Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 18 Dec 1995 15:33:19 +0000 (GMT)
From:      Paul Richards <paul@netcraft.co.uk>
To:        FreeBSD-current@FreeBSD.org (FreeBSD current mailing list)
Subject:   Re: FreeBSD-current-stable ???
Message-ID:  <199512181533.PAA18935@server.netcraft.co.uk>
In-Reply-To: <199512171731.JAA09419@freefall.freebsd.org> from "John Dyson" at Dec 17, 95 09:31:26 am

next in thread | previous in thread | raw e-mail | index | archive | help
Trimmed an out of controll CC line to just be -current.

In reply to John Dyson who said
> 
> A make world was not done, and if someone would donate a reasonable
> machine to me to let me do so -- it would be very nice.  I probably
> have the least powerful machine of any major contributor (and have
> only one.)  Those with expensive high power machines are welcome to
> help.  (Machines bigger than a 3 yr old 20MB 486/66 :-)).
> 

It's not *that* small. Geeze, remember when we started, I had a 4Mb 25Mhz
486 when we did 1.0, things have changed a lot in the last few years :-)

> The kernel did work -- and there was some chaff (a bug in sys_process.c)
> -- oh, by the way did the 1Tb changes break things -- or was it the header file
> changes/improvements????  All I had to do to get ps working again was to
> rebuild libkvm/ps....

The vm header file changes, I didn't think you were responsible
for that, I was using the 1Tb changes as an example of something
where I would expect the odd bug to slip through since simply
checking it compiles and even running it for a while is not going
to flush out every possible situation. This is exactly what -current
IS for.  Testing out totally new code to flush out those unexpected
bugs.  You do run current at a risk to your system since a filesystem
or vm bug could be very nasty, that's the sacrifice you make. What
I don't like to see though is a situation where the tree simply
wouldn't compile. It wasn't even a simple oversight, it was damn
clear no attempt had been made to compile the tree because a whole
load of utils just plain simple wouldn't build because the vm header
files had changed. There's just no excuse for that. If you don't
have a machine that you're willing to run a make world on then
don't play in areas that require a make world to validate them.

I also disagree with Poul's suggestion that people go hack user-land
code in 2.1 for 2 reasons.

1) You'd have to be pretty selective in what you did for it to
seamlessly drop into -current since all sorts of fundamental things
can change in -current, like header files, mk files, malloc, sysctl
:-) etc. It's not really a good idea to tell people to go and work
in 2.1 and really is just a cop out from addressing the real problem
which is that -current is being hacked without any clear focus as to
where it's headed.

2) It's *IMPORTANT* to have people actually running -current even if they're
not actively developing. If only a tiny number of active developers are
hacking -current then we'll never have enough of an user base for that
branch to find all the bugs that crop up from diverse use.

I've advocated having an actual plan of what features to bring in for
a particular release in the past but -core always felt this was a
restriction of volunteers freedom to give what they can when they can.

I still think that for a single release there should be one or two
main developments and that others should wait until the next release.
Releases would happen more often that way so people wouldn't have
to wait that long before it was their turn to bring in their idea.

While they're waiting for their turn they could get it that much more
polished and perhaps it would come into -current mostly working :-)

-- 
  Paul Richards, Netcraft Ltd.
  Internet: paul@netcraft.co.uk, http://www.netcraft.co.uk
  Phone: 0370 462071 (Mobile), +44 1225 447500 (work)



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199512181533.PAA18935>