Re: Updated I18N proposal

On Thu, May 11, 2006 at 04:56:14PM -0400, Jeffrey Hutzelman wrote:
> Right, and my point is that "compile time detection" cannot consist of 
> compiling and running a test program that uses the run-time detection to 
> tell you if the feature is available, because that doesn't work in 
> cross-compile environments.
> 
> It turns out that the art of build-time feature detection is fairly well 
> understood, at least in the UNIX/C world, and not breaking it is nice.  It 
> turns out that tests that check for existence of a type, function, 
> preprocessor symbol, etc are fairly easy to do.  Tests that check the value 
> of a preprocessor symbol are a bit harder, and tests that require that you 
> _call_ a function and check its behavior or return value are very poor.

I see this point, but C code that would use functions that may
optionally do nothing and return errors SHOULD always do run-time
detection of this and SHOULD NOT rely on compile-time detection of such
features.

Again, I'm not remotely opposed to have pre-processor defines for
compile-time detection of features.

Nico
-- 

_______________________________________________
Kitten mailing list
Kitten@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/kitten