update bundled libdeflate to v1.11

Signed-off-by: Ivailo Monev <xakepa10@gmail.com>

src/3rdparty/libdeflate/NOTE

@@ -1,2 +1,2 @@
-This is Git checkout 3cc3608e9c340e4996dff3d0633acf2ec537e12a
+This is Git checkout b01537448e8eaf0803e38bdba5acef1d1c8effba
 from https://github.com/ebiggers/libdeflate that has not been modified.

src/3rdparty/libdeflate/common/compiler_gcc.h

@@ -122,15 +122,30 @@
 #    endif
 
      /*
-      * Determine whether CRC32 intrinsics are supported.
+      * Determine whether ARM CRC32 intrinsics are supported.
       *
-      * With gcc r274827 or later (gcc 10.1+, 9.3+, or 8.4+), or with clang,
-      * they work as expected.  (Well, not quite.  There's still a bug, but we
-      * have to work around it later when including arm_acle.h.)
+      * This support has been affected by several gcc bugs, which we must avoid
+      * by only allowing gcc versions that have the corresponding fixes.  First,
+      * gcc commit 943766d37ae4 ("[arm] Fix use of CRC32 intrinsics with Armv8-a
+      * and hard-float"), i.e. gcc 8.4+, 9.3+, 10.1+, or 11+, is needed.
+      * Second, gcc commit c1cdabe3aab8 ("arm: reorder assembler architecture
+      * directives [PR101723]"), i.e. gcc 9.5+, 10.4+, 11.3+, or 12+, is needed
+      * when binutils is 2.34 or later, due to
+      * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104439.  We use the second
+      * set of prerequisites, as they are stricter and we have no way to detect
+      * the binutils version in C source without requiring a configure script.
+      *
+      * Yet another gcc bug makes arm_acle.h sometimes not define the crc
+      * functions even when the corresponding builtins are available.  However,
+      * we work around this later when including arm_acle.h.
+      *
+      * Things are a bit easier with clang -- we can just check whether the
+      * crc builtins are available.  However, clang's arm_acle.h is broken in
+      * the same way as gcc's, which we work around later in the same way.
       */
-#    if GCC_PREREQ(10, 1) || \
-        (GCC_PREREQ(9, 3) && !GCC_PREREQ(10, 0)) || \
-        (GCC_PREREQ(8, 4) && !GCC_PREREQ(9, 0)) || \
+#    if GCC_PREREQ(11, 3) || \
+        (GCC_PREREQ(10, 4) && !GCC_PREREQ(11, 0)) || \
+        (GCC_PREREQ(9, 5) && !GCC_PREREQ(10, 0)) || \
         (defined(__clang__) && __has_builtin(__builtin_arm_crc32b))
 #      define COMPILER_SUPPORTS_CRC32_TARGET_INTRINSICS 1
 #    endif
@@ -199,3 +214,22 @@ typedef char  __v64qi __attribute__((__vector_size__(64)));
 #define bsr64(n)	(63 - __builtin_clzll(n))
 #define bsf32(n)	__builtin_ctz(n)
 #define bsf64(n)	__builtin_ctzll(n)
+
+#if defined(__arm__) && \
+	(__ARM_ARCH >= 7 || (__ARM_ARCH == 6 && defined(__ARM_ARCH_6T2__)))
+static forceinline unsigned int
+rbit32(unsigned int v)
+{
+	__asm__("rbit %0, %1\n" : "=r" (v) : "r" (v));
+	return v;
+}
+#define rbit32 rbit32
+#elif defined(__aarch64__)
+static forceinline unsigned int
+rbit32(unsigned int v)
+{
+	__asm__("rbit %w0, %w1\n" : "=r" (v) : "r" (v));
+	return v;
+}
+#define rbit32 rbit32
+#endif /* __aarch64__ */

src/3rdparty/libdeflate/lib/arm/cpu_features.c

@@ -26,13 +26,8 @@
  */
 
 /*
- * ARM processors don't have a standard way for unprivileged programs to detect
- * processor features.  But, on Linux we can read the AT_HWCAP and AT_HWCAP2
- * values from /proc/self/auxv.
- *
- * Ideally we'd use the C library function getauxval(), but it's not guaranteed
- * to be available: it was only added to glibc in 2.16, and in Android it was
- * added to API level 18 for ARM and level 21 for AArch64.
+ * ARM CPUs don't have a standard way for unprivileged programs to detect CPU
+ * features.  But an OS-specific way can be used when available.
  */
 
