Merge branch 'bjorn/erts/zlib-1.2.8'

* bjorn/erts/zlib-1.2.8:
  configure: Prefer the system's zlib over own our zlib source
  Update preloaded modules
  Add the 'rle' zstrategy
  Don't make gzio.c dependent on the zutil.h header file
  Update zlib to 1.2.8
  erts/zlib: Remove unused file example.c

1 files changed, 375 insertions, 146 deletions

diff --git a/erts/emulator/zlib/deflate.c b/erts/emulator/zlib/deflate.c
index 92f4be57c5..943c26dfb2 100644
--- a/erts/emulator/zlib/deflate.c
+++ b/erts/emulator/zlib/deflate.c
@@ -1,10 +1,8 @@
 /* deflate.c -- compress data using the deflation algorithm
- * Copyright (C) 1995-2005 Jean-loup Gailly.
+ * Copyright (C) 1995-2013 Jean-loup Gailly and Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
-/* %ExternalCopyright% */
-
 /*
  *  ALGORITHM
  *
@@ -39,7 +37,7 @@
  *  REFERENCES
  *
  *      Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
- *      Available in http://www.ietf.org/rfc/rfc1951.txt
+ *      Available in http://tools.ietf.org/html/rfc1951
  *
  *      A description of the Rabin and Karp algorithm is given in the book
  *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
@@ -57,7 +55,7 @@
 #include "deflate.h"
 
 const char deflate_copyright[] =
-   " deflate 1.2.3 Copyright 1995-2005 Jean-loup Gailly ";
+   " deflate 1.2.8 Copyright 1995-2013 Jean-loup Gailly and Mark Adler ";
 /*
   If you use the zlib library in a product, an acknowledgment is welcome
   in the documentation of your product. If for some reason you cannot
@@ -84,19 +82,18 @@ local block_state deflate_fast   OF((deflate_state *s, int flush));
 #ifndef FASTEST
 local block_state deflate_slow   OF((deflate_state *s, int flush));
 #endif
+local block_state deflate_rle    OF((deflate_state *s, int flush));
+local block_state deflate_huff   OF((deflate_state *s, int flush));
 local void lm_init        OF((deflate_state *s));
 local void putShortMSB    OF((deflate_state *s, uInt b));
 local void flush_pending  OF((z_streamp strm));
 local int read_buf        OF((z_streamp strm, Bytef *buf, unsigned size));
-#ifndef FASTEST
 #ifdef ASMV
       void match_init OF((void)); /* asm code initialization */
       uInt longest_match  OF((deflate_state *s, IPos cur_match));
 #else
 local uInt longest_match  OF((deflate_state *s, IPos cur_match));
 #endif
-#endif
-local uInt longest_match_fast OF((deflate_state *s, IPos cur_match));
 
 #ifdef DEBUG
 local  void check_match OF((deflate_state *s, IPos start, IPos match,
@@ -115,11 +112,6 @@ local  void check_match OF((deflate_state *s, IPos start, IPos match,
 #endif
 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
 
-#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.
- */
-
 /* Values for max_lazy_match, good_match and max_chain_length, depending on
  * the desired pack level (0..9). The values given below have been tuned to
  * exclude worst case performance for pathological files. Better values may be
@@ -166,6 +158,9 @@ local const config configuration_table[10] = {
 struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
 #endif
 
+/* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */
+#define RANK(f) (((f) << 1) - ((f) > 4 ? 9 : 0))
+
 /* ===========================================================================
  * Update a hash value with the given input byte
  * IN  assertion: all calls to to UPDATE_HASH are made with consecutive
@@ -246,10 +241,19 @@ int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
 
     strm->msg = Z_NULL;
     if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
         strm->zalloc = zcalloc;
         strm->opaque = (voidpf)0;
+#endif
     }
-    if (strm->zfree == (free_func)0) strm->zfree = zcfree;
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zfree = zcfree;
+#endif
 
 #ifdef FASTEST
     if (level != 0) level = 1;
@@ -293,6 +297,8 @@ int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
     s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
     s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
 
