autoconf
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: better param checking?


From: Harlan Stenn
Subject: Re: better param checking?
Date: Sun, 13 Oct 2002 20:31:51 -0400
User-agent: EMH/1.10.0 SEMI/1.13.7 (Awazu) FLIM/1.13.2 (Kasanui) XEmacs/21.1 (patch 12) (Channel Islands) (i386--freebsd)

> | is there a way to tell configure to check params like --with-foo?
> | 
> | ./configure --foo failes with an error
> | ./configure --with-foo doesn't fail ... it just ignores the
> | "--with-foo" option.
> | 
> | this might also apply to --enable-foo
> | 
> | ./configure --with-foo should fail with the message, that the option
> | called "foo" is not available.
> 
> That's on purpose: in case you are embeded in a huge tree, with a main
> configure dispatching arguments to everybody...  I have not found a
> means to have both this property, and some form of checking.  Expect
> maybe a simple warning, but we cannot make it an error.

There may be 2 separate issues here:

- if a package sees --with-foo and it does an AC_ARG_WITH(foo), it may be
  perfectly reasonable for the package author to say "I was told to build
  with foo and foo is not there, I will abort." and code it that way.

- if a package sees --with-foo and it does not contain AC_ARG_WITH(foo) then
  is should be ignored, for the reason given by Akim.

The same logic applies to --enable.

Now one could, for example, also code --with-fatal-missing-with to control
if my first case would be a fatal or not.

And that sort of behavior would best come from stock autoconf, as otherwise
every package maintainer (or worse, end-user package builders) would have to
add that capability.

H





reply via email to

[Prev in Thread] Current Thread [Next in Thread]