 #include "../cpu_features_common.h" /* must be included first */
@@ -40,6 +35,16 @@
 
 #if ARM_CPU_FEATURES_ENABLED
 
+#ifdef __linux__
+/*
+ * On Linux, arm32 and arm64 CPU features can be detected by reading the
+ * AT_HWCAP and AT_HWCAP2 values from /proc/self/auxv.
+ *
+ * Ideally we'd use the C library function getauxval(), but it's not guaranteed
+ * to be available: it was only added to glibc in 2.16, and in Android it was
+ * added to API level 18 for arm32 and level 21 for arm64.
+ */
+
 #include <errno.h>
 #include <fcntl.h>
 #include <string.h>
@@ -48,8 +53,6 @@
 #define AT_HWCAP	16
 #define AT_HWCAP2	26
 
-volatile u32 _cpu_features = 0;
-
 static void scan_auxv(unsigned long *hwcap, unsigned long *hwcap2)
 {
 	int fd;
@@ -92,13 +95,7 @@ out:
 	close(fd);
 }
 
-static const struct cpu_feature arm_cpu_feature_table[] = {
-	{ARM_CPU_FEATURE_NEON,		"neon"},
-	{ARM_CPU_FEATURE_PMULL,		"pmull"},
-	{ARM_CPU_FEATURE_CRC32,		"crc32"},
-};
-
-void setup_cpu_features(void)
+static u32 get_arm_cpu_features(void)
 {
 	u32 features = 0;
 	unsigned long hwcap = 0;
@@ -123,6 +120,56 @@ void setup_cpu_features(void)
 	if (hwcap & (1 << 7))	/* HWCAP_CRC32 */
 		features |= ARM_CPU_FEATURE_CRC32;
 #endif
+	return features;
+}
+
+#elif defined(__APPLE__)
+/* On Apple platforms, arm64 CPU features can be detected via sysctlbyname(). */
+
+#include <sys/types.h>
+#include <sys/sysctl.h>
+
+static const struct {
+	const char *name;
+	u32 feature;
+} feature_sysctls[] = {
+	{ "hw.optional.neon",		ARM_CPU_FEATURE_NEON },
+	{ "hw.optional.AdvSIMD",	ARM_CPU_FEATURE_NEON },
+	{ "hw.optional.arm.FEAT_PMULL",	ARM_CPU_FEATURE_PMULL },
+	{ "hw.optional.armv8_crc32",	ARM_CPU_FEATURE_CRC32 },
+};
+
+static u32 get_arm_cpu_features(void)
+{
+	u32 features = 0;
+	size_t i;
+
+	for (i = 0; i < ARRAY_LEN(feature_sysctls); i++) {
+		const char *name = feature_sysctls[i].name;
+		u32 val = 0;
+		size_t valsize = sizeof(val);
+
+		if (sysctlbyname(name, &val, &valsize, NULL, 0) == 0 &&
+		    valsize == sizeof(val) && val == 1)
+			features |= feature_sysctls[i].feature;
+	}
+	return features;
+}
+#else
+#error "unhandled case"
+#endif
+
+static const struct cpu_feature arm_cpu_feature_table[] = {
+	{ARM_CPU_FEATURE_NEON,		"neon"},
+	{ARM_CPU_FEATURE_PMULL,		"pmull"},
+	{ARM_CPU_FEATURE_CRC32,		"crc32"},
+};
+
+volatile u32 _cpu_features = 0;
+
+void setup_cpu_features(void)
+{
+	u32 features = get_arm_cpu_features();
 
 	disable_cpu_features_for_testing(&features, arm_cpu_feature_table,
 					 ARRAY_LEN(arm_cpu_feature_table));

src/3rdparty/libdeflate/lib/arm/cpu_features.h

@@ -8,7 +8,8 @@
 #include "../lib_common.h"
 