+    s->high_water = 0;      /* nothing written to s->window yet */
+
     s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
 
     overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
@@ -302,7 +308,7 @@ int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
         s->pending_buf == Z_NULL) {
         s->status = FINISH_STATE;
-        strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
+        strm->msg = ERR_MSG(Z_MEM_ERROR);
         deflateEnd (strm);
         return Z_MEM_ERROR;
     }
@@ -323,43 +329,70 @@ int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
     uInt  dictLength;
 {
     deflate_state *s;
-    uInt length = dictLength;
-    uInt n;
-    IPos hash_head = 0;
+    uInt str, n;
+    int wrap;
+    unsigned avail;
+    z_const unsigned char *next;
 
-    if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
-        strm->state->wrap == 2 ||
-        (strm->state->wrap == 1 && strm->state->status != INIT_STATE))
+    if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL)
         return Z_STREAM_ERROR;
-
     s = strm->state;
-    if (s->wrap)
-        strm->adler = adler32(strm->adler, dictionary, dictLength);
+    wrap = s->wrap;
+    if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead)
+        return Z_STREAM_ERROR;
 
-    if (length < MIN_MATCH) return Z_OK;
-    if (length > MAX_DIST(s)) {
-        length = MAX_DIST(s);
-        dictionary += dictLength - length; /* use the tail of the dictionary */
+    /* when using zlib wrappers, compute Adler-32 for provided dictionary */
+    if (wrap == 1)
+        strm->adler = adler32(strm->adler, dictionary, dictLength);
+    s->wrap = 0;                    /* avoid computing Adler-32 in read_buf */
+
+    /* if dictionary would fill window, just replace the history */
+    if (dictLength >= s->w_size) {
+        if (wrap == 0) {            /* already empty otherwise */
+            CLEAR_HASH(s);
+            s->strstart = 0;
+            s->block_start = 0L;
+            s->insert = 0;
+        }
+        dictionary += dictLength - s->w_size;  /* use the tail */
+        dictLength = s->w_size;
     }
-    zmemcpy(s->window, dictionary, length);
-    s->strstart = length;
-    s->block_start = (long)length;
 
-    /* Insert all strings in the hash table (except for the last two bytes).
-     * s->lookahead stays null, so s->ins_h will be recomputed at the next
-     * call of fill_window.
-     */
-    s->ins_h = s->window[0];
-    UPDATE_HASH(s, s->ins_h, s->window[1]);
-    for (n = 0; n <= length - MIN_MATCH; n++) {
-        INSERT_STRING(s, n, hash_head);
+    /* insert dictionary into window and hash */
+    avail = strm->avail_in;
+    next = strm->next_in;
+    strm->avail_in = dictLength;
+    strm->next_in = (z_const Bytef *)dictionary;
+    fill_window(s);
+    while (s->lookahead >= MIN_MATCH) {
+        str = s->strstart;
+        n = s->lookahead - (MIN_MATCH-1);
+        do {
+            UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+            s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+            s->head[s->ins_h] = (Pos)str;
+            str++;
+        } while (--n);
+        s->strstart = str;
+        s->lookahead = MIN_MATCH-1;
+        fill_window(s);
     }
-    if (hash_head) hash_head = 0;  /* to make compiler happy */
+    s->strstart += s->lookahead;
+    s->block_start = (long)s->strstart;
+    s->insert = s->lookahead;
+    s->lookahead = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    strm->next_in = next;
+    strm->avail_in = avail;
+    s->wrap = wrap;
     return Z_OK;
 }
 
 /* ========================================================================= */
-int ZEXPORT deflateReset (strm)
+int ZEXPORT deflateResetKeep (strm)
     z_streamp strm;
 {
     deflate_state *s;
@@ -389,12 +422,23 @@ int ZEXPORT deflateReset (strm)
     s->last_flush = Z_NO_FLUSH;
 
