[cvsnt] Re: CVSNT 2.0.34 puts extra info in RCS files that CVSGraph does not like...

[Bo Berglund]

As I stated this was a copy of a ViewCvs list mail because I thought
that the CVSNT list would have better possibilities of answering.
I will forward your reply to the ViewCvs list and hopefully the
developers there can fix this.
They have already swiched from RCS to using the cvs.exe for parsing
the RCS files, now someone also needs to check up on how cvsgraph is
called.
If we are lucky then cvsgraph does not call RCS by itself but rather
use suff sent in by the ViewCvs system. If we are unlucky then the
cvsgraph developer has to be activated and switch from RCS to cvs.exe
too.

/Bo

On Fri, 02 Apr 2004 10:50:55 +0100, Tony Hoyle <tmh at nodomain.org>
wrote:

>On Thu, 01 Apr 2004 19:43:22 +0200, Bo Berglund
><bo.berglund at telia.com> wrote:
>
>>This is a copy of an email that was posted on the ViewCvs mailing
>>list.
>>I thought that I should re-post it here in order to have Tony or
>>someone else look at it and see if anything can be done about it.
>>I have looked at some of my RCS files and indeed there is no @
>>characters enclosing the committing user.
>>/Bo
>
>The @ characters are necessary when then username contains
>non-alphanumeric tokens.  It's actually an error in the original RCS
>specs - they treated the author name as a token which fails horribly
>if the name has a space or several other characters in it (@, ;, a few
>others).
>
>Even in the 'old' CVS if you had a username with spaces you'd get a
>broken RCS file which CVS could read but RCS couldn't, so there's no
>real change to the behaviour in that sense... CVSNT tries to avoid
>using the @ bracketing if your name is also a valid token.
>
>Really you must use the RCS wrappers with CVSNT repositories.  These
>handle such things (and more major changes, like binary deltas)
>transparently.
>
>Tony


/Bo
(Bo Berglund, developer in Sweden)