php patches + optimized checkout

patches to specifically address a gcc -Werror=logical-op issue explained in BLAKE3-team/BLAKE3#380
and a gcc -Wunused-function issue explained in BLAKE3-team/BLAKE3#382
and optimized upstream git checkout to only fetch the files we want.

ext/hash/blake3/fetch_upstream_blake3.sh

@@ -2,12 +2,17 @@
 # afaik the PHP project doesn't allow git submodules, so we do this fetcher script instead.
 cd "$(dirname "$0")"
 rm -rf upstream_blake3
-# fancy way of just fetching the "c" folder (the only thing we want)
+# fancy way of fetching only the files we want
 git clone --branch '1.5.0' -n --depth=1 --filter=tree:0 'https://github.com/BLAKE3-team/BLAKE3.git' 'upstream_blake3'
 cd upstream_blake3
-git sparse-checkout set --no-cone c LICENSE
+git sparse-checkout set --no-cone LICENSE c/blake3.c  c/blake3.h  c/blake3_avx2.c  c/blake3_avx2_x86-64_unix.S \
+ c/blake3_avx2_x86-64_windows_gnu.S  c/blake3_avx2_x86-64_windows_msvc.asm  c/blake3_avx512.c \
+ c/blake3_avx512_x86-64_unix.S  c/blake3_avx512_x86-64_windows_gnu.S  c/blake3_avx512_x86-64_windows_msvc.asm \
+ c/blake3_dispatch.c  c/blake3_impl.h  c/blake3_neon.c  c/blake3_portable.c  c/blake3_sse2.c \
+ c/blake3_sse2_x86-64_unix.S  c/blake3_sse2_x86-64_windows_gnu.S  c/blake3_sse2_x86-64_windows_msvc.asm \
+ c/blake3_sse41.c  c/blake3_sse41_x86-64_unix.S  c/blake3_sse41_x86-64_windows_gnu.S \
+ c/blake3_sse41_x86-64_windows_msvc.asm
 git checkout
 rm -rf .git
-cd c
-# some stuff we don't need
-rm -rf blake3_c_rust_bindings test.py example.c main.c Makefile.testing CMakeLists.txt blake3-config.cmake.in README.md .gitignore libblake3.pc.in
+cd ..
+git apply patches.diff

ext/hash/blake3/patches.diff

@@ -0,0 +1,56 @@
+diff --git a/ext/hash/blake3/upstream_blake3/c/blake3.c b/ext/hash/blake3/upstream_blake3/c/blake3.c
+index 692f4b0216..3591ba245a 100644
+--- a/ext/hash/blake3/upstream_blake3/c/blake3.c
++++ b/ext/hash/blake3/upstream_blake3/c/blake3.c
+@@ -341,21 +341,23 @@ INLINE void compress_subtree_to_parent_node(
+   size_t num_cvs = blake3_compress_subtree_wide(input, input_len, key,
+                                                 chunk_counter, flags, cv_array);
+   assert(num_cvs <= MAX_SIMD_DEGREE_OR_2);
+-
+-  // If MAX_SIMD_DEGREE is greater than 2 and there's enough input,
+-  // compress_subtree_wide() returns more than 2 chaining values. Condense
+-  // them into 2 by forming parent nodes repeatedly.
+-  uint8_t out_array[MAX_SIMD_DEGREE_OR_2 * BLAKE3_OUT_LEN / 2];
+-  // The second half of this loop condition is always true, and we just
++  // https://github.com/BLAKE3-team/BLAKE3/pull/380
++  // This condition is always true, and we just
+   // asserted it above. But GCC can't tell that it's always true, and if NDEBUG
+   // is set on platforms where MAX_SIMD_DEGREE_OR_2 == 2, GCC emits spurious
+   // warnings here. GCC 8.5 is particularly sensitive, so if you're changing
+   // this code, test it against that version.
+-  while (num_cvs > 2 && num_cvs <= MAX_SIMD_DEGREE_OR_2) {
++#if MAX_SIMD_DEGREE_OR_2 > 2
++  // If MAX_SIMD_DEGREE_OR_2 is greater than 2 and there's enough input,
++  // compress_subtree_wide() returns more than 2 chaining values. Condense
++  // them into 2 by forming parent nodes repeatedly.
++  uint8_t out_array[MAX_SIMD_DEGREE_OR_2 * BLAKE3_OUT_LEN / 2];
++  while (num_cvs > 2) {
+     num_cvs =
+         compress_parents_parallel(cv_array, num_cvs, key, flags, out_array);
+     memcpy(cv_array, out_array, num_cvs * BLAKE3_OUT_LEN);
+   }
++#endif
+   memcpy(out, cv_array, 2 * BLAKE3_OUT_LEN);
+ }
+
+diff --git a/ext/hash/blake3/upstream_blake3/c/blake3_dispatch.c b/ext/hash/blake3/upstream_blake3/c/blake3_dispatch.c
+index af6c3dadc7..af3bf17bbe 100644
+--- a/ext/hash/blake3/upstream_blake3/c/blake3_dispatch.c
++++ b/ext/hash/blake3/upstream_blake3/c/blake3_dispatch.c
+@@ -86,7 +86,6 @@ static void cpuidex(uint32_t out[4], uint32_t id, uint32_t sid) {
+ #endif
+ }
+
+-#endif
+
+ enum cpu_feature {
+   SSE2 = 1 << 0,
+@@ -161,6 +160,8 @@ static
+ #endif
+   }
+ }
++// https://github.com/BLAKE3-team/BLAKE3/pull/382
++#endif
+
+ void blake3_compress_in_place(uint32_t cv[8],
+                               const uint8_t block[BLAKE3_BLOCK_LEN],