     _tr_init(s);
-    lm_init(s);
 
     return Z_OK;
 }
 
 /* ========================================================================= */
+int ZEXPORT deflateReset (strm)
+    z_streamp strm;
+{
+    int ret;
+
+    ret = deflateResetKeep(strm);
+    if (ret == Z_OK)
+        lm_init(strm->state);
+    return ret;
+}
+
+/* ========================================================================= */
 int ZEXPORT deflateSetHeader (strm, head)
     z_streamp strm;
     gz_headerp head;
@@ -406,14 +450,42 @@ int ZEXPORT deflateSetHeader (strm, head)
 }
 
 /* ========================================================================= */
+int ZEXPORT deflatePending (strm, pending, bits)
+    unsigned *pending;
+    int *bits;
+    z_streamp strm;
+{
+    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
+    if (pending != Z_NULL)
+        *pending = strm->state->pending;
+    if (bits != Z_NULL)
+        *bits = strm->state->bi_valid;
+    return Z_OK;
+}
+
+/* ========================================================================= */
 int ZEXPORT deflatePrime (strm, bits, value)
     z_streamp strm;
     int bits;
     int value;
 {
+    deflate_state *s;
+    int put;
+
     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
-    strm->state->bi_valid = bits;
-    strm->state->bi_buf = (ush)(value & ((1 << bits) - 1));
+    s = strm->state;
+    if ((Bytef *)(s->d_buf) < s->pending_out + ((Buf_size + 7) >> 3))
+        return Z_BUF_ERROR;
+    do {
+        put = Buf_size - s->bi_valid;
+        if (put > bits)
+            put = bits;
+        s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid);
+        s->bi_valid += put;
+        _tr_flush_bits(s);
+        value >>= put;
+        bits -= put;
+    } while (bits);
     return Z_OK;
 }
 
@@ -440,9 +512,12 @@ int ZEXPORT deflateParams(strm, level, strategy)
     }
     func = configuration_table[s->level].func;
 
-    if (func != configuration_table[level].func && strm->total_in != 0) {
+    if ((strategy != s->strategy || func != configuration_table[level].func) &&
+        strm->total_in != 0) {
         /* Flush the last buffer: */
-        err = deflate(strm, Z_PARTIAL_FLUSH);
+        err = deflate(strm, Z_BLOCK);
+        if (err == Z_BUF_ERROR && s->pending == 0)
+            err = Z_OK;
     }
     if (s->level != level) {
         s->level = level;
@@ -486,33 +561,66 @@ int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
  * resulting from using fixed blocks instead of stored blocks, which deflate
  * can emit on compressed data for some combinations of the parameters.
  *
- * This function could be more sophisticated to provide closer upper bounds
- * for every combination of windowBits and memLevel, as well as wrap.
- * But even the conservative upper bound of about 14% expansion does not
- * seem onerous for output buffer allocation.
+ * This function could be more sophisticated to provide closer upper bounds for
+ * every combination of windowBits and memLevel.  But even the conservative
+ * upper bound of about 14% expansion does not seem onerous for output buffer
+ * allocation.
  */
 uLong ZEXPORT deflateBound(strm, sourceLen)
     z_streamp strm;
     uLong sourceLen;
 {
     deflate_state *s;
-    uLong destLen;
+    uLong complen, wraplen;
+    Bytef *str;
 
-    /* conservative upper bound */
-    destLen = sourceLen +
-              ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 11;
+    /* conservative upper bound for compressed data */
+    complen = sourceLen +
+              ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5;
 
-    /* if can't get parameters, return conservative bound */
+    /* if can't get parameters, return conservative bound plus zlib wrapper */
     if (strm == Z_NULL || strm->state == Z_NULL)
-        return destLen;
+        return complen + 6;
 
