grpc 1.71.0 → 1.72.0

data/third_party/utf8_range/utf8_range_neon.inc

@@ -0,0 +1,117 @@
+#include <arm_neon.h>
+
+/* This code is almost the same as SSE implementation, please reference
+ * utf8-range-sse.inc for detailed explanation.
+ * The only difference is the range adjustment step. NEON code is more
+ * straightforward.
+ */
+
+static FORCE_INLINE_ATTR inline size_t utf8_range_ValidateUTF8Simd(
+    const char* data_original, const char* data, const char* end,
+    int return_position) {
+  const uint8x16_t first_len_tbl = {
+      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 2, 3,
+  };
+  const uint8x16_t first_range_tbl = {
+      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8,
+  };
+  const uint8x16_t range_min_tbl = {
+      0x00, 0x80, 0x80, 0x80, 0xA0, 0x80, 0x90, 0x80,
+      0xC2, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+  };
+  const uint8x16_t range_max_tbl = {
+      0x7F, 0xBF, 0xBF, 0xBF, 0xBF, 0x9F, 0xBF, 0x8F,
+      0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+  };
+  /* Range adjustment in NEON uint8x16x2 table. Note that lanes are interleaved
+   * in register. The table below is plotted vertically to ease understanding.
+   * The 1st column is for E0~EF, 2nd column for F0~FF.
+   */
+  // clang-format off
+  const uint8_t range_adjust_tbl_data[] = {
+    /* index -> 0~15  16~31 <- index */
+    /*  E0 -> */ 2,     3, /* <- F0  */
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     4, /* <- F4  */
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     0,
+                 0,     0,
+    /*  ED -> */ 3,     0,
+                 0,     0,
+                 0,     0,
+  };
+  // clang-format on
+  const uint8x16x2_t range_adjust_tbl = vld2q_u8(range_adjust_tbl_data);
+
+  const uint8x16_t const_1 = vdupq_n_u8(1);
+  const uint8x16_t const_2 = vdupq_n_u8(2);
+  const uint8x16_t const_e0 = vdupq_n_u8(0xE0);
+
+  uint8x16_t prev_input = vdupq_n_u8(0);
+  uint8x16_t prev_first_len = vdupq_n_u8(0);
+  uint8x16_t error = vdupq_n_u8(0);
+
+  while (end - data >= 16) {
+    const uint8x16_t input = vld1q_u8((const uint8_t*)data);
+
+    const uint8x16_t high_nibbles = vshrq_n_u8(input, 4);
+
+    const uint8x16_t first_len = vqtbl1q_u8(first_len_tbl, high_nibbles);
+
+    uint8x16_t range = vqtbl1q_u8(first_range_tbl, high_nibbles);
+
+    range = vorrq_u8(range, vextq_u8(prev_first_len, first_len, 15));
+
+    uint8x16_t shift2 = vextq_u8(prev_first_len, first_len, 14);
+    shift2 = vqsubq_u8(shift2, const_1);
+    range = vorrq_u8(range, shift2);
+
+    uint8x16_t shift3 = vextq_u8(prev_first_len, first_len, 13);
+    shift3 = vqsubq_u8(shift3, const_2);
+    range = vorrq_u8(range, shift3);
+
+    uint8x16_t shift1 = vextq_u8(prev_input, input, 15);
+    shift1 = vsubq_u8(shift1, const_e0);
+    range = vaddq_u8(range, vqtbl2q_u8(range_adjust_tbl, shift1));
+
+    const uint8x16_t min_range = vqtbl1q_u8(range_min_tbl, range);
+    const uint8x16_t max_range = vqtbl1q_u8(range_max_tbl, range);
+
+    if (return_position) {
+      error = vcltq_u8(input, min_range);
+      error = vorrq_u8(error, vcgtq_u8(input, max_range));
+      if (vmaxvq_u32(vreinterpretq_u32_u8(error))) {
+        break;
+      }
+    } else {
+      error = vorrq_u8(error, vcltq_u8(input, min_range));
+      error = vorrq_u8(error, vcgtq_u8(input, max_range));
+    }
+
+    prev_input = input;
+    prev_first_len = first_len;
+
+    data += 16;
+  }
+
+  if (return_position && data == data_original) {
+    return utf8_range_ValidateUTF8Naive(data, end, return_position);
+  }
+  const int32_t prev = vgetq_lane_s32(vreinterpretq_s32_u8(prev_input), 3);
+  data -= utf8_range_CodepointSkipBackwards(prev);
+  if (return_position) {
+    return (data - data_original) +
+           utf8_range_ValidateUTF8Naive(data, end, return_position);
+  }
+  if (vmaxvq_u32(vreinterpretq_u32_u8(error))) {
+    return 0;
+  }
+  return utf8_range_ValidateUTF8Naive(data, end, return_position);
+}

