Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 27 May 2008 16:22:55 -0700
From:      Marcel Moolenaar <xcllnt@mac.com>
To:        Peter Wemm <peter@wemm.org>
Cc:        src-committers@freebsd.org, cvs-src@freebsd.org, cvs-all@freebsd.org, Poul-Henning Kamp <phk@phk.freebsd.dk>, John Birrell <jb@freebsd.org>, Robert Watson <rwatson@freebsd.org>, John Birrell <jb@what-creek.com>
Subject:   Re: cvs commit: src Makefile
Message-ID:  <C5830363-526E-4D79-99CD-C56A57C000D6@mac.com>
In-Reply-To: <e7db6d980805271535h5533862bk44d7167c77106852@mail.gmail.com>
References:  <20080525221112.GH5179@what-creek.com> <21823.1211785618@critter.freebsd.dk> <e7db6d980805262223y4ee76253u1ad9d29213fd580@mail.gmail.com> <20080527110625.GA97301@zim.MIT.EDU> <e7db6d980805271535h5533862bk44d7167c77106852@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help

On May 27, 2008, at 3:35 PM, Peter Wemm wrote:

> On Tue, May 27, 2008 at 4:06 AM, David Schultz <das@freebsd.org>  
> wrote:
>> On Mon, May 26, 2008, Peter Wemm wrote:
>>> On Mon, May 26, 2008 at 12:06 AM, Poul-Henning Kamp <phk@phk.freebsd.dk 
>>> > wrote:
>>>> In message <20080525221112.GH5179@what-creek.com>, John Birrell  
>>>> writes:
>>>>
>>>>> I will back out the change, but I think you are making the  
>>>>> 'universe' target
>>>>> out to be more than was intended. We used to talk about tiers.  
>>>>> We seem to have
>>>>> lost sight of that.
>>>>
>>>> No, architecture tiers is about code how well the code runs, make
>>>> universe is simply a way to keep it compiling.
>>>
>>> sparc64 and sun4v share userland.  The sparc64 in universe overs 99%
>>> of the compile test for sun4v already.
>>>
>>> It seems a shame to compile sparc64 userland twice for universe.   
>>> And
>>> on that note, do we compile i386 twice for i386 and pc98?
>>>
>>> I might find myself more inclined to use 'universe' if it had less
>>> duplicated work.
>>
>> I've always thought it would be nice to have a stripped-down
>> version of make universe (make galaxy?) that compiled for a
>> representative sample of platforms, and with only one or two
>> kernels per platform instead of 3 or 4 or 5. For small changes,
>> this represents a better tradeoff in time spent compiling vs. the
>> cost of things breaking occasionally. For actively developed
>> platforms, breaking the build wastes lots of people's time; for
>> everything else, there's tinderbox.
>
> "make tier1" ?  (Stuff which must not be broken)
> "make tier2" etc.

These are bad ideas, because people *WILL* do the absolute minimum
and as such will end up breaking non-tier1 platforms even more
often than they do already.

People need to remember that maintainers of non-tier1 platforms
spend most of their time fixing problems that can easily be dealt
with by the 300+ developers not worrying about non-tier1 (as it
hardly ever requires in-depth knowledge of the platform). For the
non-tier1 maintainers this is a *BIG* waste of their time...

FYI,

-- 
Marcel Moolenaar
xcllnt@mac.com






Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?C5830363-526E-4D79-99CD-C56A57C000D6>