-    /* if not default parameters, return conservative bound */
+    /* compute wrapper length */
     s = strm->state;
+    switch (s->wrap) {
+    case 0:                                 /* raw deflate */
+        wraplen = 0;
+        break;
+    case 1:                                 /* zlib wrapper */
+        wraplen = 6 + (s->strstart ? 4 : 0);
+        break;
+    case 2:                                 /* gzip wrapper */
+        wraplen = 18;
+        if (s->gzhead != Z_NULL) {          /* user-supplied gzip header */
+            if (s->gzhead->extra != Z_NULL)
+                wraplen += 2 + s->gzhead->extra_len;
+            str = s->gzhead->name;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            str = s->gzhead->comment;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            if (s->gzhead->hcrc)
+                wraplen += 2;
+        }
+        break;
+    default:                                /* for compiler happiness */
+        wraplen = 6;
+    }
+
+    /* if not default parameters, return conservative bound */
     if (s->w_bits != 15 || s->hash_bits != 8 + 7)
-        return destLen;
+        return complen + wraplen;
 
     /* default settings: return tight bound for that case */
-    return compressBound(sourceLen);
+    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
+           (sourceLen >> 25) + 13 - 6 + wraplen;
 }
 
 /* =========================================================================
@@ -537,19 +645,22 @@ local void putShortMSB (s, b)
 local void flush_pending(strm)
     z_streamp strm;
 {
-    unsigned len = strm->state->pending;
+    unsigned len;
+    deflate_state *s = strm->state;
 
+    _tr_flush_bits(s);
+    len = s->pending;
     if (len > strm->avail_out) len = strm->avail_out;
     if (len == 0) return;
 
-    zmemcpy(strm->next_out, strm->state->pending_out, len);
+    zmemcpy(strm->next_out, s->pending_out, len);
     strm->next_out  += len;
-    strm->state->pending_out  += len;
+    s->pending_out  += len;
     strm->total_out += len;
     strm->avail_out  -= len;
-    strm->state->pending -= len;
-    if (strm->state->pending == 0) {
-        strm->state->pending_out = strm->state->pending_buf;
+    s->pending -= len;
+    if (s->pending == 0) {
+        s->pending_out = s->pending_buf;
     }
 }
 
@@ -562,7 +673,7 @@ int ZEXPORT deflate (strm, flush)
     deflate_state *s;
 
     if (strm == Z_NULL || strm->state == Z_NULL ||
-        flush > Z_FINISH || flush < 0) {
+        flush > Z_BLOCK || flush < 0) {
         return Z_STREAM_ERROR;
     }
     s = strm->state;
@@ -586,7 +697,7 @@ int ZEXPORT deflate (strm, flush)
             put_byte(s, 31);
             put_byte(s, 139);
             put_byte(s, 8);
-            if (s->gzhead == NULL) {
+            if (s->gzhead == Z_NULL) {
                 put_byte(s, 0);
                 put_byte(s, 0);
                 put_byte(s, 0);
@@ -613,7 +724,7 @@ int ZEXPORT deflate (strm, flush)
                             (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
                              4 : 0));
                 put_byte(s, s->gzhead->os & 0xff);
-                if (s->gzhead->extra != NULL) {
+                if (s->gzhead->extra != Z_NULL) {
                     put_byte(s, s->gzhead->extra_len & 0xff);
                     put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
                 }
@@ -655,7 +766,7 @@ int ZEXPORT deflate (strm, flush)
     }
 #ifdef GZIP
     if (s->status == EXTRA_STATE) {
-        if (s->gzhead->extra != NULL) {
+        if (s->gzhead->extra != Z_NULL) {
             uInt beg = s->pending;  /* start of bytes to update crc */
 
             while (s->gzindex < (s->gzhead->extra_len & 0xffff)) {
@@ -683,7 +794,7 @@ int ZEXPORT deflate (strm, flush)
             s->status = NAME_STATE;
     }
     if (s->status == NAME_STATE) {
-        if (s->gzhead->name != NULL) {
+        if (s->gzhead->name != Z_NULL) {
             uInt beg = s->pending;  /* start of bytes to update crc */
             int val;
 
@@ -714,7 +825,7 @@ int ZEXPORT deflate (strm, flush)
             s->status = COMMENT_STATE;
     }
     if (s->status == COMMENT_STATE) {
-        if (s->gzhead->comment != NULL) {
+        if (s->gzhead->comment != Z_NULL) {
             uInt beg = s->pending;  /* start of bytes to update crc */
             int val;
 
