Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 4 Feb 2006 10:42:08 -0800 (PST)
From:      Matthew Jacob <mj@feral.com>
To:        Wilko Bulte <wb@freebie.xs4all.nl>
Cc:        src-committers@FreeBSD.org, harti@FreeBSD.org, sparc64@FreeBSD.org, stable@FreeBSD.org, kris@obsecurity.org, des@des.no
Subject:   Re: [releng_6 tinderbox] failure on sparc64/sparc64
Message-ID:  <20060204103132.D88429@ns1.feral.com>
In-Reply-To: <20060204092225.GB46310@freebie.xs4all.nl>
References:  <861wykr9vx.fsf@xps.des.no> <20060203.105305.71186162.imp@bsdimp.com> <43E4142A.4@samsco.org> <20060203.215549.74746986.imp@bsdimp.com> <20060204092225.GB46310@freebie.xs4all.nl>

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

>
> I would think that the tinderboxes should run 100% the same flags as
> what normal release builds use.  Nothing more, nothing less.

What I would like to see is a pointer to a procedure and tools to make 
sure builds aren't broken.

I've been refreshing my memory about email going back about 5 years, and 
at that time there was a lot of wrangle over people not doing adequate 
checking for at least syntactic correctness for multiple platforms.

I certainly have broken a lot more than I would like lately, and part of 
this (other than being too stupid and hasty) came about because it 
wasn't actually obvious what would be a good pre-commit compile check 
for kernels for me to follow.

At the very least I've now come up with:

  compile GENERIC (easy enough to do)
  compile LINT (this wasn't obvious how to make LINT)
  compile PAE (for i386 at least)

Since the complaint of 5 years ago by many was that they didn't have 
alphas to compile on is still relatively true (that is, few people have 
more than one architecture) has been addressed in two ways (tinderbox, 
and cross-compilers), the issue should be better, but for three things 
I've observed:

a) The tinderbox breakage is being treated as bad as stop ship type of 
bug rather than being informative as it should be. I feel I got roasted 
and slammed for what should have been simply a "hey- Matt- come fix this 
please!".

b) It's instantly not obvious to me (being lazy and not having kept all 
my committer mail in a way I can find) how *I* can do a tinderbox run 
myself.

c) Similarily, I don't know how to build a cross-build environment. I 
should, and I bet if grovel around a bit I can find out how to do so.

The point here is that if well-meaning and moderately intelligent 
committers miss steps that are important to keeping the quality up, 
please point them at documentation that gives a reasonably coherent set 
of steps as to how to correct their errors. I'm sure that most of those 
who err will spend the extra late night hours to get it right then.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060204103132.D88429>