RE: libgetopt++ and setup and libstdc++

["Robert Collins" <robert dot collins at itdomain dot com dot au>]

> -----Original Message-----
> From: Earnie Boyd [mailto:earnie_boyd@yahoo.com] 
> Sent: Sunday, April 28, 2002 4:23 AM

> > However, CVS is _not_ a distribution, and anyone developing 
> from CVS 
> > will require the autotool suite if they make any 
> non-trivial changes.
> > 
> 
> A source distribution is a source distribution regardless of 
> how the source is distributed.  

Granted. 

> Cvs is a viable source 
> distribution medium even though it's purpose is to aid 
> developers in development.

Distribution is an act that developers (and any user once they have the
source) takes. If I chose to distribute directly from CVS, then I would
do things differently. However, I choose to distribute from packages
(for several reasons(*)). If a user chooses to get the code from CVS (or
even bothers getting the code at all) then I expect them to have the
appropiate tools. I think this is a reasonable approach, and not
burdensome on users or developers. 

As usual, nothing is fixed in stone, and if this is a significant issue,
it can be revisited.

Rob

*: Why don't I encourage users to get everything from CVS?
1) If users expect CVS to be the means for distribution it makes
development harder. An expectation arises that HEAD won't get broken,
because if it does, users can't grab the source and build. Experiments
can not be made as easily. Developers understand this, and can roll CVS
forward and backwards to play with their local stuff and avoid breakage,
whereas users don't want or need that potential hassle. If they do want
that, then they are (in my book) developers, and the bar for getting the
source can be a little higher without undue concern.
2) CVS can be flakey. Even when not broken it may not 'do the right
thing'. This is why I branch everything before releasing, so that the
used code can be made stable and correct, even if that involves kludges
that are inappropriate for the HEAD code.