Re: [jamboost] Pitfall -- environment variables

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-03-28 11:11:02

----- Original Message -----
From: "Richard Hadsell" <hadsell_at_[hidden]>

> That sounds like a major improvement. In my case, however, even when
it built a
> bin.digital directory, I didn't recognize any problem. It was only
when the
> build failed with incorrect command options and I noticed bin.irix2,
that I
> realized my own OS was conflicting with Jam's usage.
>
> Regardless of how you decide to control the build with environment
variables, I
> recommend (as did Rene Rivera) that you name them all with BOOST_ or
JAM_, so
> that their purpose is clear and less likely to conflict with a user's
own
> environment. You could ask the user to prefix all his own env
variables, but
> that's not very user friendly.
>
> If you want to make the transition to new variable names less painful,
you could
> accept OS in the absence of JAM_OS, but then eventually discontinue
use of OS.

We're going to stop relying on environment variables for
boost.build-specific configuration (except for BOOST_BUILD_PATH, which
will be optional), so there will be no need for BOOST_xxx or JAM_xxx.
For more details, www.boost.org/tools/build/architecture.html.

We're never going to stop responding to "standard" environment variables
such as $(OS) when they can be used to generate a correct configuration
without user intervention on 99% of all systems. However, as I said,
we'll provide a way for users to override those settings by supplying
explicit configuration (in a configuration file).

-Dave