Subject: Re: Running -current.
To: , <>
From: Jonathan Cline <>
List: current-users
Date: 05/20/2003 17:26:09
> From: Richard Rauch <>
>  * The laptop is a 233MHz plain Pentium; no speed demon by present
>    standards.  Should I consider building -current daily?  Or is that
>    not a good idea even with a fast machine?

You can always build on a faster machine and install to the laptop as the
second step?

I don't see -current changing so fast that daily rebuilds are necessary.
Unless there's a specific change you're waiting for.  Perhaps weekly, or
other week.  'update'ing faster than that reduces productivity as you spend
more time syncing and merging than actually developing.  It's best to update
when you've reached a two week milestone and have tested your stuff (in my
experience), so you know when everything breaks that it isn't your stuff
is breaking.

So I guess that means "No, it's not worth rebuilding every day".

>    backup for re-installing?  If so, the easiest backup target medium
>    is an NFS-mounted partition on a 1.6 server; is tar good for a complete
>    system backup, or should I consider using something else?)

tar is the best backup.  As long as you're comfortable rebuilding your
disks with a rescue floppy.  Otherwise keep a dump'ed backup.

>    What's the best way to keep a local hack (yes, my change *IS* a hack,
>    but it makes the driver work for me) in?

When you 'cvs update' your fix will remain (cvs will merge it).  Only
if someone else hacks the same area of that source file will cvs have
a conflict and require manual intervention:
   * cvs update -dP | grep '^[MC]'
will tell you.

>  * A little related to part of the above: The reason that I want to run
>    -current is to do some belated work on window(1).  The hoops to make
>    it all work on a 1.6 system were enough that I just want to put a
>    development system on -current.  So, what's a good way (without my
>    own branch on the repository and without write access to the main
>    branch) to do this?  I'd rather synchronize at least periodically (if
>    not regularly) as opposed to only after making changes.

I use sourceforge.
I sync with netbsd cvs regularly:
    * cd /usr/src; cvs -d"blah" update -dP
and, I commit/sync with my sourceforge cvs regularly:
    * cd /usr/work; cvs -d"blah" commit

I just put symlinks in where necessary:
    * ln -s /usr/work/usr.bin/blah /usr/src/usr.bin/blah

The repositories are seperate and cvs seems to do the right thing.

>    I have the means to set up my own repository, but am not sure how
>    I would go about tracking *that* to the real repository to keep myself
>    in synch.  Is that the way to go?

It's not worth it.  There are several free cvs servers out there which have
small-scale development repositories set up.  If anyone else ends up
with your development, they use the same server.  Plus all the things which
with cvs are already integrated on these servers (i.e. cvsweb,
somewhat-automated backups, rudimentary task reminders, etc).