data/third_party/utf8_range/utf8_range_sse.inc

@@ -0,0 +1,272 @@
+#include <emmintrin.h>
+#include <smmintrin.h>
+#include <tmmintrin.h>
+
+static FORCE_INLINE_ATTR inline size_t utf8_range_ValidateUTF8Simd(
+    const char* data_original, const char* data, const char* end,
+    int return_position) {
+  /* This code checks that utf-8 ranges are structurally valid 16 bytes at once
+   * using superscalar instructions.
+   * The mapping between ranges of codepoint and their corresponding utf-8
+   * sequences is below.
+   */
+
+  /*
+   * U+0000...U+007F     00...7F
+   * U+0080...U+07FF     C2...DF 80...BF
+   * U+0800...U+0FFF     E0      A0...BF 80...BF
+   * U+1000...U+CFFF     E1...EC 80...BF 80...BF
+   * U+D000...U+D7FF     ED      80...9F 80...BF
+   * U+E000...U+FFFF     EE...EF 80...BF 80...BF
+   * U+10000...U+3FFFF   F0      90...BF 80...BF 80...BF
+   * U+40000...U+FFFFF   F1...F3 80...BF 80...BF 80...BF
+   * U+100000...U+10FFFF F4      80...8F 80...BF 80...BF
+   */
+
+  /* First we compute the type for each byte, as given by the table below.
+   * This type will be used as an index later on.
+   */
+
+  /*
+   * Index  Min Max Byte Type
+   *  0     00  7F  Single byte sequence
+   *  1,2,3 80  BF  Second, third and fourth byte for many of the sequences.
+   *  4     A0  BF  Second byte after E0
+   *  5     80  9F  Second byte after ED
+   *  6     90  BF  Second byte after F0
+   *  7     80  8F  Second byte after F4
+   *  8     C2  F4  First non ASCII byte
+   *  9..15 7F  80  Invalid byte
+   */
+
+  /* After the first step we compute the index for all bytes, then we permute
+     the bytes according to their indices to check the ranges from the range
+     table.
+   * The range for a given type can be found in the range_min_table and
+     range_max_table, the range for type/index X is in range_min_table[X] ...
+     range_max_table[X].
+   */
+
+  /* Algorithm:
+   * Put index zero to all bytes.
+   * Find all non ASCII characters, give them index 8.
+   * For each tail byte in a codepoint sequence, give it an index corresponding
+     to the 1 based index from the end.
+   * If the first byte of the codepoint is in the [C0...DF] range, we write
+     index 1 in the following byte.
+   * If the first byte of the codepoint is in the range [E0...EF], we write
+     indices 2 and 1 in the next two bytes.
+   * If the first byte of the codepoint is in the range [F0...FF] we write
+     indices 3,2,1 into the next three bytes.
+   * For finding the number of bytes we need to look at high nibbles (4 bits)
+     and do the lookup from the table, it can be done with shift by 4 + shuffle
+     instructions. We call it `first_len`.
+   * Then we shift first_len by 8 bits to get the indices of the 2nd bytes.
+   * Saturating sub 1 and shift by 8 bits to get the indices of the 3rd bytes.
+   * Again to get the indices of the 4th bytes.
+   * Take OR of all that 4 values and check within range.
+   */
