[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Adding an option to configure?
From: |
Keith MARSHALL |
Subject: |
Re: Adding an option to configure? |
Date: |
Tue, 3 Jan 2006 16:53:14 +0000 |
Stepan Kasal wrote:
> even though `--without-gcclib' might sound better to a newcomer, the
> convention is that:
>
> `--without-gcclib' means that gcclib is not installed on the host
system
> `--disable-gcclib' implies that this package contains a subsystem
called
> gcclib, and it means that this subsystem should not be
built
>
> This convention is not obvious until someone tells you about it; yet it
> is good to follow it because of people who know about it and expect it.
Exactly. It's an accepted convention, and in such a case, I agree that it
is best to adopt it; but within that convention, there are some decisions
which will be less clear cut, and then the choice comes down to personal
taste. For example, I'm currently autoconfiscating `man', to make it port
more readily to Win32 using MinGW/MSYS. `man' requires that some variant
of troff be installed -- it won't work without it -- but there are user
choices to be made in the options with which troff will be invoked. I've
chosen to use configure options such as
--with-nroff='nroff -Tascii -mandoc'
to achieve this. This usage doesn't strictly conform to the convention,
but then, IMHO, the requirement doesn't strictly fit the guidelines for
either of the documented conventions either; neither is there any other
technique, of which I am aware, which better fits my needs -- note that,
in this example, neither `--without-nroff' nor `--disable-nroff' would
have any practical application.
> That's why I jumped in: to encourage following the documented semantics,
> instead of mere guess.
I appreciate your objective; I'm merely suggesting that, in some cases,
it may not be appropriate to be too dogmatic about conventions.
Regards,
Keith.