@@ -776,7 +887,7 @@ int ZEXPORT deflate (strm, flush)
      * flushes. For repeated and useless calls with Z_FINISH, we keep
      * returning Z_STREAM_END instead of Z_BUF_ERROR.
      */
-    } else if (strm->avail_in == 0 && flush <= old_flush &&
+    } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) &&
                flush != Z_FINISH) {
         ERR_RETURN(strm, Z_BUF_ERROR);
     }
@@ -792,7 +903,9 @@ int ZEXPORT deflate (strm, flush)
         (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
         block_state bstate;
 
-        bstate = (*(configuration_table[s->level].func))(s, flush);
+        bstate = s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) :
+                    (s->strategy == Z_RLE ? deflate_rle(s, flush) :
+                        (*(configuration_table[s->level].func))(s, flush));
 
         if (bstate == finish_started || bstate == finish_done) {
             s->status = FINISH_STATE;
@@ -813,13 +926,18 @@ int ZEXPORT deflate (strm, flush)
         if (bstate == block_done) {
             if (flush == Z_PARTIAL_FLUSH) {
                 _tr_align(s);
-            } else { /* FULL_FLUSH or SYNC_FLUSH */
+            } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */
                 _tr_stored_block(s, (char*)0, 0L, 0);
                 /* For a full flush, this empty block will be recognized
                  * as a special marker by inflate_sync().
                  */
                 if (flush == Z_FULL_FLUSH) {
                     CLEAR_HASH(s);             /* forget history */
+                    if (s->lookahead == 0) {
+                        s->strstart = 0;
+                        s->block_start = 0L;
+                        s->insert = 0;
+                    }
                 }
             }
             flush_pending(strm);
@@ -914,12 +1032,12 @@ int ZEXPORT deflateCopy (dest, source)
 
     ss = source->state;
 
-    zmemcpy(dest, source, sizeof(z_stream));
+    zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
 
     ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
     if (ds == Z_NULL) return Z_MEM_ERROR;
     dest->state = (struct internal_state FAR *) ds;
-    zmemcpy(ds, ss, sizeof(deflate_state));
+    zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state));
     ds->strm = dest;
 
     ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
@@ -935,8 +1053,8 @@ int ZEXPORT deflateCopy (dest, source)
     }
     /* following zmemcpy do not work for 16-bit MSDOS */
     zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
-    zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
-    zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
+    zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos));
+    zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos));
     zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
 
     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
@@ -970,15 +1088,15 @@ local int read_buf(strm, buf, size)
 
     strm->avail_in  -= len;
 
+    zmemcpy(buf, strm->next_in, len);
     if (strm->state->wrap == 1) {
-        strm->adler = adler32(strm->adler, strm->next_in, len);
+        strm->adler = adler32(strm->adler, buf, len);
     }
 #ifdef GZIP
     else if (strm->state->wrap == 2) {
-        strm->adler = crc32(strm->adler, strm->next_in, len);
+        strm->adler = crc32(strm->adler, buf, len);
     }
 #endif
-    zmemcpy(buf, strm->next_in, len);
     strm->next_in  += len;
     strm->total_in += len;
 
@@ -1005,6 +1123,7 @@ local void lm_init (s)
     s->strstart = 0;
     s->block_start = 0L;
     s->lookahead = 0;
+    s->insert = 0;
     s->match_length = s->prev_length = MIN_MATCH-1;
     s->match_available = 0;
     s->ins_h = 0;
@@ -1172,12 +1291,13 @@ local uInt longest_match(s, cur_match)
     return s->lookahead;
 }
 #endif /* ASMV */
-#endif /* FASTEST */
+
+#else /* FASTEST */
 
 /* ---------------------------------------------------------------------------
- * Optimized version for level == 1 or strategy == Z_RLE only
+ * Optimized version for FASTEST only
  */
-local uInt longest_match_fast(s, cur_match)
+local uInt longest_match(s, cur_match)
     deflate_state *s;
     IPos cur_match;                             /* current match */
 {
@@ -1230,6 +1350,8 @@ local uInt longest_match_fast(s, cur_match)
     return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
 }
 
