diff --git a/utf8.c b/utf8.c
index 2b89fae4314e..5dbee8d4047d 100644
--- a/utf8.c
+++ b/utf8.c
@@ -3269,29 +3269,19 @@ They differ in that C<bytes_to_utf8_free_me> takes an extra parameter
 C<free_me>.  If that parameter is NULL, this function behaves identically to
 C<bytes_to_utf8>.
 
-But if not NULL, the function skips allocating new memory if the input string
-already is C<NUL>-terminated, and its UTF-8 representation is the same as its
-native representation.  In other words it returns the input string if
-converting the string would be a no-op.  It sets C<*free_me> to NULL in that
-case.  Otherwise C<*free_me> is set to the address of the newly allocated
-memory.  Note that in both cases, you can pass that result to C<L</Safefree>>
-and it will do the right thing.  
+But if not NULL, the function skips allocating new memory if the UTF-8
+representation of the input string is the same as its native representation.
+In other words it returns the input string if converting the string would be a
+no-op.  It sets C<*free_me> to NULL in that case.  Otherwise C<*free_me> is set
+to the address of the newly allocated memory.  Note that in both cases, you can
+pass that result to C<L</Safefree>> and it will do the right thing.  
+
+If new memory is allocated, it will be C<NUL>-terminated.
 
 Note that when new memory is allocated, the caller is responsible for arranging
 for that memory to get freed.  (This is transparent to the caller if
 C<Safefree> is called with C<free_me>.)
 
-The two forms have subtle differences in trailing C<NUL> handling.
-C<bytes_to_utf8> does not look for any trailing C<NUL>.  Instead it
-overallocates space for the copy by 1 byte and adds a C<NUL> to the end of it.
-If the input C<*lenp> included a C<NUL>, there will be two trailing C<NUL>
-characters; otherwise just 1.  There is code that depends on this behavior.
-C<bytes_to_utf8_free_me> has to look for a trailing C<NUL> in order to be able
-to guarantee that the result has one if no copy is made.  If the input C<*lenp>
-doesn't include any C<NUL> character at the end, this form has to assume that
-one doesn't exist, and will create a copy.  Only if C<*lenp> does include the
-C<NUL>, does this form check to see if a copy can be avoided.
-
 Upon return, the number of variants in the string can be computed by
 having saved the value of C<*lenp> before the call, and subtracting it from the
 after-call value of C<*lenp>.
@@ -3310,15 +3300,13 @@ Perl_bytes_to_utf8_free_me(pTHX_ const U8 *s, Size_t *lenp,
     PERL_ARGS_ASSERT_BYTES_TO_UTF8_FREE_ME;
     PERL_UNUSED_CONTEXT;
 
-    const Size_t len = *lenp;
-    const U8 * const send = s + len;
+    const U8 * const send = s + (*lenp);
     const Size_t variant_count = variant_under_utf8_count(s, send);
-    const bool has_trailing_NUL = free_me_ptr && len > 0 && *(send - 1) == '\0';
 
     /* Return the input unchanged if the flag indicates to do so, and there
      * are no characters that differ when represented in UTF-8, and the
      * original is NUL-terminated */
-    if (free_me_ptr != NULL && variant_count == 0 && has_trailing_NUL) {
+    if (free_me_ptr != NULL && variant_count == 0) {
         *free_me_ptr = NULL;
         return (U8 *) s;
     }
@@ -3326,13 +3314,11 @@ Perl_bytes_to_utf8_free_me(pTHX_ const U8 *s, Size_t *lenp,
     U8 *d;
     U8 *dst;
 
-    /*   1 for each byte except any trailing NUL
-     * + 1 for each byte that expands to two
-     * + 1 for the trailing NUL */
-    Newx(d, len - has_trailing_NUL + variant_count + 1, U8);
+    /* 1 for each byte + 1 for each byte that expands to two, + trailing NUL */
+    Newx(d, (*lenp) + variant_count + 1, U8);
     dst = d;
 
-    while (s < send - has_trailing_NUL) {
+    while (s < send) {
         append_utf8_from_native_byte(*s, &d);
         s++;
     }