Re: patches to cvshome cvs sources... when is it okay?

[Derek Robert Price]

My apologies for the brevity of my response.  I have a few deadlines 
coming up.

Mark D. Baushke wrote:

> Hi Derek & Larry,
>
> In the absence of a 'no', should I assume a 'yes' for committing the
> currently proposed patches:
>
> -> CVS_LOCAL_BRANCH_NUM, pay attention to existing environment variable

This is great.

> -> CVS_PID, newly generated enviornment variables

Sorry, I'm still unconvinced on this one, but I still haven't had time 
to go through your last email.  Unless Larry wants to take over the 
conversation, I'll try and get to it this week.  Failing that it will be 
the week after next.

> I am still trying to understand the etiquette for committing. :-)

If you think everyone will agree, go ahead.  If not, discuss it.  If no 
one argues too much, or you seem to win the argument, go ahead.  :)

> I'd also like your opionion as to if I should consider spending time on
> breaking up the FreeBSD features into small patch sets for commit.

Sure.  I annotated my individual opinions below.

> For my employer, I need for the next version of cvs I deploy to have
> CVS_LOCAL_BRANCH_NUM support and would find it highly desirable to have
> the variable keyword expansion controls and the new cvs diff options.
>
> I would like to deploy the cvshome.org sources as we are in a
> heterogeneous environment, so I'd like to see these patches added to the
> baseline cvshome version.
>
>         Thanks,
>         -- Mark
>
> Here is a brief description of the outstandings feature differences
> between the FreeBSD cvs and the cvshome cvs.
>
>  - CVSREADONLYFS environment variable and global option -R to enable
>    no-locking readonly mode (eg: cvs repo is a cdrom or mirror)
>         [A minor feature that could be misused as it does not create
>         any lock files in the repository at all.]

Sure.

>  - CVS_OPTIONS environment variable read before the .cvsrcfile
>         [Useful if you need to override the .cvsrcfile at times, say by
>         telling it to ignore the .cvsrc file .]

What's wrong with -f?

>  - CVSROOT/options file
>         [Useful for the CVSHeader and other keyword expansion controls.]
>
>  - Variable keyword expansion controls including custom keywords.
>         [Being able to add new keywords like $FreeBSD$ $OpenBSD$ $NetBSD$
>         for keeping track of the version of code imported from someone
>         else that uses cvs source control.]
>
>  - $ CVSHeader$ keyword - like Header, but with $CVSROOT stripped off.
>         [New cvs keywords that help cvs mirrors that keep files in trees
>         that are different than the master repository looking the same.]

I'd like more information on these.

>  - CVS_RSH is ssh rather than rsh
>         [This is the right default, but not everyone agrees yet.]

Sure.  It's probably about time.

>  - -T switch for checkout and update to Create/Update the
>    CVS/Template file.
>         [Sadly, this feature is intended for a local cvs mirror and does
>         not work in client/server mode. It would be nice to setup a new
>         client/server message to update the CVS/Template file when a cvs
>         update is performed, but that work has not yet happened.]

I think I discussed this with you before.  I think this fix should be 
trivial and could even be made backwards compatible with old clients 
since (I think) clients send the entire contents of their CVS 
directories and (I know) that server can create files with arbitrary 
names on the client.

>  - cvs diff accepts two -j rev:date options to diff against a
>    revision relative to a branch tag in the past.
>         [This general kind of rev:date format needs more discussion,
>         but is very nice to use in practice.]

I assume you mean two -r rev:date options using the -j format from 
update.  I think this is useful and actually think it should be a change 
made across all the commands that accept -r.  And -r:date should specify 
a date on the trunk.

>  - New global -U switch for server mode to skip CVSROOT/password and
>    require /etc/passwd users only.
>         [I am not sure how to justify this patch.]

Maybe a switch.  I think this would be more appropriately stored in the 
CVSROOT/config file, probably as a new value, "only", to the SystemAuth key.

>  - New global -g switch to force group-write permissions on working
>    files.
>         [Let your umask be relaxed for some of your cvs sandboxes.]

Maybe.  I'd like to hear more justification.

Derek

--
               *8^)

Email: derek@ximbiot.com

Get CVS support at <http://ximbiot.com>!
--
It's always darkest before dawn. So if you're going to steal the
neighbor's newspaper, that's the time to do it.