+#endif /* FASTEST */
+
 #ifdef DEBUG
 /* ===========================================================================
  * Check that the match at match_start is indeed a match.
@@ -1276,6 +1398,8 @@ local void fill_window(s)
     unsigned more;    /* Amount of free space at the end of the window. */
     uInt wsize = s->w_size;
 
+    Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead");
+
     do {
         more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
 
@@ -1308,7 +1432,6 @@ local void fill_window(s)
                later. (Using level 0 permanently is not an optimal usage of
                zlib, so we don't care about this pathological case.)
              */
-            /* %%% avoid this when Z_RLE */
             n = s->hash_size;
             p = &s->head[n];
             do {
@@ -1329,7 +1452,7 @@ local void fill_window(s)
 #endif
             more += wsize;
         }
-        if (s->strm->avail_in == 0) return;
+        if (s->strm->avail_in == 0) break;
 
         /* If there was no sliding:
          *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
@@ -1348,39 +1471,88 @@ local void fill_window(s)
         s->lookahead += n;
 
         /* Initialize the hash value now that we have some input: */
-        if (s->lookahead >= MIN_MATCH) {
-            s->ins_h = s->window[s->strstart];
-            UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
+        if (s->lookahead + s->insert >= MIN_MATCH) {
+            uInt str = s->strstart - s->insert;
+            s->ins_h = s->window[str];
+            UPDATE_HASH(s, s->ins_h, s->window[str + 1]);
 #if MIN_MATCH != 3
             Call UPDATE_HASH() MIN_MATCH-3 more times
 #endif
+            while (s->insert) {
+                UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+                s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+                s->head[s->ins_h] = (Pos)str;
+                str++;
+                s->insert--;
+                if (s->lookahead + s->insert < MIN_MATCH)
+                    break;
+            }
         }
         /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
          * but this is not important since only literal bytes will be emitted.
          */
 
     } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
+
+    /* If the WIN_INIT bytes after the end of the current data have never been
+     * written, then zero those bytes in order to avoid memory check reports of
+     * the use of uninitialized (or uninitialised as Julian writes) bytes by
+     * the longest match routines.  Update the high water mark for the next
+     * time through here.  WIN_INIT is set to MAX_MATCH since the longest match
+     * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead.
+     */
+    if (s->high_water < s->window_size) {
+        ulg curr = s->strstart + (ulg)(s->lookahead);
+        ulg init;
+
+        if (s->high_water < curr) {
+            /* Previous high water mark below current data -- zero WIN_INIT
+             * bytes or up to end of window, whichever is less.
+             */
+            init = s->window_size - curr;
+            if (init > WIN_INIT)
+                init = WIN_INIT;
+            zmemzero(s->window + curr, (unsigned)init);
+            s->high_water = curr + init;
+        }
+        else if (s->high_water < (ulg)curr + WIN_INIT) {
+            /* High water mark at or above current data, but below current data
+             * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up
+             * to end of window, whichever is less.
+             */
+            init = (ulg)curr + WIN_INIT - s->high_water;
+            if (init > s->window_size - s->high_water)
+                init = s->window_size - s->high_water;
+            zmemzero(s->window + s->high_water, (unsigned)init);
+            s->high_water += init;
+        }
+    }
+
+    Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD,
+           "not enough room for search");
 }
 
 /* ===========================================================================
  * Flush the current block, with given end-of-file flag.
  * IN assertion: strstart is set to the end of the current match.
  */
-#define FLUSH_BLOCK_ONLY(s, eof) { \
+#define FLUSH_BLOCK_ONLY(s, last) { \
    _tr_flush_block(s, (s->block_start >= 0L ? \
                    (charf *)&s->window[(unsigned)s->block_start] : \
                    (charf *)Z_NULL), \
                 (ulg)((long)s->strstart - s->block_start), \
-                (eof)); \
+                (last)); \
    s->block_start = s->strstart; \
    flush_pending(s->strm); \
    Tracev((stderr,"[FLUSH]")); \
 }
 
