#ifndef DEFLATE_H_ #define DEFLATE_H_ /* deflate.h -- internal compression state * Copyright (C) 1995-2016 Jean-loup Gailly * For conditions of distribution and use, see copyright notice in zlib.h */ /* WARNING: this file should *not* be used by applications. It is part of the implementation of the compression library and is subject to change. Applications should only use zlib.h. */ #include "zutil.h" #include "zendian.h" /* define NO_GZIP when compiling if you want to disable gzip header and trailer creation by deflate(). NO_GZIP would be used to avoid linking in the crc code when it is not needed. For shared libraries, gzip encoding should be left enabled. */ #ifndef NO_GZIP # define GZIP #endif #define NIL 0 /* Tail of hash chains */ /* =========================================================================== * Internal compression state. */ #define LENGTH_CODES 29 /* number of length codes, not counting the special END_BLOCK code */ #define LITERALS 256 /* number of literal bytes 0..255 */ #define L_CODES (LITERALS+1+LENGTH_CODES) /* number of Literal or Length codes, including the END_BLOCK code */ #define D_CODES 30 /* number of distance codes */ #define BL_CODES 19 /* number of codes used to transfer the bit lengths */ #define HEAP_SIZE (2*L_CODES+1) /* maximum heap size */ #define MAX_BITS 15 /* All codes must not exceed MAX_BITS bits */ #define BIT_BUF_SIZE 64 /* size of bit buffer in bi_buf */ #define END_BLOCK 256 /* end of block literal code */ #define INIT_STATE 42 /* zlib header -> BUSY_STATE */ #ifdef GZIP # define GZIP_STATE 57 /* gzip header -> BUSY_STATE | EXTRA_STATE */ #endif #define EXTRA_STATE 69 /* gzip extra block -> NAME_STATE */ #define NAME_STATE 73 /* gzip file name -> COMMENT_STATE */ #define COMMENT_STATE 91 /* gzip comment -> HCRC_STATE */ #define HCRC_STATE 103 /* gzip header CRC -> BUSY_STATE */ #define BUSY_STATE 113 /* deflate -> FINISH_STATE */ #define FINISH_STATE 666 /* stream complete */ /* Stream status */ #define HASH_BITS 16u /* log2(HASH_SIZE) */ #define HASH_SIZE 65536u /* number of elements in hash table */ #define HASH_MASK (HASH_SIZE - 1u) /* HASH_SIZE-1 */ /* Data structure describing a single value and its code string. */ typedef struct ct_data_s { union { uint16_t freq; /* frequency count */ uint16_t code; /* bit string */ } fc; union { uint16_t dad; /* father node in Huffman tree */ uint16_t len; /* length of bit string */ } dl; } ct_data; #define Freq fc.freq #define Code fc.code #define Dad dl.dad #define Len dl.len typedef struct static_tree_desc_s static_tree_desc; typedef struct tree_desc_s { ct_data *dyn_tree; /* the dynamic tree */ int max_code; /* largest code with non zero frequency */ const static_tree_desc *stat_desc; /* the corresponding static tree */ } tree_desc; typedef uint16_t Pos; /* A Pos is an index in the character window. We use short instead of int to * save space in the various tables. */ typedef struct internal_state { PREFIX3(stream) *strm; /* pointer back to this zlib stream */ unsigned char *pending_buf; /* output still pending */ unsigned char *pending_out; /* next pending byte to output to the stream */ uint32_t pending_buf_size; /* size of pending_buf */ uint32_t pending; /* nb of bytes in the pending buffer */ int wrap; /* bit 0 true for zlib, bit 1 true for gzip */ uint32_t gzindex; /* where in extra, name, or comment */ PREFIX(gz_headerp) gzhead; /* gzip header information to write */ int status; /* as the name implies */ int last_flush; /* value of flush param for previous deflate call */ int reproducible; /* Whether reproducible compression results are required. */ int block_open; /* Whether or not a block is currently open for the QUICK deflation scheme. * This is set to 1 if there is an active block, or 0 if the block was just closed. */ /* used by deflate.c: */ unsigned int w_size; /* LZ77 window size (32K by default) */ unsigned int w_bits; /* log2(w_size) (8..16) */ unsigned int w_mask; /* w_size - 1 */ unsigned int lookahead; /* number of valid bytes ahead in window */ unsigned int high_water; /* High water mark offset in window for initialized bytes -- bytes above * this are set to zero in order to avoid memory check warnings when * longest match routines access bytes past the input. This is then * updated to the new high water mark. */ unsigned int window_size; /* Actual size of window: 2*wSize, except when the user input buffer * is directly used as sliding window. */ unsigned char *window; /* Sliding window. Input bytes are read into the second half of the window, * and move to the first half later to keep a dictionary of at least wSize * bytes. With this organization, matches are limited to a distance of * wSize-MAX_MATCH bytes, but this ensures that IO is always * performed with a length multiple of