(after-insert-file-set-buffer-file-coding-system):

Don't make buffer unibyte unless we seem to be visiting a file.

1 files changed, 6 insertions, 4 deletions

diff --git a/lisp/international/mule.el b/lisp/international/mule.el
index 82d9cbcddc9..831d2811897 100644
--- a/lisp/international/mule.el
+++ b/lisp/international/mule.el
@@ -845,10 +845,12 @@ function by default."
             (modified-p (buffer-modified-p)))
         (when coding-system
           (set-buffer-file-coding-system coding-system)
-          (if (or (eq coding-system 'no-conversion)
-                  (eq (coding-system-type coding-system) 5))
-              ;; It seems that random 8-bit codes are read.  We had
-              ;; better edit this buffer without multibyte characters.
+          (if (and (or (eq coding-system 'no-conversion)
+                       (eq (coding-system-type coding-system) 5))
+                   ;; If buffer was unmodified, we must be visiting it.
+                   (not modified-p))
+              ;; For coding systems no-conversion and raw-text...,
+              ;; edit the buffer as unibyte.
               (set-buffer-multibyte nil))
           (set-buffer-modified-p modified-p))))
   nil)