 /* Same but force premature exit if necessary. */
-#define FLUSH_BLOCK(s, eof) { \
-   FLUSH_BLOCK_ONLY(s, eof); \
-   if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
+#define FLUSH_BLOCK(s, last) { \
+   FLUSH_BLOCK_ONLY(s, last); \
+   if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \
 }
 
 /* ===========================================================================
@@ -1439,8 +1611,14 @@ local block_state deflate_stored(s, flush)
             FLUSH_BLOCK(s, 0);
         }
     }
-    FLUSH_BLOCK(s, flush == Z_FINISH);
-    return flush == Z_FINISH ? finish_done : block_done;
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if ((long)s->strstart > s->block_start)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
 }
 
 /* ===========================================================================
@@ -1454,7 +1632,7 @@ local block_state deflate_fast(s, flush)
     deflate_state *s;
     int flush;
 {
-    IPos hash_head = NIL; /* head of the hash chain */
+    IPos hash_head;       /* head of the hash chain */
     int bflush;           /* set if current block must be flushed */
 
     for (;;) {
@@ -1474,6 +1652,7 @@ local block_state deflate_fast(s, flush)
         /* Insert the string window[strstart .. strstart+2] in the
          * dictionary, and set hash_head to the head of the hash chain:
          */
+        hash_head = NIL;
         if (s->lookahead >= MIN_MATCH) {
             INSERT_STRING(s, s->strstart, hash_head);
         }
@@ -1486,19 +1665,8 @@ local block_state deflate_fast(s, flush)
              * of window index 0 (in particular we have to avoid a match
              * of the string with itself at the start of the input file).
              */
-#ifdef FASTEST
-            if ((s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) ||
-                (s->strategy == Z_RLE && s->strstart - hash_head == 1)) {
-                s->match_length = longest_match_fast (s, hash_head);
-            }
-#else
-            if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
-                s->match_length = longest_match (s, hash_head);
-            } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
-                s->match_length = longest_match_fast (s, hash_head);
-            }
-#endif
-            /* longest_match() or longest_match_fast() sets match_start */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
         }
         if (s->match_length >= MIN_MATCH) {
             check_match(s, s->strstart, s->match_start, s->match_length);
@@ -1546,8 +1714,14 @@ local block_state deflate_fast(s, flush)
         }
         if (bflush) FLUSH_BLOCK(s, 0);
     }
-    FLUSH_BLOCK(s, flush == Z_FINISH);
-    return flush == Z_FINISH ? finish_done : block_done;
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
 }
 
 #ifndef FASTEST
@@ -1560,7 +1734,7 @@ local block_state deflate_slow(s, flush)
     deflate_state *s;
     int flush;
 {
-    IPos hash_head = NIL;    /* head of hash chain */
+    IPos hash_head;          /* head of hash chain */
     int bflush;              /* set if current block must be flushed */
 
     /* Process the input block. */
@@ -1581,6 +1755,7 @@ local block_state deflate_slow(s, flush)
         /* Insert the string window[strstart .. strstart+2] in the
          * dictionary, and set hash_head to the head of the hash chain:
          */
+        hash_head = NIL;
         if (s->lookahead >= MIN_MATCH) {
             INSERT_STRING(s, s->strstart, hash_head);
         }
@@ -1596,12 +1771,8 @@ local block_state deflate_slow(s, flush)
              * of window index 0 (in particular we have to avoid a match
              * of the string with itself at the start of the input file).
              */
-            if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
-                s->match_length = longest_match (s, hash_head);
-            } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
-                s->match_length = longest_match_fast (s, hash_head);
-            }
-            /* longest_match() or longest_match_fast() sets match_start */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
 
             if (s->match_length <= 5 && (s->strategy == Z_FILTERED
 #if TOO_FAR <= 32767
@@ -1674,12 +1845,17 @@ local block_state deflate_slow(s, flush)
         _tr_tally_lit(s, s->window[s->strstart-1], bflush);
         s->match_available = 0;
     }
-    FLUSH_BLOCK(s, flush == Z_FINISH);
-    return flush == Z_FINISH ? finish_done : block_done;
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
 }
 #endif /* FASTEST */
 
