diff --git a/pkgs/racket-test-core/tests/racket/unicode.rktl b/pkgs/racket-test-core/tests/racket/unicode.rktl
index f9c3cdb379..b90757b298 100644
--- a/pkgs/racket-test-core/tests/racket/unicode.rktl
+++ b/pkgs/racket-test-core/tests/racket/unicode.rktl
@@ -996,6 +996,17 @@
       (let-values ([(s2 n2 status2) (bytes-convert c2 s)])
         (bytes->string/utf-8 s2)))))
 
+;; Check that `bytes-convert-end` does nothing for UTF-8 and UTF-16 conversion:
+(let ([c (bytes-open-converter "platform-UTF-16" "platform-UTF-8")])
+  (test-values '(#"" complete)
+               (lambda () (bytes-convert-end c))))
+(let ([c (bytes-open-converter "platform-UTF-8" "platform-UTF-16")])
+  (test-values '(#"" complete)
+               (lambda () (bytes-convert-end c))))
+(let ([c (bytes-open-converter "UTF-8-permissive" "UTF-8")])
+  (test-values '(#"" complete)
+               (lambda () (bytes-convert-end c))))
+
 (when (eq? (system-type) 'windows)
   (let ([c (bytes-open-converter "platform-UTF-8-permissive" "platform-UTF-16")])
     ;; Check that we use all 6 bytes of #"\355\240\200\355\260\200" or none
diff --git a/racket/src/racket/src/string.c b/racket/src/racket/src/string.c
index de05025887..7b86e0dd10 100644
--- a/racket/src/racket/src/string.c
+++ b/racket/src/racket/src/string.c
@@ -4696,7 +4696,7 @@ static Scheme_Object *convert_one(const char *who, int opos, int argc, Scheme_Ob
     scheme_get_substring_indices(who, argv[1], argc, argv, 2, 3, &istart, &ifinish);
   } else {
     istart = 0;
-    ifinish = 4; /* This is really a guess about how much space we need for a shift terminator */
+    ifinish = 0;
   }
 
   if (argc > opos) {