+  /* For example:
+   * input       C3 80 68 E2 80 20 A6 F0 A0 80 AC 20 F0 93 80 80
+   * first_len   1  0  0  2  0  0  0  3  0  0  0  0  3  0  0  0
+   * 1st byte    8  0  0  8  0  0  0  8  0  0  0  0  8  0  0  0
+   * 2nd byte    0  1  0  0  2  0  0  0  3  0  0  0  0  3  0  0 // Shift + sub
+   * 3rd byte    0  0  0  0  0  1  0  0  0  2  0  0  0  0  2  0 // Shift + sub
+   * 4th byte    0  0  0  0  0  0  0  0  0  0  1  0  0  0  0  1 // Shift + sub
+   * Index       8  1  0  8  2  1  0  8  3  2  1  0  8  3  2  1 // OR of results
+   */
+
+  /* Checking for errors:
+   * Error checking is done by looking up the high nibble (4 bits) of each byte
+     against an error checking table.
+   * Because the lookup value for the second byte depends of the value of the
+     first byte in codepoint, we use saturated operations to adjust the index.
+   * Specifically we need to add 2 for E0, 3 for ED, 3 for F0 and 4 for F4 to
+     match the correct index.
+       * If we subtract from all bytes EF then EO -> 241, ED -> 254, F0 -> 1,
+         F4 -> 5
+       * Do saturating sub 240, then E0 -> 1, ED -> 14 and we can do lookup to
+         match the adjustment
+       * Add saturating 112, then F0 -> 113, F4 -> 117, all that were > 16 will
+         be more 128 and lookup in ef_fe_table will return 0 but for F0
+         and F4 it will be 4 and 5 accordingly
+   */
+  /*
+   * Then just check the appropriate ranges with greater/smaller equal
+     instructions. Check tail with a naive algorithm.
+   * To save from previous 16 byte checks we just align previous_first_len to
+     get correct continuations of the codepoints.
+   */
+
+  /*
+   * Map high nibble of "First Byte" to legal character length minus 1
+   * 0x00 ~ 0xBF --> 0
+   * 0xC0 ~ 0xDF --> 1
+   * 0xE0 ~ 0xEF --> 2
+   * 0xF0 ~ 0xFF --> 3
+   */
+  const __m128i first_len_table =
+      _mm_setr_epi8(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 2, 3);
+
+  /* Map "First Byte" to 8-th item of range table (0xC2 ~ 0xF4) */
+  const __m128i first_range_table =
+      _mm_setr_epi8(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8);
+
+  /*
+   * Range table, map range index to min and max values
+   */
+  const __m128i range_min_table =
+      _mm_setr_epi8(0x00, 0x80, 0x80, 0x80, 0xA0, 0x80, 0x90, 0x80, 0xC2, 0x7F,
+                    0x7F, 0x7F, 0x7F, 0x7F, 0x7F, 0x7F);
+
+  const __m128i range_max_table =
+      _mm_setr_epi8(0x7F, 0xBF, 0xBF, 0xBF, 0xBF, 0x9F, 0xBF, 0x8F, 0xF4, 0x80,
+                    0x80, 0x80, 0x80, 0x80, 0x80, 0x80);
+
+  /*
+   * Tables for fast handling of four special First Bytes(E0,ED,F0,F4), after
+   * which the Second Byte are not 80~BF. It contains "range index adjustment".
+   * +------------+---------------+------------------+----------------+
+   * | First Byte | original range| range adjustment | adjusted range |
+   * +------------+---------------+------------------+----------------+
+   * | E0         | 2             | 2                | 4              |
