[Fwd: Re: [21.2]: VC/RCS changes default branch when dealing with branch re vision]

[Fwd: Re: [21.2]: VC/RCS changes default branch when dealing with branch re vision]

[Andre Spiegel]

[-- Attachment #1: Type: text/plain, Size: 48 bytes --]

Here is a copy of my reply to Simon's message.


[-- Attachment #2: Forwarded message - Re: [21.2]: VC/RCS changes default branch when dealing with branch re vision --]
[-- Type: message/rfc822, Size: 2228 bytes --]

From: Andre Spiegel <spiegel@gnu.org>
To: "Marshall, Simon" <simon.marshall@misys.com>
Subject: Re: [21.2]: VC/RCS changes default branch when dealing with branch re vision
Date: 15 Jun 2002 13:20:11 +0200
Message-ID: <1024139335.3519.30.camel@eagle>

On Fri, 2002-06-14 at 18:16, Marshall, Simon wrote:

> Emacs has set the default branch revision in RCS.  AFAICS this happens
> when I C-u C-x v v to change to edit the branch revision.  Perhaps Emacs
> has (or had) good reason to change the default branch, but I think it is
> wrong.  C-u C-x v v is the only way to checkout a branch revision, e.g.,
> for editing, but it doesn't mean I want the branch revision to be the
> default revision in RCS.

Thanks for your analysis.  This behavior is fully intentional, though.
The reason is, as we put it in the Emacs manual, that after you switch
to a branch, you "stay" on it for subsequent editing.  And yes, Emacs
achieves this by setting the RCS default branch.

To go back to the trunk in Emacs (and clear the default branch), you'd
have to do C-u C-x v v RET once.  You could of course also do this from
the shell using rcs -b; I'm not sure if plain `co' also has a
corresponding option.

The reason we set the default branch is that without it, at the
beginning of a new session, Emacs would always think it is working on
the trunk, even if you've left the branch version checked out, and
wanted to stay on the branch.  Emacs _can_ detect that if you use
version headers ($Id$) in your file.  In that case, setting the default
branch wouldn't be necessary, but we decided to do it because the
feature needs to work even in the absence of version headers.

This has been the behavior ever since RCS branches were supported, i.e.
for more than five years.  So far, people seem to have been content with
it.  Can you make a compelling case against it?  I'm open to
suggestions, but as far as I can think now, the current behavior does
make a lot of sense.