From: Nobuyoshi Nakada <nobu@...>
Date: 2007-08-30T11:04:32+09:00
Subject: Re: Whatever happened to have_const?

Hi,

At Thu, 30 Aug 2007 05:29:57 +0900,
Daniel Berger wrote in [ruby-core:12027]:
> What ever happened to the have_const patch that was supposed to get 
> added to mkmf.rb?

Sorry, but nothing yet.

> I ask because I've hit yet another scenario where I need to check for an 
> enum value.
> 
> BTW, there's a bug in that diff - the word "type" should be replaced 
> with "const".

For the case of non-integral value as asked by Matz in
[ruby-core:04428], I'd change the check code like this.

If no objection, I want to commit it to 1.8 and trunk.
Any thoughts?


def try_const(const, header = nil, opt = "", &b)
  const, type = *const
  if try_compile(<<"SRC", opt, &b)
#{COMMON_HEADERS}
#{cpp_include(header)}
/*top*/
typedef #{type || 'int'} conftest_type;
conftest_type conftestval = #{type ? '' : '(int)'}#{const};
SRC
    $defs.push(format("-DHAVE_CONST_%s", const.strip.upcase.tr_s("^A-Z0-9_", "_")))
    true
  else
    false
  end
end

# Returns whether or not the constant +const+ is defined.  You may
# optionally pass the +type+ of +const+ as <code>[const, type]</code>,
# like as:
#
#   have_const(%w[PTHREAD_MUTEX_INITIALIZER pthread_mutex_t], "pthread.h")
#
# You may also pass additional +headers+ to check against in addition
# to the common header files, and additional flags to +opt+ which are
# then passed along to the compiler.
#
# If found, a macro is passed as a preprocessor constant to the compiler using
# the type name, in uppercase, prepended with 'HAVE_CONST_'.
#
# For example, if have_const('foo') returned true, then the HAVE_CONST_FOO
# preprocessor macro would be passed to the compiler.
#
def have_const(const, header = nil, opt = "", &b)
  checking_for checking_message([*const].compact.join(' '), header, opt) do
    try_const(const, header, opt, &b)
  end
end


-- 
Nobu Nakada