Subject: Re: wrt: various platforms
To: Blair Sadewitz <firstname.lastname@example.org>
From: Julio M. Merino Vidal <email@example.com>
Date: 09/16/2007 23:32:24
On Sep 16, 2007, at 11:26 PM, Blair Sadewitz wrote:
> Now and then, I come across a package which uses configure to test for
> the presence/usability of a certain header file, yet mysteriously does
> not use the definition in config.h in the source itself.
> For instance, X.org's pixman library tests for inttypes.h, but in
> pixman.h, they only include it if __OpenBSD__ is defined.
> Are there cases where this is the right thing to do? Not using
> HAVE_INTTYPES_H when configure tests for it seems so peculiar to me I
> can't help but suspect I'm missing something.
They have probably overlooked something, and I think you should
report this as a bug. Checking for __*__ macros directly from the
code is generally bad, as you cannot predict how systems will be
modified in the future (whereas well-coded configure tests can do a
good job at this) and it is extremely hard to track all their
previous versions and variants straight from preprocessor conditionals.
Julio M. Merino Vidal <firstname.lastname@example.org>