the block size. Also, it limits * the window size to 64K, which is quite useful on MSDOS. * To do: use the user input buffer as sliding window. */ Pos *prev; /* Link to older string with same hash index. To limit the size of this * array to 64K, this link is maintained only for the last 32K strings. * An index in this array is thus a window index modulo 32K. */ Pos *head; /* Heads of the hash chains or NIL. */ int block_start; /* Window position at the beginning of the current output block. Gets * negative when the window is moved backwards. */ unsigned int match_length; /* length of best match */ Pos prev_match; /* previous match */ int match_available; /* set if previous match exists */ unsigned int strstart; /* start of string to insert */ unsigned int match_start; /* start of matching string */ unsigned int prev_length; /* Length of the best match at previous step. Matches not greater than this * are discarded. This is used in the lazy match evaluation. */ unsigned int max_chain_length; /* To speed up deflation, hash chains are never searched beyond this length. * A higher limit improves compression ratio but degrades the speed. */ unsigned int max_lazy_match; /* Attempt to find a better match only when the current match is strictly smaller * than this value. This mechanism is used only for compression levels >= 4. */ # define max_insert_length max_lazy_match /* Insert new strings in the hash table only if the match length is not * greater than this length. This saves time but degrades compression. * max_insert_length is used only for compression levels <= 3. */ int level; /* compression level (1..9) */ int strategy; /* favor or force Huffman coding*/ unsigned int good_match; /* Use a faster search when the previous match is longer than this */ int nice_match; /* Stop searching when current match exceeds this */ #ifdef X86_PCLMULQDQ_CRC unsigned crc0[4 * 5]; #endif /* used by trees.c: */ /* Didn't use ct_data typedef below to suppress compiler warning */ struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */ struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */ struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */ struct tree_desc_s l_desc; /* desc. for literal tree */ struct tree_desc_s d_desc; /* desc. for distance tree */ struct tree_desc_s bl_desc; /* desc. for bit length tree */ uint16_t bl_count[MAX_BITS+1]; /* number of codes at each bit length for an optimal tree */ int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */ int heap_len; /* number of elements in the heap */ int heap_max; /* element of largest frequency */ /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used. * The same heap array is used to build all trees. */ unsigned char depth[2*L_CODES+1]; /* Depth of each subtree used as tie breaker for trees of equal frequency */ unsigned int lit_bufsize; /* Size of match buffer for literals/lengths. There are 4 reasons for * limiting lit_bufsize to 64K: * - frequencies can be kept in 16 bit counters * - if compression is not successful for the first block, all input * data is still in the window so we can still emit a stored block even * when input comes from standard input. (This can also be done for * all blocks if lit_bufsize is not greater than 32K.) * - if compression is not successful for a file smaller than 64K, we can * even emit a stored file instead of a stored block (saving 5 bytes). * This is applicable only for zip (not gzip or zlib). * - creating new Huffman trees less frequently may not provide fast * adaptation to changes in the input data statistics. (Take for * example a binary file with poorly compressible code followed by * a highly compressible string table.) Smaller buffer sizes give * fast adaptation but have of course the overhead of transmitting * trees more frequently. * - I can't count above 4 */ unsigned char *sym_buf; /* buffer for distances and literals/lengths */ unsigned int sym_next; /* running index in sym_buf */ unsigned int sym_end; /* symbol table full when sym_next reaches this */ unsigned long opt_len; /* bit length of current block with optimal trees */ unsigned long static_len; /* bit length of current block with static trees */ unsigned int matches; /* number of string matches in current block */ unsigned int insert; /* bytes at end of window left to insert */ #ifdef ZLIB_DEBUG unsigned long compressed_len; /* total bit length of compressed file mod 2^32 */ unsigned long bits_sent; /* bit length of compressed data sent mod 2^32 */ #endif uint64_t bi_buf; /* Output buffer. bits are inserted starting at the bottom (least significant bits). */ int32_t bi_valid; /* Number of valid bits in bi_buf. All bits above the last valid bit are always zero. */ } deflate_state; typedef enum { need_more, /* block not completed, need