+   * +------------+---------------+------------------+----------------+
+   * | ED         | 2             | 3                | 5              |
+   * +------------+---------------+------------------+----------------+
+   * | F0         | 3             | 3                | 6              |
+   * +------------+---------------+------------------+----------------+
+   * | F4         | 4             | 4                | 8              |
+   * +------------+---------------+------------------+----------------+
+   */
+
+  /* df_ee_table[1] -> E0, df_ee_table[14] -> ED as ED - E0 = 13 */
+  // The values represent the adjustment in the Range Index table for a correct
+  // index.
+  const __m128i df_ee_table =
+      _mm_setr_epi8(0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0);
+
+  /* ef_fe_table[1] -> F0, ef_fe_table[5] -> F4, F4 - F0 = 4 */
+  // The values represent the adjustment in the Range Index table for a correct
+  // index.
+  const __m128i ef_fe_table =
+      _mm_setr_epi8(0, 3, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+
+  __m128i prev_input = _mm_set1_epi8(0);
+  __m128i prev_first_len = _mm_set1_epi8(0);
+  __m128i error = _mm_set1_epi8(0);
+
+  while (end - data >= 16) {
+    const __m128i input = _mm_loadu_si128((const __m128i*)(data));
+
+    /* high_nibbles = input >> 4 */
+    const __m128i high_nibbles =
+        _mm_and_si128(_mm_srli_epi16(input, 4), _mm_set1_epi8(0x0F));
+
+    /* first_len = legal character length minus 1 */
+    /* 0 for 00~7F, 1 for C0~DF, 2 for E0~EF, 3 for F0~FF */
+    /* first_len = first_len_table[high_nibbles] */
+    __m128i first_len = _mm_shuffle_epi8(first_len_table, high_nibbles);
+
+    /* First Byte: set range index to 8 for bytes within 0xC0 ~ 0xFF */
+    /* range = first_range_table[high_nibbles] */
+    __m128i range = _mm_shuffle_epi8(first_range_table, high_nibbles);
+
+    /* Second Byte: set range index to first_len */
+    /* 0 for 00~7F, 1 for C0~DF, 2 for E0~EF, 3 for F0~FF */
+    /* range |= (first_len, prev_first_len) << 1 byte */
+    range = _mm_or_si128(range, _mm_alignr_epi8(first_len, prev_first_len, 15));
+
+    /* Third Byte: set range index to saturate_sub(first_len, 1) */
+    /* 0 for 00~7F, 0 for C0~DF, 1 for E0~EF, 2 for F0~FF */
+    __m128i tmp1;
+    __m128i tmp2;
+    /* tmp1 = saturate_sub(first_len, 1) */
+    tmp1 = _mm_subs_epu8(first_len, _mm_set1_epi8(1));
+    /* tmp2 = saturate_sub(prev_first_len, 1) */
+    tmp2 = _mm_subs_epu8(prev_first_len, _mm_set1_epi8(1));
+    /* range |= (tmp1, tmp2) << 2 bytes */
+    range = _mm_or_si128(range, _mm_alignr_epi8(tmp1, tmp2, 14));
+
+    /* Fourth Byte: set range index to saturate_sub(first_len, 2) */
+    /* 0 for 00~7F, 0 for C0~DF, 0 for E0~EF, 1 for F0~FF */
+    /* tmp1 = saturate_sub(first_len, 2) */
+    tmp1 = _mm_subs_epu8(first_len, _mm_set1_epi8(2));
+    /* tmp2 = saturate_sub(prev_first_len, 2) */
