Remove extra padding element from simple-character-string

There is no convention for UCS-4 strings to be terminated by 4 zero bytes.
Unlike with ASCII strings, it doesn't make Lisp more C-compatible.

src/compiler/generic/vm-array.lisp

@@ -78,7 +78,6 @@
                     :complex-typecode #.complex-base-string-widetag)
          #+sb-unicode
          (character ,(code-char 0) 32 simple-character-string
-                    :n-pad-elements 1
                     :complex-typecode #.complex-character-string-widetag)
          (single-float $0.0f0 32 simple-array-single-float)
          (double-float $0.0d0 64 simple-array-double-float)

src/runtime/gc-common.c

@@ -938,12 +938,11 @@ NWORDS(uword_t x, uword_t n_bits)
 
 DEF_SPECIALIZED_VECTOR(vector_nil, 0*length)
 DEF_SPECIALIZED_VECTOR(vector_bit, NWORDS(length,1))
-/* NOTE: strings contain one more element of data (a terminating '\0'
+/* NOTE: base strings contain one more element of data (a terminating '\0'
  * to help interface with C functions) than indicated by the length slot.
- * This is true even for UCS4 strings, despite that C APIs are unlikely
- * to have a convention that expects 4 zero bytes. */
+ * UCS4 strings do not get a terminator element */
 DEF_SPECIALIZED_VECTOR(base_string, NWORDS((length+1), 8))
-DEF_SPECIALIZED_VECTOR(character_string, NWORDS((length+1), 32))
+DEF_SPECIALIZED_VECTOR(character_string, NWORDS(length, 32))
 DEF_SCAV_TRANS_SIZE_UB(2)
 DEF_SCAV_TRANS_SIZE_UB(4)
 DEF_SCAV_TRANS_SIZE_UB(8)

tests/array.pure.lisp

@@ -433,7 +433,7 @@
   (checked-compile-and-assert (:optimize '(:safety 0))
       `(lambda (x)
          ;; Strings are null-terminated for C interoperability
-         (char "abcd" x))
+         (char #.(coerce "abcd" 'simple-base-string) x))
     ((4) #\Nul)))
 
 (with-test (:name (adjust-array :transform))