Re: [trotts@MIT.EDU: bad line in /usr/include/math.h]

[Chris Faylor <cgf at cygnus dot com>]

Wouldn't the 'extern inline log2()' trick work for c++?

cgf

On Thu, Jul 20, 2000 at 10:42:14PM -0400, DJ Delorie wrote:
>Perhaps using #defines for functions is a bad idea?
>
>------- Start of forwarded message -------
>Date: Thu, 20 Jul 2000 22:14:03 -0400
>To: cygwin@sourceware.cygnus.com
>From: Issac Trotts <trotts@MIT.EDU>
>Subject: bad line in /usr/include/math.h
>
>I got a very cryptic error message while compiling a C++ class library
>called GiNaC (http://www.ginac.de/) that overloads the function log2 
>because of the following sinister line in /usr/include/math.h:
>
>#define log2(x) (log(x) / M_LOG2_E)
>
>I first tried to replace the line with inline versions of log2 for float
>and double but
>the compiler complained that log2 was being redefined.  So I changed the line
>to
>
>#ifdef __cplusplus
>// Do not define log2
>#else
>#define log2(x) (log(x) / M_LOG2_E)
>#endif
>
>and now the code compiles.  Btw, I also put the following code:
>
>#ifdef log2
>#undef log2
>#endif
>
>above the line
>
>unsigned log2(unsigned n); 
>
>in the GiNaC header file that originally had the problem, just in case
>I get another infected math.h file later on.
>
>- -Issac
>
>------- End of forwarded message -------

-- 
cgf@cygnus.com                        Cygnus Solutions, a Red Hat company
http://sourceware.cygnus.com/         http://www.redhat.com/