decode-coding-region returns octal escapes

[Eli Zaretskii <eliz@gnu.org>]

To reproduce:

  emacs -Q
  M-x find-file-literally RET etc/tutorials/TUTORIAL.de RET

(The German tutorial is just an example, you can use any one, but
others might need different values of the 3rd argument to
decode-coding-region below.)

Now mark a region that includes non-ASCII characters, for example,
this one, on line 7 of the file:

    da\337 die CONTROL-Taste gedr\374ckt sein mu\337

and type

    M-: (decode-coding-region (mark) (point) 'latin-1 t) RET

The result is that the echo area shows this:

    #("da\337 die CONTROL-Taste gedr\374ckt sein mu\337" 0 39 (charset iso-8859-1))

and the *Messages* buffer shows this:

    #("da\303\237 die CONTROL-Taste gedr\303\274ckt sein mu\303\237" 0 39 (charset iso-8859-1))

But I expected to see this in both cases:

    #("daß die CONTROL-Taste gedrückt sein muß" 0 39 (charset iso-8859-1))

Is this a bug?  If not, what is the explanation for what I see?  Why
are raw bytes inserted instead of decoded characters?