[cvsnt] Re: Problem on CVSNT as client + CVS as server

[Tony Hoyle]

Sean Choi wrote:

>              <<<<<<< abc.txt
>               my changes=======
>               cvs source changes>>>>>>>cvs version

It's saying that you have a nonterminated line at the end of your file, so
it's reflecting that accurately in the two versions.  It's a choice between
accurate&ugly and innacurate&pretty - cvs has historically gone for the
'accurate' version.

>   2. unicode file
>      I find that after updating and checking out an unicode file, the
> unicode file is corrupted.
>      Besides there are two bytes "FFFE" at the beginning of the file as
> unicode header in windows, two more bytes "FFEE" is added to the unicode
> file.
> 
Unicode has FEFF or FFFE depending on the endianness of the file (not FFEE -
if any software is adding this then it is broken).

cvsnt internally converts this to UTF8, which has a marker of EFBBBF.  This
is what you'll see if you forget to mark the file with '-ku', as it'll
checkout as UFT8 by default.

*however* your problem is you are trying to use unicode with Unix CVS - you
cannot use unicode with the cvshome.org version.  Your only option (apart
from using cvsnt on the server side) is to make the files as binary.

Tony