 #if (defined(__arm__) || defined(__aarch64__)) && \
-	defined(__linux__) && \
+	(defined(__linux__) || \
+	 (defined(__aarch64__) && defined(__APPLE__))) && \
 	COMPILER_SUPPORTS_TARGET_FUNCTION_ATTRIBUTE && \
 	!defined(FREESTANDING)
 #  define ARM_CPU_FEATURES_ENABLED 1

src/3rdparty/libdeflate/lib/arm/crc32_impl.h

@@ -80,13 +80,17 @@ crc32_arm(u32 remainder, const u8 *p, size_t size)
 		size--;
 	}
 
-	while (size >= 32) {
+	while (size >= 64) {
 		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 0)));
 		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 1)));
 		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 2)));
 		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 3)));
-		p += 32;
-		size -= 32;
+		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 4)));
+		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 5)));
+		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 6)));
+		remainder = __crc32d(remainder, le64_bswap(*((u64 *)p + 7)));
+		p += 64;
+		size -= 64;
 	}
 
 	while (size >= 8) {

src/3rdparty/libdeflate/lib/bt_matchfinder.h

@@ -139,14 +139,14 @@ bt_right_child(struct bt_matchfinder *mf, s32 node)
 /* Advance the binary tree matchfinder by one byte, optionally recording
  * matches.  @record_matches should be a compile-time constant.  */
 static forceinline struct lz_match *
-bt_matchfinder_advance_one_byte(struct bt_matchfinder * const restrict mf,
-				const u8 * const restrict in_base,
+bt_matchfinder_advance_one_byte(struct bt_matchfinder * const mf,
+				const u8 * const in_base,
 				const ptrdiff_t cur_pos,
 				const u32 max_len,
 				const u32 nice_len,
 				const u32 max_search_depth,
-				u32 * const restrict next_hashes,
-				struct lz_match * restrict lz_matchptr,
+				u32 * const next_hashes,
+				struct lz_match *lz_matchptr,
 				const bool record_matches)
 {
 	const u8 *in_next = in_base + cur_pos;

src/3rdparty/libdeflate/lib/deflate_decompress.c

@@ -313,7 +313,7 @@ do {									\
 
 /*
  * Read a 16-bit value from the input.  This must have been preceded by a call
- * to ALIGN_INPUT(), and the caller must have already checked for overrun.
+ * to ALIGN_INPUT(), and the caller must have already checked for overread.
  */
 #define READ_U16() (tmp16 = get_unaligned_le16(in_next), in_next += 2, tmp16)
 
@@ -886,9 +886,8 @@ copy_word_unaligned(const void *src, void *dst)
  *****************************************************************************/
 
 typedef enum libdeflate_result (*decompress_func_t)
-	(struct libdeflate_decompressor * restrict d,
-	 const void * restrict in, size_t in_nbytes,
-	 void * restrict out, size_t out_nbytes_avail,
+	(struct libdeflate_decompressor *d,
+	 const void *in, size_t in_nbytes, void *out, size_t out_nbytes_avail,
 	 size_t *actual_in_nbytes_ret, size_t *actual_out_nbytes_ret);
 
 #undef DEFAULT_IMPL
@@ -906,18 +905,16 @@ typedef enum libdeflate_result (*decompress_func_t)
 
 #ifdef DISPATCH
 static enum libdeflate_result
-dispatch(struct libdeflate_decompressor * restrict d,
-	 const void * restrict in, size_t in_nbytes,
-	 void * restrict out, size_t out_nbytes_avail,
+dispatch(struct libdeflate_decompressor *d,
+	 const void *in, size_t in_nbytes, void *out, size_t out_nbytes_avail,
 	 size_t *actual_in_nbytes_ret, size_t *actual_out_nbytes_ret);
 
 static volatile decompress_func_t decompress_impl = dispatch;
 