+    tmp2 = _mm_subs_epu8(prev_first_len, _mm_set1_epi8(2));
+    /* range |= (tmp1, tmp2) << 3 bytes */
+    range = _mm_or_si128(range, _mm_alignr_epi8(tmp1, tmp2, 13));
+
+    /*
+     * Now we have below range indices calculated
+     * Correct cases:
+     * - 8 for C0~FF
+     * - 3 for 1st byte after F0~FF
+     * - 2 for 1st byte after E0~EF or 2nd byte after F0~FF
+     * - 1 for 1st byte after C0~DF or 2nd byte after E0~EF or
+     *         3rd byte after F0~FF
+     * - 0 for others
+     * Error cases:
+     *   >9 for non ascii First Byte overlapping
+     *   E.g., F1 80 C2 90 --> 8 3 10 2, where 10 indicates error
+     */
+
+    /* Adjust Second Byte range for special First Bytes(E0,ED,F0,F4) */
+    /* Overlaps lead to index 9~15, which are illegal in range table */
+    __m128i shift1;
+    __m128i pos;
+    __m128i range2;
+    /* shift1 = (input, prev_input) << 1 byte */
+    shift1 = _mm_alignr_epi8(input, prev_input, 15);
+    pos = _mm_sub_epi8(shift1, _mm_set1_epi8(0xEF));
+    /*
+     * shift1:  | EF  F0 ... FE | FF  00  ... ...  DE | DF  E0 ... EE |
+     * pos:     | 0   1      15 | 16  17           239| 240 241    255|
+     * pos-240: | 0   0      0  | 0   0            0  | 0   1      15 |
+     * pos+112: | 112 113    127|       >= 128        |     >= 128    |
+     */
+    tmp1 = _mm_subs_epu8(pos, _mm_set1_epi8(-16));
+    range2 = _mm_shuffle_epi8(df_ee_table, tmp1);
+    tmp2 = _mm_adds_epu8(pos, _mm_set1_epi8(112));
+    range2 = _mm_add_epi8(range2, _mm_shuffle_epi8(ef_fe_table, tmp2));
+
+    range = _mm_add_epi8(range, range2);
+
+    /* Load min and max values per calculated range index */
+    __m128i min_range = _mm_shuffle_epi8(range_min_table, range);
+    __m128i max_range = _mm_shuffle_epi8(range_max_table, range);
+
+    /* Check value range */
+    if (return_position) {
+      error = _mm_cmplt_epi8(input, min_range);
+      error = _mm_or_si128(error, _mm_cmpgt_epi8(input, max_range));
+      /* 5% performance drop from this conditional branch */
+      if (!_mm_testz_si128(error, error)) {
+        break;
+      }
+    } else {
+      error = _mm_or_si128(error, _mm_cmplt_epi8(input, min_range));
+      error = _mm_or_si128(error, _mm_cmpgt_epi8(input, max_range));
+    }
+
+    prev_input = input;
+    prev_first_len = first_len;
+
+    data += 16;
+  }
+  /* If we got to the end, we don't need to skip any bytes backwards */
+  if (return_position && data == data_original) {
+    return utf8_range_ValidateUTF8Naive(data, end, return_position);
+  }
+  /* Find previous codepoint (not 80~BF) */
+  data -= utf8_range_CodepointSkipBackwards(_mm_extract_epi32(prev_input, 3));
+  if (return_position) {
+    return (data - data_original) +
+           utf8_range_ValidateUTF8Naive(data, end, return_position);
+  }
+  /* Test if there was any error */
+  if (!_mm_testz_si128(error, error)) {
+    return 0;
+  }
+  /* Check the tail */
+  return utf8_range_ValidateUTF8Naive(data, end, return_position);
+}