-#if 0
 /* ===========================================================================
  * For Z_RLE, simply look for runs of bytes, generate matches only of distance
  * one.  Do not maintain a hash table.  (It will be regenerated if this run of
@@ -1689,43 +1865,52 @@ local block_state deflate_rle(s, flush)
     deflate_state *s;
     int flush;
 {
-    int bflush;         /* set if current block must be flushed */
-    uInt run;           /* length of run */
-    uInt max;           /* maximum length of run */
-    uInt prev;          /* byte at distance one to match */
-    Bytef *scan;        /* scan for end of run */
+    int bflush;             /* set if current block must be flushed */
+    uInt prev;              /* byte at distance one to match */
+    Bytef *scan, *strend;   /* scan goes up to strend for length of run */
 
     for (;;) {
         /* Make sure that we always have enough lookahead, except
          * at the end of the input file. We need MAX_MATCH bytes
-         * for the longest encodable run.
+         * for the longest run, plus one for the unrolled loop.
          */
-        if (s->lookahead < MAX_MATCH) {
+        if (s->lookahead <= MAX_MATCH) {
             fill_window(s);
-            if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) {
+            if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) {
                 return need_more;
             }
             if (s->lookahead == 0) break; /* flush the current block */
         }
 
         /* See how many times the previous byte repeats */
-        run = 0;
-        if (s->strstart > 0) {      /* if there is a previous byte, that is */
-            max = s->lookahead < MAX_MATCH ? s->lookahead : MAX_MATCH;
+        s->match_length = 0;
+        if (s->lookahead >= MIN_MATCH && s->strstart > 0) {
             scan = s->window + s->strstart - 1;
-            prev = *scan++;
-            do {
-                if (*scan++ != prev)
-                    break;
-            } while (++run < max);
+            prev = *scan;
+            if (prev == *++scan && prev == *++scan && prev == *++scan) {
+                strend = s->window + s->strstart + MAX_MATCH;
+                do {
+                } while (prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         scan < strend);
+                s->match_length = MAX_MATCH - (int)(strend - scan);
+                if (s->match_length > s->lookahead)
+                    s->match_length = s->lookahead;
+            }
+            Assert(scan <= s->window+(uInt)(s->window_size-1), "wild scan");
         }
 
         /* Emit match if have run of MIN_MATCH or longer, else emit literal */
-        if (run >= MIN_MATCH) {
-            check_match(s, s->strstart, s->strstart - 1, run);
-            _tr_tally_dist(s, 1, run - MIN_MATCH, bflush);
-            s->lookahead -= run;
-            s->strstart += run;
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->strstart - 1, s->match_length);
+
+            _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+            s->strstart += s->match_length;
+            s->match_length = 0;
         } else {
             /* No match, output a literal byte */
             Tracevv((stderr,"%c", s->window[s->strstart]));
@@ -1735,7 +1920,51 @@ local block_state deflate_rle(s, flush)
         }
         if (bflush) FLUSH_BLOCK(s, 0);
     }
-    FLUSH_BLOCK(s, flush == Z_FINISH);
-    return flush == Z_FINISH ? finish_done : block_done;
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+
+/* ===========================================================================
+ * For Z_HUFFMAN_ONLY, do not look for matches.  Do not maintain a hash table.
+ * (It will be regenerated if this run of deflate switches away from Huffman.)
+ */
+local block_state deflate_huff(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    int bflush;             /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we have a literal to write. */
+        if (s->lookahead == 0) {
+            fill_window(s);
+            if (s->lookahead == 0) {
+                if (flush == Z_NO_FLUSH)
+                    return need_more;
+                break;      /* flush the current block */
+            }
+        }
+
+        /* Output a literal byte */
+        s->match_length = 0;
+        Tracevv((stderr,"%c", s->window[s->strstart]));
+        _tr_tally_lit (s, s->window[s->strstart], bflush);
+        s->lookahead--;
+        s->strstart++;
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
 }
-#endif