zlib-streams 1.0.7 → 1.1.0

package/src/deflate.h

@@ -0,0 +1,420 @@
+/* deflate.h -- internal compression state
+ * Copyright (C) 1995-2026 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.
+ */
+
+/* @(#) $Id$ */
+
+#ifndef DEFLATE_H
+#define DEFLATE_H
+
+#if defined(DEFLATE_CHUNK_WRITE_64LE)
+#include <stdint.h>
+#endif
+
+#include "zutil.h"
+
+/* Chromium's deflate.c calls zmemcpy/zmemzero on deflate_state, Posf and
+   z_streamp pointers, because its own zutil.h declares them as taking void*.
+   This build keeps madler's zutil.h, where they take Bytef*, so cast at the
+   call sites. The inner name is not re-expanded, so these stay function calls. */
+#define zmemcpy(dest, source, len) zmemcpy((Bytef *)(dest), (const Bytef *)(source), (len))
+#define zmemzero(dest, len) zmemzero((Bytef *)(dest), (len))
+
+/* 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 LIT_MEM to slightly increase the speed of deflate (order 1% to 2%) at
+   the cost of a larger memory footprint */
+/* #define LIT_MEM */
+
+/* ===========================================================================
+ * 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 */
+
+#if defined(DEFLATE_CHUNK_WRITE_64LE)
+#define Buf_size 64
+#else
+#define Buf_size 16
+#endif
+/* size of bit buffer in bi_buf */
+
+#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 */
+
+
+/* Data structure describing a single value and its code string. */
+typedef struct ct_data_s {
+    union {
+        ush  freq;       /* frequency count */
+        ush  code;       /* bit string */
+    } fc;
+    union {
+        ush  dad;        /* father node in Huffman tree */
+        ush  len;        /* length of bit string */
+    } dl;
+} FAR 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 */
+} FAR tree_desc;
+
+typedef ush Pos;
+typedef Pos FAR Posf;
+typedef unsigned IPos;
+
+/* A Pos is an index in the character window. We use short instead of int to
+ * save space in the various tables. IPos is used only for parameter passing.
+ */
+
+typedef struct internal_state {
+    z_streamp strm;      /* pointer back to this zlib stream */
+    int   status;        /* as the name implies */
+    Bytef *pending_buf;  /* output still pending */
+    ulg   pending_buf_size; /* size of pending_buf */
+    Bytef *pending_out;  /* next pending byte to output to the stream */
+    ulg   pending;       /* nb of bytes in the pending buffer */
+    int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
+    gz_headerp  gzhead;  /* gzip header information to write */
+    ulg   gzindex;       /* where in extra, name, or comment */
+    Byte  method;        /* can only be DEFLATED */
+    int   last_flush;    /* value of flush param for previous deflate call */
+    unsigned crc0[4 * 5];
+    /* used by deflate.c: */
+
+    uInt  w_size;        /* LZ77 window size (32K by default) */
+    uInt  w_bits;        /* log2(w_size)  (8..16) */
+    uInt  w_mask;        /* w_size - 1 */
+
+    Bytef *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.
+     */
+
+    ulg window_size;
+    /* Actual size of window: 2*wSize, except when the user input buffer
+     * is directly used as sliding window.
+     */
+
+    Posf *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.
+     */
+
+    Posf *head; /* Heads of the hash chains or NIL. */
+
+    uInt  ins_h;          /* hash index of string to be inserted */
+    uInt  hash_size;      /* number of elements in hash table */
+    uInt  hash_bits;      /* log2(hash_size) */
+    uInt  hash_mask;      /* hash_size-1 */
+
+    uInt  hash_shift;
+    /* Number of bits by which ins_h must be shifted at each input
+     * step. It must be such that after MIN_MATCH steps, the oldest
+     * byte no longer takes part in the hash key, that is:
+     *   hash_shift * MIN_MATCH >= hash_bits
+     */
+
+    long block_start;
+    /* Window position at the beginning of the current output block. Gets
+     * negative when the window is moved backwards.
+     */
+
+    uInt match_length;           /* length of best match */
+    IPos prev_match;             /* previous match */
+    int match_available;         /* set if previous match exists */
+    uInt strstart;               /* start of string to insert */
+    uInt match_start;            /* start of matching string */
+    uInt lookahead;              /* number of valid bytes ahead in window */
+
+    uInt 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.
+     */
+
+    uInt 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.
+     */
+
+    uInt 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*/
+
+    uInt good_match;
+    /* Use a faster search when the previous match is longer than this */
+
+    int nice_match; /* Stop searching when current match exceeds this */
+
+                /* 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 */
+
+    ush 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.
+     */
+
+    uch depth[2*L_CODES+1];
+    /* Depth of each subtree used as tie breaker for trees of equal frequency
+     */
+
+#ifdef LIT_MEM
+    ushf *d_buf;          /* buffer for distances */
+    uchf *l_buf;          /* buffer for literals/lengths */
+#else
+    uchf *sym_buf;        /* buffer for distances and literals/lengths */
+#endif
+
+    uInt  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
+     */
+
+    uInt sym_next;      /* running index in symbol buffer */
+    uInt sym_end;       /* symbol table full when sym_next reaches this */
+
+    ulg opt_len;        /* bit length of current block with optimal trees */
+    ulg static_len;     /* bit length of current block with static trees */
+    uInt matches;       /* number of string matches in current block */
+    uInt insert;        /* bytes at end of window left to insert */
+
+#ifdef ZLIB_DEBUG
+    ulg compressed_len; /* total bit length of compressed file mod 2^32 */
+    ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
+#endif
+
+#if defined(DEFLATE_CHUNK_WRITE_64LE)
+    uint64_t bi_buf;
+#else
+    ush bi_buf;
+#endif
+    /* Output buffer. bits are inserted starting at the bottom (least
+     * significant bits).
+     */
+    int bi_valid;
+    /* Number of valid bits in bi_buf.  All bits above the last valid bit
+     * are always zero.
+     */
+    int bi_used;
+    /* Last number of used bits when going to a byte boundary.
+     */
+
+    ulg 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.
+     */
+
+    uInt chromium_zlib_hash;
+    /* 0 if Rabin-Karp rolling hash is enabled, non-zero if chromium zlib
+     * hash is enabled.
+     */
+
+#if defined(QAT_COMPRESSION_ENABLED)
+    /* Pointer to a struct that contains the current state of the QAT
+     * stream.
+     */
+    struct qat_deflate *qat_s;
+#endif
+
+    int slid;
+    /* True if the hash table has been slid since it was cleared. */
+
+} FAR deflate_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++] = (Bytef)(c);}
+
+
+#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 */
+
+        /* in trees.c */
+void ZLIB_INTERNAL _tr_init(deflate_state *s);
+int ZLIB_INTERNAL _tr_tally(deflate_state *s, unsigned dist, unsigned lc);
+void ZLIB_INTERNAL _tr_flush_block(deflate_state *s, charf *buf,
+                                   ulg stored_len, int last);
+void ZLIB_INTERNAL _tr_flush_bits(deflate_state *s);
+void ZLIB_INTERNAL _tr_align(deflate_state *s);
+void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
+                                    ulg stored_len, int last);
+
+#define d_code(dist) \
+   ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
+/* Mapping from a distance to a distance code. dist is the distance - 1 and
+ * must not have side effects. _dist_code[256] and _dist_code[257] are never
+ * used.
+ */
+
+#ifndef ZLIB_DEBUG
+/* Inline versions of _tr_tally for speed: */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+  extern uch ZLIB_INTERNAL _length_code[];
+  extern uch ZLIB_INTERNAL _dist_code[];
+#else
+  extern const uch ZLIB_INTERNAL _length_code[];
+  extern const uch ZLIB_INTERNAL _dist_code[];
+#endif
+
+#ifdef LIT_MEM
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->d_buf[s->sym_next] = 0; \
+    s->l_buf[s->sym_next++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (uch)(length); \
+    ush dist = (ush)(distance); \
+    s->d_buf[s->sym_next] = dist; \
+    s->l_buf[s->sym_next++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+  }
+#else
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->sym_buf[s->sym_next++] = 0; \
+    s->sym_buf[s->sym_next++] = 0; \
+    s->sym_buf[s->sym_next++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (uch)(length); \
+    ush dist = (ush)(distance); \
+    s->sym_buf[s->sym_next++] = (uch)dist; \
+    s->sym_buf[s->sym_next++] = (uch)(dist >> 8); \
+    s->sym_buf[s->sym_next++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+  }
+#endif
+#else
+# define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
+# define _tr_tally_dist(s, distance, length, flush) \
+              flush = _tr_tally(s, distance, length)
+#endif
+
+/* Functions that are SIMD optimised on x86 */
+void ZLIB_INTERNAL crc_fold_init(deflate_state* const s);
+void ZLIB_INTERNAL crc_fold_copy(deflate_state* const s,
+                                 unsigned char* dst,
+                                 const unsigned char* src,
+                                 long len);
+unsigned ZLIB_INTERNAL crc_fold_512to32(deflate_state* const s);
+
+#endif /* DEFLATE_H */