data/third_party/zlib/deflate.c

@@ -1,5 +1,5 @@
 /* deflate.c -- compress data using the deflation algorithm
- * Copyright (C) 1995-2023 Jean-loup Gailly and Mark Adler
+ * Copyright (C) 1995-2024 Jean-loup Gailly and Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -52,7 +52,7 @@
 #include "deflate.h"
 
 const char deflate_copyright[] =
-   " deflate 1.3 Copyright 1995-2023 Jean-loup Gailly and Mark Adler ";
+   " deflate 1.3.1.1 Copyright 1995-2024 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
@@ -493,7 +493,7 @@ int ZEXPORT deflateInit2_(z_streamp strm, int level, int method,
      * symbols from which it is being constructed.
      */
 
-    s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, 4);
+    s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, LIT_BUFS);
     s->pending_buf_size = (ulg)s->lit_bufsize * 4;
 
     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
@@ -503,8 +503,14 @@ int ZEXPORT deflateInit2_(z_streamp strm, int level, int method,
         deflateEnd (strm);
         return Z_MEM_ERROR;
     }
+#ifdef LIT_MEM
+    s->d_buf = (ushf *)(s->pending_buf + (s->lit_bufsize << 1));
+    s->l_buf = s->pending_buf + (s->lit_bufsize << 2);
+    s->sym_end = s->lit_bufsize - 1;
+#else
     s->sym_buf = s->pending_buf + s->lit_bufsize;
     s->sym_end = (s->lit_bufsize - 1) * 3;
+#endif
     /* We avoid equality with lit_bufsize*3 because of wraparound at 64K
      * on 16 bit machines and because stored blocks are restricted to
      * 64K-1 bytes.
@@ -720,9 +726,15 @@ int ZEXPORT deflatePrime(z_streamp strm, int bits, int value) {
 
     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
     s = strm->state;
+#ifdef LIT_MEM
+    if (bits < 0 || bits > 16 ||
+        (uchf *)s->d_buf < s->pending_out + ((Buf_size + 7) >> 3))
+        return Z_BUF_ERROR;
+#else
     if (bits < 0 || bits > 16 ||
         s->sym_buf < s->pending_out + ((Buf_size + 7) >> 3))
         return Z_BUF_ERROR;
+#endif
     do {
         put = Buf_size - s->bi_valid;
         if (put > bits)
@@ -1294,7 +1306,7 @@ int ZEXPORT deflateCopy(z_streamp dest, z_streamp source) {
     ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
     ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
     ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
-    ds->pending_buf = (uchf *) ZALLOC(dest, ds->lit_bufsize, 4);
+    ds->pending_buf = (uchf *) ZALLOC(dest, ds->lit_bufsize, LIT_BUFS);
 
     if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
         ds->pending_buf == Z_NULL) {
@@ -1305,10 +1317,15 @@ int ZEXPORT deflateCopy(z_streamp dest, z_streamp source) {
     zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
     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);
+    zmemcpy(ds->pending_buf, ss->pending_buf, ds->lit_bufsize * LIT_BUFS);
 
     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
+#ifdef LIT_MEM
+    ds->d_buf = (ushf *)(ds->pending_buf + (ds->lit_bufsize << 1));
+    ds->l_buf = ds->pending_buf + (ds->lit_bufsize << 2);
+#else
     ds->sym_buf = ds->pending_buf + ds->lit_bufsize;
+#endif
 
     ds->l_desc.dyn_tree = ds->dyn_ltree;
     ds->d_desc.dyn_tree = ds->dyn_dtree;
@@ -1539,13 +1556,21 @@ local uInt longest_match(deflate_state *s, IPos cur_match) {
  */
 local void check_match(deflate_state *s, IPos start, IPos match, int length) {
     /* check that the match is indeed a match */
-    if (zmemcmp(s->window + match,
-                s->window + start, length) != EQUAL) {
-        fprintf(stderr, " start %u, match %u, length %d\n",
-                start, match, length);
+    Bytef *back = s->window + (int)match, *here = s->window + start;
+    IPos len = length;
+    if (match == (IPos)-1) {
+        /* match starts one byte before the current window -- just compare the
+           subsequent length-1 bytes */
+        back++;
+        here++;
+        len--;
+    }
+    if (zmemcmp(back, here, len) != EQUAL) {
+        fprintf(stderr, " start %u, match %d, length %d\n",
+                start, (int)match, length);
         do {
-            fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
-        } while (--length != 0);
+            fprintf(stderr, "(%02x %02x)", *back++, *here++);
+        } while (--len != 0);
         z_error("invalid match");
     }
     if (z_verbose > 1) {
@@ -1646,10 +1671,10 @@ local block_state deflate_stored(deflate_state *s, int flush) {
         _tr_stored_block(s, (char *)0, 0L, last);
 
         /* Replace the lengths in the dummy stored block with len. */
-        s->pending_buf[s->pending - 4] = len;
-        s->pending_buf[s->pending - 3] = len >> 8;
-        s->pending_buf[s->pending - 2] = ~len;
-        s->pending_buf[s->pending - 1] = ~len >> 8;
+        s->pending_buf[s->pending - 4] = (Bytef)len;
+        s->pending_buf[s->pending - 3] = (Bytef)(len >> 8);
+        s->pending_buf[s->pending - 2] = (Bytef)~len;
+        s->pending_buf[s->pending - 1] = (Bytef)(~len >> 8);
 