ext/hash/blake3/upstream_blake3/c/blake3.c

@@ -341,21 +341,23 @@ INLINE void compress_subtree_to_parent_node(
   size_t num_cvs = blake3_compress_subtree_wide(input, input_len, key,
                                                 chunk_counter, flags, cv_array);
   assert(num_cvs <= MAX_SIMD_DEGREE_OR_2);
-
-  // If MAX_SIMD_DEGREE is greater than 2 and there's enough input,
-  // compress_subtree_wide() returns more than 2 chaining values. Condense
-  // them into 2 by forming parent nodes repeatedly.
-  uint8_t out_array[MAX_SIMD_DEGREE_OR_2 * BLAKE3_OUT_LEN / 2];
-  // The second half of this loop condition is always true, and we just
+  // https://github.com/BLAKE3-team/BLAKE3/pull/380
+  // This condition is always true, and we just
   // asserted it above. But GCC can't tell that it's always true, and if NDEBUG
   // is set on platforms where MAX_SIMD_DEGREE_OR_2 == 2, GCC emits spurious
   // warnings here. GCC 8.5 is particularly sensitive, so if you're changing
   // this code, test it against that version.
-  while (num_cvs > 2 && num_cvs <= MAX_SIMD_DEGREE_OR_2) {
+#if MAX_SIMD_DEGREE_OR_2 > 2
+  // If MAX_SIMD_DEGREE_OR_2 is greater than 2 and there's enough input,
+  // compress_subtree_wide() returns more than 2 chaining values. Condense
+  // them into 2 by forming parent nodes repeatedly.
+  uint8_t out_array[MAX_SIMD_DEGREE_OR_2 * BLAKE3_OUT_LEN / 2];
+  while (num_cvs > 2) {
     num_cvs =
         compress_parents_parallel(cv_array, num_cvs, key, flags, out_array);
     memcpy(cv_array, out_array, num_cvs * BLAKE3_OUT_LEN);
   }
+#endif
   memcpy(out, cv_array, 2 * BLAKE3_OUT_LEN);
 }
 

ext/hash/blake3/upstream_blake3/c/blake3_dispatch.c

@@ -86,7 +86,6 @@ static void cpuidex(uint32_t out[4], uint32_t id, uint32_t sid) {
 #endif
 }
 
-#endif
 
 enum cpu_feature {
   SSE2 = 1 << 0,
@@ -161,6 +160,8 @@ static
 #endif
   }
 }
+// https://github.com/BLAKE3-team/BLAKE3/pull/382
+#endif
 
 void blake3_compress_in_place(uint32_t cv[8],
                               const uint8_t block[BLAKE3_BLOCK_LEN],