Re: Interix identification.

[Dan Kegel]

Thomas Dickey wrote:
> On Mon, Jul 22, 2002 at 10:22:52AM -0700, Paul Eggert wrote:
>
> > > From: Dan Kegel <address@hidden>
> > >
> > > On Linux, it might be more standards-friendly to do
> > > #define _XOPEN_SOURCE 600
> > > instead of _GNU_SOURCE.  If we're trying to provide an illusion of
> > > uniform behavior, we should avoid enabling GNU extensions, no?
>
> > First, a common case in Autoconf is to use all available extensions.
>
> bad idea (but the notion that the developer should be choosing features
> seems to have been discarded)
>
> > Second, we already tried defining _XOPEN_SOURCE a while ago, and it
> > didn't work well, precisely because it disabled desirable extensions:
>
> that's certainly an understatement (the ifdef's in glibc probably have been
> reasonably well-tested only for the combination that corresponds to
> _GNU_SOURCE - after seeing a number of obvious errors, I decided that it
> must be intentional).

Really?  Do you have a test case?  Maybe I can add a regression test to glibc
for it.  Here, I use

#define _XOPEN_SOURCE 500

to get access to pread() without any trouble.
- Dan