         /* Write the stored block header bytes. */
         flush_pending(s->strm);

data/third_party/zlib/deflate.h

@@ -1,5 +1,5 @@
 /* deflate.h -- internal compression state
- * Copyright (C) 1995-2018 Jean-loup Gailly
+ * Copyright (C) 1995-2024 Jean-loup Gailly
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -23,6 +23,10 @@
 #  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.
  */
@@ -217,7 +221,14 @@ typedef struct internal_state {
     /* Depth of each subtree used as tie breaker for trees of equal frequency
      */
 
+#ifdef LIT_MEM
+#   define LIT_BUFS 5
+    ushf *d_buf;          /* buffer for distances */
+    uchf *l_buf;          /* buffer for literals/lengths */
+#else
+#   define LIT_BUFS 4
     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
@@ -239,7 +250,7 @@ typedef struct internal_state {
      *   - I can't count above 4
      */
 
-    uInt sym_next;      /* running index in sym_buf */
+    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 */
@@ -318,6 +329,25 @@ void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
   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; \
@@ -337,6 +367,7 @@ void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
     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) \

data/third_party/zlib/gzguts.h

@@ -1,5 +1,5 @@
 /* gzguts.h -- zlib internal header definitions for gz* operations
- * Copyright (C) 2004-2019 Mark Adler
+ * Copyright (C) 2004-2024 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -210,9 +210,5 @@ char ZLIB_INTERNAL *gz_strwinerror(DWORD error);
 /* GT_OFF(x), where x is an unsigned value, is true if x > maximum z_off64_t
    value -- needed when comparing unsigned to z_off64_t, which is signed
    (possible z_off64_t types off_t, off64_t, and long are all signed) */
-#ifdef INT_MAX
-#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > INT_MAX)
-#else
 unsigned ZLIB_INTERNAL gz_intmax(void);
-#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > gz_intmax())
-#endif
+#define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > gz_intmax())

data/third_party/zlib/inflate.c

@@ -1387,7 +1387,7 @@ int ZEXPORT inflateSync(z_streamp strm) {
     /* if first time, start search in bit buffer */
     if (state->mode != SYNC) {
         state->mode = SYNC;
-        state->hold <<= state->bits & 7;
+        state->hold >>= state->bits & 7;
         state->bits -= state->bits & 7;
         len = 0;
         while (state->bits >= 8) {

data/third_party/zlib/inftrees.c

@@ -1,5 +1,5 @@
 /* inftrees.c -- generate Huffman trees for efficient decoding
- * Copyright (C) 1995-2023 Mark Adler
+ * Copyright (C) 1995-2024 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -9,7 +9,7 @@
 #define MAXBITS 15
 
 const char inflate_copyright[] =
-   " inflate 1.3 Copyright 1995-2023 Mark Adler ";
+   " inflate 1.3.1.1 Copyright 1995-2024 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
@@ -57,7 +57,7 @@ int ZLIB_INTERNAL inflate_table(codetype type, unsigned short FAR *lens,
         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
     static const unsigned short lext[31] = { /* Length codes 257..285 extra */
         16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
-        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 198, 203};
+        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 73, 200};
     static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,

data/third_party/zlib/inftrees.h

@@ -41,8 +41,8 @@ typedef struct {
    examples/enough.c found in the zlib distribution.  The arguments to that
    program are the number of symbols, the initial root table size, and the
    maximum bit length of a code.  "enough 286 9 15" for literal/length codes
-   returns returns 852, and "enough 30 6 15" for distance codes returns 592.
-   The initial root table size (9 or 6) is found in the fifth argument of the
+   returns 852, and "enough 30 6 15" for distance codes returns 592. The
+   initial root table size (9 or 6) is found in the fifth argument of the
    inflate_table() calls in inflate.c and infback.c.  If the root table size is
    changed, then these maximum sizes would be need to be recalculated and
    updated. */