 /* Choose the fastest implementation at runtime */
 static enum libdeflate_result
-dispatch(struct libdeflate_decompressor * restrict d,
-	 const void * restrict in, size_t in_nbytes,
-	 void * restrict out, size_t out_nbytes_avail,
+dispatch(struct libdeflate_decompressor *d,
+	 const void *in, size_t in_nbytes, void *out, size_t out_nbytes_avail,
 	 size_t *actual_in_nbytes_ret, size_t *actual_out_nbytes_ret)
 {
 	decompress_func_t f = arch_select_decompress_func();
@@ -943,9 +940,9 @@ dispatch(struct libdeflate_decompressor * restrict d,
  * at runtime.
  */
 LIBDEFLATEEXPORT enum libdeflate_result LIBDEFLATEAPI
-libdeflate_deflate_decompress_ex(struct libdeflate_decompressor * restrict d,
-				 const void * restrict in, size_t in_nbytes,
-				 void * restrict out, size_t out_nbytes_avail,
+libdeflate_deflate_decompress_ex(struct libdeflate_decompressor *d,
+				 const void *in, size_t in_nbytes,
+				 void *out, size_t out_nbytes_avail,
 				 size_t *actual_in_nbytes_ret,
 				 size_t *actual_out_nbytes_ret)
 {
@@ -954,9 +951,9 @@ libdeflate_deflate_decompress_ex(struct libdeflate_decompressor * restrict d,
 }
 
 LIBDEFLATEEXPORT enum libdeflate_result LIBDEFLATEAPI
-libdeflate_deflate_decompress(struct libdeflate_decompressor * restrict d,
-			      const void * restrict in, size_t in_nbytes,
-			      void * restrict out, size_t out_nbytes_avail,
+libdeflate_deflate_decompress(struct libdeflate_decompressor *d,
+			      const void *in, size_t in_nbytes,
+			      void *out, size_t out_nbytes_avail,
 			      size_t *actual_out_nbytes_ret)
 {
 	return libdeflate_deflate_decompress_ex(d, in, in_nbytes,

src/3rdparty/libdeflate/lib/hc_matchfinder.h

@@ -181,15 +181,15 @@ hc_matchfinder_slide_window(struct hc_matchfinder *mf)
  * 'best_len' was found.
  */
 static forceinline u32
-hc_matchfinder_longest_match(struct hc_matchfinder * const restrict mf,
-			     const u8 ** const restrict in_base_p,
-			     const u8 * const restrict in_next,
+hc_matchfinder_longest_match(struct hc_matchfinder * const mf,
+			     const u8 ** const in_base_p,
+			     const u8 * const in_next,
 			     u32 best_len,
 			     const u32 max_len,
 			     const u32 nice_len,
 			     const u32 max_search_depth,
-			     u32 * const restrict next_hashes,
-			     u32 * const restrict offset_ret)
+			     u32 * const next_hashes,
+			     u32 * const offset_ret)
 {
 	u32 depth_remaining = max_search_depth;
 	const u8 *best_matchptr = in_next;
@@ -359,12 +359,12 @@ out:
  *	the sequence beginning at @in_next + @count.
  */
 static forceinline void
-hc_matchfinder_skip_bytes(struct hc_matchfinder * const restrict mf,
-			  const u8 ** const restrict in_base_p,
+hc_matchfinder_skip_bytes(struct hc_matchfinder * const mf,
+			  const u8 ** const in_base_p,
 			  const u8 *in_next,
 			  const u8 * const in_end,
 			  const u32 count,
-			  u32 * const restrict next_hashes)
+			  u32 * const next_hashes)
 {
 	u32 cur_pos;
 	u32 hash3, hash4;

src/3rdparty/libdeflate/lib/ht_matchfinder.h

@@ -75,13 +75,13 @@ ht_matchfinder_slide_window(struct ht_matchfinder *mf)
 
