On Apr 13, 2011, at 9:41 AM, Guy Harris wrote:
> Of course, none of that addresses any configuration tests we might do that involve compiling *and* running code; if we do any of that, cross-compilation is even harder to make work.
And we do, in fact, do tests such as that. For example, we run code to test whether the GLib module mechanism, which we use as a platform-independent mechanism for loading modules (plugins) at run time, is available. If it's always present in GLib 2.x - i.e., if GLib won't compile on a platform that has no module-loading mechanism the GLib module mechanism can use - or if there's a better way to test whether the host-system supports run-time module loading, we could fix that.
We also test whether dladdr() can be used to find a file name for the executable. That one might be harder to do for cross compilation.
There's also a check for a pre-BIND82 inet_pton() bug - perhaps that's so old and crufty that we don't need to care about it any more.