more input or more output */ block_done, /* block flush performed */ finish_started, /* finish started, need only more output at next deflate */ finish_done /* finish done, accept no more input or output */ } block_state; /* Output a byte on the stream. * IN assertion: there is enough room in pending_buf. */ #define put_byte(s, c) { \ s->pending_buf[s->pending++] = (unsigned char)(c); \ } /* =========================================================================== * Output a short LSB first on the stream. * IN assertion: there is enough room in pending_buf. */ static inline void put_short(deflate_state *s, uint16_t w) { #if defined(UNALIGNED_OK) *(uint16_t *)(&s->pending_buf[s->pending]) = w; s->pending += 2; #else put_byte(s, (w & 0xff)); put_byte(s, ((w >> 8) & 0xff)); #endif } /* =========================================================================== * Output a short MSB first on the stream. * IN assertion: there is enough room in pending_buf. */ static inline void put_short_msb(deflate_state *s, uint16_t w) { put_byte(s, ((w >> 8) & 0xff)); put_byte(s, (w & 0xff)); } /* =========================================================================== * Output a 32-bit unsigned int LSB first on the stream. * IN assertion: there is enough room in pending_buf. */ static inline void put_uint32(deflate_state *s, uint32_t dw) { #if defined(UNALIGNED_OK) *(uint32_t *)(&s->pending_buf[s->pending]) = dw; s->pending += 4; #else put_byte(s, (dw & 0xff)); put_byte(s, ((dw >> 8) & 0xff)); put_byte(s, ((dw >> 16) & 0xff)); put_byte(s, ((dw >> 24) & 0xff)); #endif } /* =========================================================================== * Output a 32-bit unsigned int MSB first on the stream. * IN assertion: there is enough room in pending_buf. */ static inline void put_uint32_msb(deflate_state *s, uint32_t dw) { #if defined(UNALIGNED_OK) *(uint32_t *)(&s->pending_buf[s->pending]) = ZSWAP32(dw); s->pending += 4; #else put_byte(s, ((dw >> 24) & 0xff)); put_byte(s, ((dw >> 16) & 0xff)); put_byte(s, ((dw >> 8) & 0xff)); put_byte(s, (dw & 0xff)); #endif } /* =========================================================================== * Output a 64-bit unsigned int LSB first on the stream. * IN assertion: there is enough room in pending_buf. */ static inline void put_uint64(deflate_state *s, uint64_t lld) { #if defined(UNALIGNED64_OK) *(uint64_t *)(&s->pending_buf[s->pending]) = lld; s->pending += 8; #elif defined(UNALIGNED_OK) *(uint32_t *)(&s->pending_buf[s->pending]) = lld & 0xffffffff; s->pending += 4; *(uint32_t *)(&s->pending_buf[s->pending]) = (lld >> 32) & 0xffffffff; s->pending += 4; #else put_byte(s, (lld & 0xff)); put_byte(s, ((lld >> 8) & 0xff)); put_byte(s, ((lld >> 16) & 0xff)); put_byte(s, ((lld >> 24) & 0xff)); put_byte(s, ((lld >> 32) & 0xff)); put_byte(s, ((lld >> 40) & 0xff)); put_byte(s, ((lld >> 48) & 0xff)); put_byte(s, ((lld >> 56) & 0xff)); #endif } #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) /* Minimum amount of lookahead, except at the end of the input file. * See deflate.c for comments about the MIN_MATCH+1. */ #define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD) /* In order to simplify the code, particularly on 16 bit machines, match * distances are limited to MAX_DIST instead of WSIZE. */ #define WIN_INIT MAX_MATCH /* Number of bytes after end of data in window to initialize in order to avoid memory checker errors from longest match routines */ void Z_INTERNAL fill_window(deflate_state *s); void Z_INTERNAL slide_hash_c(deflate_state *s); /* in trees.c */ void Z_INTERNAL zng_tr_init(deflate_state *s); void Z_INTERNAL zng_tr_flush_block(deflate_state *s, char *buf, uint32_t stored_len, int last); void Z_INTERNAL zng_tr_flush_bits(deflate_state *s); void Z_INTERNAL zng_tr_align(deflate_state *s); void Z_INTERNAL zng_tr_stored_block(deflate_state *s, char *buf, uint32_t stored_len, int last); unsigned Z_INTERNAL bi_reverse(unsigned code, int len); void Z_INTERNAL flush_pending(PREFIX3(streamp) strm); #define d_code(dist) ((dist) < 256 ? zng_dist_code[dist] : zng_dist_code[256+((dist)>>7)]) /* Mapping from a distance to a distance code. dist is the distance - 1 and * must not have side effects. zng_dist_code[256] and zng_dist_code[257] are never * used. */ /* Bit buffer and compress bits calculation debugging */ #ifdef ZLIB_DEBUG # define cmpr_bits_add(s, len) s->compressed_len += (len) # define cmpr_bits_align(s) s->compressed_len = (s->compressed_len + 7) & ~7L # define sent_bits_add(s, bits) s->bits_sent += (bits) # define sent_bits_align(s) s->bits_sent = (s->bits_sent + 7) & ~7L #else # define cmpr_bits_add(s, len) (void)(len) # define cmpr_bits_align(s) # define sent_bits_add(s, bits) (void)(bits) # define sent_bits_align(s) #endif #endif /* DEFLATE_H_ */