From: Ian Macdonald <ian@...>
Date: 2002-06-22T07:41:31+09:00
Subject: Re: Why is I/O slow?

On Sat 22 Jun 2002 at 06:21:28 +0900, Chris Ross wrote:

> Yes completely possible. But that is what I am saying matz wants to 
> avoid. Currently if there is a bug three main questions are asked,
> "what version of ruby, what os and what architecture". Those questions
> would then be added to if it's a compile time option "what flags where 
> used, what versions of the libraries, what optimisations were compiled
> in",

Just run ./configure --help to see all the possible compile-time
options that can be set. They're already part of the troubleshooting
equation.

Libraries, too, can make a big difference. There might be a bug in db1
that isn't present in db3 or db4, for example.

Furthermore, some systems don't have certain syscalls, so #ifdef
clauses are already needed there.

You just can't get a single unified binary for all systems.

Ian
-- 
Ian Macdonald               | Some men are so macho they'll get you
ian@caliban.org             | pregnant just to kill a rabbit.   --
                            | Maureen Murphy 
                            | 
                            |