bug#2741: Mixed UTF-8 and raw bytes in output of vc-annotate after (set-language-environment "UTF-8")

[Juanma Barranquero]

1) Create a Git repository and add a Latin-1 file with some non-ASCII
characters. In my example, the archive test.txt contains the following
text:

    A few Spanish characters: áéíóúüñ

2) Execute "emacs -Q test.txt -f vc-annotate". The resulting *Annotate
test.txt* buffer has buffer-file-coding-system `iso-latin-1-dos' and
shows:

    ^7fb00c1 (Juanma Barranquero 2009-03-22 00:01:39 +0100 1) A few
Spanish characters: áéíóúüñ

3) Set LANG to UTF-8 (for example, "set LANG=en_US.UTF-8"), and repeat
"emacs -Q test.txt -f vc-annotate". Now the *Annotate* buffer is in
`utf-8-dos', and shows:

    ^7fb00c1 (Juanma Barranquero 2009-03-22 00:01:39 +0100 1) A few
Spanish characters: áéíóúüñ

4) Finally, after unsetting LANG or not (it is irrelevant) do

    emacs -Q --eval "(set-language-environment \"UTF-8\")" test.txt -f
vc-annotate

  Now the *Annotate* buffer is in `utf-8-dos', but contains a mixture
of utf-8 and raw bytes:

    ^7fb00c1 (Juanma Barranquero 2009-03-22 00:01:39 +0100 1) A few
Spanish characters: \341\351\355\363\372\374\361

    Juanma