Re: The right configuration mechanism

From: Jonathan Turkanis (technews_at_[hidden])
Date: 2005-08-06 23:16:52

David Abrahams wrote:
> "Jonathan Turkanis" writes:
>> David Abrahams wrote:
>>> "Jonathan Turkanis" writes:
>>>

>> I've written two chapters of O'Reilly C++ Cookbook, which is
>> scheduled to come out in Nov 2005. It contains five "recipes" about
>> using Boost.Build, the first of which described installation and
>> configuration.
>
> Wow, that is too cool!

I tried to put in as much Boost stuff as I could. I know a number of other boost
libraries are dicussed as well, including Date Time, String Algorithms and
Conversion.

>
>> The types of sytax I describe are
>>
>> using toolset-name ;
>>
>> if there's just one version of a toolset and it's installed in a
>> standard location,
>>
>> using toolset-name : : command ;
>>
>> if there's just one version of a toolset but its installed in a
>> non-standard location, and
>>
>> using toolset-name : version-name-1 : command-1 ;
>> ...
>> using toolset-name : version-name-N : command-N ;
>>
>> if there's more than one version of a toolset.
>>
>> If this syntax might not work after the official release of BB.v2,
>> I'll just say "follow the instructions in user-config.jam" (assuming
>> that some good instructions will be put there).
>
> There's always some risk in documenting evolving software...

This is a major concern, not just with Boost.Build but with some other software
projects I discuss. I'm trying to keep in close contact with the developers to
make sure the book will not be obsolete before it's released.

> but I
> think the plan was to keep "using" around for backward compatibility
> and introduce "configure."

Ideally, I'd like to use the "prefered" syntax, but I don't want to mention it
if it isn't implemented yet. As long as the syntax I presented will continue to
work, that's good enough for me; readers can always read the documentation for
the most up-to-date and flexible syntax.

I'm giving very detailed installation instructions because I don't want readers
to get stuck trying to set up Boost.Build and never actually try using it.

Jonathan