diff --git a/README.md b/README.md
index 089dd1a..7e63a2a 100644
--- a/README.md
+++ b/README.md
@@ -775,6 +775,11 @@ sized deallocation sanity checks for C. A performance-oriented allocator could
 use the same API as an optimization to avoid a potential cache miss from
 reading the size from metadata.
 
+The `void free_aligned_sized(void *ptr, size_t alignment, size_t expected_size)`
+function is the aligned counterpart to `free_sized`, checking both the size and
+alignment passed at allocation time. It is used to implement C++17 sized
+`operator delete` for over-aligned types.
+
 The `size_t malloc_object_size(void *ptr)` function returns an *upper bound* on
 the accessible size of the relevant object (if any) by querying the malloc
 implementation. It's similar to the `__builtin_object_size` intrinsic used by
diff --git a/h_malloc.c b/h_malloc.c
index 88d50c9..070f5d1 100644
--- a/h_malloc.c
+++ b/h_malloc.c
@@ -1364,7 +1364,7 @@ static size_t get_large_size_class(size_t size) {
         // Continue small size class growth pattern of power of 2 spacing classes:
         //
         // 4 KiB [20 KiB, 24 KiB, 28 KiB, 32 KiB]
-        // 8 KiB [40 KiB, 48 KiB, 54 KiB, 64 KiB]
+        // 8 KiB [40 KiB, 48 KiB, 56 KiB, 64 KiB]
         // 16 KiB [80 KiB, 96 KiB, 112 KiB, 128 KiB]
         // 32 KiB [160 KiB, 192 KiB, 224 KiB, 256 KiB]
         // 512 KiB [2560 KiB, 3 MiB, 3584 KiB, 4 MiB]
@@ -1796,7 +1796,9 @@ EXPORT void h_free_aligned_sized(void *p, size_t alignment, size_t expected_size
         return;
     }
 
+    int saved_errno = errno;
     deallocate_large(p, &expected_size);
+    errno = saved_errno;
 
     thread_seal_metadata();
 }