* get literal and write it */
            symbol = lit ? decode(s, &litcode) : bits(s, 8);
            s->out[s->next++] = symbol;
            if (s->next == MAXWIN) {
                if (s->outfun(s->outhow, s->out, s->next)) return 1;
                s->next = 0;
                s->first = 0;
            }
        }
    } while (1);
    return 0;
}

/* See comments in blast.h */
int blast(blast_in infun, void *inhow, blast_out outfun, void *outhow)
{
    struct state s;             /* input/output state */
    int err;                    /* return value */

    /* initialize input state */
    s.infun = infun;
    s.inhow = inhow;
    s.left = 0;
    s.bitbuf = 0;
    s.bitcnt = 0;

    /* initialize output state */
    s.outfun = outfun;
    s.outhow = outhow;
    s.next = 0;
    s.first = 1;

    /* return if bits() or decode() tries to read past available input */
    if (setjmp(s.env) != 0)             /* if came back here via longjmp(), */
        err = 2;                        /*  then skip decomp(), return error */
    else
        err = decomp(&s);               /* decompress */

    /* write any leftover output and update the error code if needed */
    if (err != 1 && s.next && s.outfun(s.outhow, s.out, s.next) && err == 0)
        err = 1;
    return err;
}

#ifdef TEST
/* Example of how to use blast() */
#include <stdio.h>
#include <stdlib.h>

#define CHUNK 16384

local unsigned inf(void *how, unsigned char **buf)
{
    static unsigned char hold[CHUNK];

    *buf = hold;
    return fread(hold, 1, CHUNK, (FILE *)how);
}

local int outf(void *how, unsigned char *buf, unsigned len)
{
    return fwrite(buf, 1, len, (FILE *)how) != len;
}

/* Decompress a PKWare Compression Library stream from stdin to stdout */
int main(void)
{
    int ret, n;

    /* decompress to stdout */
    ret = blast(inf, stdin, outf, stdout);
    if (ret != 0) fprintf(stderr, "blast error: %d\n", ret);

    /* see if there are any leftover bytes */
    n = 0;
    while (getchar() != EOF) n++;
    if (n) fprintf(stderr, "blast warning: %d unused bytes of input\n", n);

    /* return blast() error code */
    return ret;
}
#endif