package/src/deflate_crc32.c

@@ -0,0 +1,22 @@
+/* deflate_crc32.c -- gzip CRC-32 helpers for the Chromium deflate.c
+ *
+ * Chromium's crc32.c owns crc_reset(), crc_finalize() and copy_with_crc(),
+ * which deflate.c calls unconditionally on the gzip path. This build keeps
+ * madler's crc32.c, so the three functions are provided here without the
+ * SIMD folding they carry upstream.
+ */
+
+#include "deflate.h"
+
+void ZLIB_INTERNAL crc_reset(deflate_state *const s) {
+    s->strm->adler = crc32(0L, Z_NULL, 0);
+}
+
+void ZLIB_INTERNAL crc_finalize(deflate_state *const s) {
+    (void)s;
+}
+
+void ZLIB_INTERNAL copy_with_crc(z_streamp strm, Bytef *dst, long size) {
+    zmemcpy(dst, strm->next_in, size);
+    strm->adler = crc32(strm->adler, dst, size);
+}

package/src/inffast.h

@@ -0,0 +1,26 @@
+/* inffast.h -- header to use inffast.c
+ * Copyright (C) 1995-2003, 2010 Mark Adler
+ * 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.
+ */
+
+/* INFLATE_FAST_MIN_INPUT: the minimum number of input bytes needed so that
+   we can safely call inflate_fast() with only one up-front bounds check. One
+   length/distance code pair (15 bits for the length code, 5 bits for length
+   extra, 15 bits for the distance code, 13 bits for distance extra) requires
+   reading up to 48 input bits (6 bytes).
+*/
+#define INFLATE_FAST_MIN_INPUT 6
+
+/* INFLATE_FAST_MIN_OUTPUT: the minimum number of output bytes needed so that
+   we can safely call inflate_fast() with only one up-front bounds check. One
+   length/distance code pair can output up to 258 bytes, which is the maximum
+   length that can be coded.
+ */
+#define INFLATE_FAST_MIN_OUTPUT 258
+
+void ZLIB_INTERNAL inflate_fast(z_streamp strm, unsigned start);