 /* Note: max_len must be >= HT_MATCHFINDER_REQUIRED_NBYTES */
 static forceinline u32
-ht_matchfinder_longest_match(struct ht_matchfinder * const restrict mf,
-			     const u8 ** const restrict in_base_p,
-			     const u8 * const restrict in_next,
+ht_matchfinder_longest_match(struct ht_matchfinder * const mf,
+			     const u8 ** const in_base_p,
+			     const u8 * const in_next,
 			     const u32 max_len,
 			     const u32 nice_len,
-			     u32 * const restrict next_hash,
-			     u32 * const restrict offset_ret)
+			     u32 * const next_hash,
+			     u32 * const offset_ret)
 {
 	u32 best_len = 0;
 	const u8 *best_matchptr = in_next;
@@ -195,12 +195,12 @@ out:
 }
 
 static forceinline void
-ht_matchfinder_skip_bytes(struct ht_matchfinder * const restrict mf,
-			  const u8 ** const restrict in_base_p,
+ht_matchfinder_skip_bytes(struct ht_matchfinder * const mf,
+			  const u8 ** const in_base_p,
 			  const u8 *in_next,
 			  const u8 * const in_end,
 			  const u32 count,
-			  u32 * const restrict next_hash)
+			  u32 * const next_hash)
 {
 	s32 cur_pos = in_next - *in_base_p;
 	u32 hash;

src/3rdparty/libdeflate/lib/lib_common.h

@@ -60,8 +60,22 @@ void *memmove(void *dest, const void *src, size_t n);
 
 int memcmp(const void *s1, const void *s2, size_t n);
 #define memcmp(s1, s2, n)	__builtin_memcmp((s1), (s2), (n))
+
+#undef LIBDEFLATE_ENABLE_ASSERTIONS
 #else
 #include <string.h>
 #endif
 
+/*
+ * Runtime assertion support.  Don't enable this in production builds; it may
+ * hurt performance significantly.
+ */
+#ifdef LIBDEFLATE_ENABLE_ASSERTIONS
+void libdeflate_assertion_failed(const char *expr, const char *file, int line);
+#define ASSERT(expr) { if (unlikely(!(expr))) \
+	libdeflate_assertion_failed(#expr, __FILE__, __LINE__); }
+#else
+#define ASSERT(expr) (void)(expr)
+#endif
+
 #endif /* LIB_LIB_COMMON_H */

src/3rdparty/libdeflate/lib/utils.c

@@ -140,3 +140,14 @@ memcmp(const void *s1, const void *s2, size_t n)
 	return 0;
 }
 #endif /* FREESTANDING */
+
+#ifdef LIBDEFLATE_ENABLE_ASSERTIONS
+#include <stdio.h>
+#include <stdlib.h>
+void
+libdeflate_assertion_failed(const char *expr, const char *file, int line)
+{
+	fprintf(stderr, "Assertion failed: %s at %s:%d\n", expr, file, line);
+	abort();
+}
+#endif /* LIBDEFLATE_ENABLE_ASSERTIONS */

src/3rdparty/libdeflate/libdeflate.h

@@ -10,8 +10,8 @@ extern "C" {
 #endif
 
 #define LIBDEFLATE_VERSION_MAJOR	1
-#define LIBDEFLATE_VERSION_MINOR	9
-#define LIBDEFLATE_VERSION_STRING	"1.9"
+#define LIBDEFLATE_VERSION_MINOR	11
+#define LIBDEFLATE_VERSION_STRING	"1.11"
 
 #include <stddef.h>
 #include <stdint.h>
@@ -31,7 +31,7 @@ extern "C" {
 #  define LIBDEFLATEEXPORT
 #endif
 
-#if defined(_WIN32) && !defined(_WIN64)
+#if defined(_WIN32) && !defined(_WIN64) && defined(LIBDEFLATE_DLL)
 #  define LIBDEFLATEAPI_ABI	__stdcall
 #else
 #  define LIBDEFLATEAPI_ABI