bug#23595: 25.1.50; file with chinese/japanse chars, vc-diff fails (HG, Git, RCS)

[Paul Eggert]

On 05/23/2016 02:02 PM, Dmitry Gutov wrote:
> Not sure what's the best place to do it, but the patch below gives me 
> 24.5's behavior (correctly decoding the short "Binary files ... 
> differ" output). Could someone try it together with Paul's solution?

It worked for me in the Bug#23595 test case, with Git configured with 
utf16<->utf8 filters as I described. However, it reintroduces a bug when 
the version-controlled uses ISO-2022-JP. If I make a trivial change to 
etc/HELLO, for example, the patch can cause vc-diff to display mojibake, 
as the output of "git diff" uses ISO0-2022-JP but vc-diff decodes it as 
UTF-8. Although this is the same mojibake that Emacs 24.5 generates so 
the behavior is not a regression from 24.5, it is a regression from 
current emacs-25.

We are on thin ice here no matter what. One idea to improve on the 
current emacs-25 behavior is to test whether a simple ASCII message like 
"Binary files differ" encodes as itself using the file's coding system, 
and to use the file's coding system if it does and locale-coding-system 
if it doesn't.

> diff --git a/lisp/vc/vc.el b/lisp/vc/vc.el
> index 25b41e3..b62b68d 100644
> --- a/lisp/vc/vc.el
> +++ b/lisp/vc/vc.el
> @@ -1696,6 +1696,8 @@ vc-diff-internal
>      (setq coding-system-for-read
>            (coding-system-change-eol-conversion coding-system-for-read
>                             'dos)))
> +    (unless (coding-system-get coding-system-for-read 
> :ascii-compatible-p)
> +      (setq coding-system-for-read nil))