Re: Re[2]: [lmi] Should we restrict cvs files to ASCII?

[Evgeniy Tarassov]

On 4/17/07, Vadim Zeitlin <address@hidden> wrote:
> On Tue, 17 Apr 2007 17:57:49 +0200 Evgeniy Tarassov <address@hidden> wrote:
>
> ET> I suspect that 'cvs protocol' has a hardcoded utf8 encoding for text files.
>
>  No, not at all, it knows nothing about encodings.
>
> ET> Otherwise any patch file containing non-ASCII characters will depend
> ET> on the way it was created ('cvs diff' will produce utf8 encoded file,
> ET> 'diff' will produce ISO-8859-1 (native lmi encoding)).
>
>  Again, no, cvs diff doesn't produce UTF-8 encoded files, the output will
> contain UTF-8 characters if and only if the input file had them.

This is the thing that puzzles me. The current encoding of ChangeLog
is 'ISO-8859-1', but 'cvs diff' produces the patch as if the original
encoding of ChangeLog was utf8.

You can see it for yourself -- just checkout a clean version of lmi,
then cheng ChangeLog file (duplicate the line with non-ASCII
character) and run:
cvs -d:pserver:address@hidden:/sources/lmi diff
ChangeLog > ChangeLog.patch

Then compare the produced ChangeLog.patch file and ChangeLog -- the
(implicit) text encoding is different. That means that somewhere along
the way (in cvs diff) the text encoding was changed somehow.

--
Best wishes,
Evgeniy Tarassov