Date: Tue, 2 May 2006 13:53:43 -0700 From: Paul Allen <nospam@ugcs.caltech.edu> To: Mark Linimon <linimon@lonesome.com> Cc: Giorgos Keramidas <keramida@ceid.upatras.gr>, joaoBR <joao@matik.com.br>, current@freebsd.org Subject: Re: cc can't build 32-bit executables on amd64 Message-ID: <20060502205343.GA28259@regurgitate.ugcs.caltech.edu> In-Reply-To: <20060502201847.GA7449@soaustin.net> References: <200605011604.26507.mi%2Bmx@aldan.algebra.com> <20060501212539.GA24193@regurgitate.ugcs.caltech.edu> <4456C439.1070500@samsco.org> <4456E860.8090308@samsco.org> <20060502163204.GB31236@soaustin.net> <4457928F.60805@matik.com.br> <20060502175502.GA31993@gothmog.pc> <20060502201847.GA7449@soaustin.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> It took me quite a while to learn it, and even that was on embedded systems
> where the code was much smaller and centrally controlled than an entire OS
> plus utilities plus applications. The problem space was orders of magnitude
> smaller, and the code only needed to run on one piece of hardware. Even so,
> there was always something else that was broken ...
Yes... usually things "work" not because they do what you intended but often
because of some unusual unplanned property. e.g., the size of types happens to
just work out.
This shouldn't be surprising. Write a bunch of code... it has many errors in it.
Stochastically fix them until it works. It goes a long way to demonstrating how
biological evolution works.
Luckily hardware designers take a more rigorous approach, but then again an ECO
could cost hundreds of thousands of dollars--if not millions.
Paul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060502205343.GA28259>