Re: Ubuntu and Debian packages / 2013-07-07

[Graham Lee]

On 9 Jul 2013, at 07:54, "Richard Frith-Macdonald" 
<richardfrithmacdonald@gmail.com> wrote:

> If we define this then presumably there's a bigger chance that they can just 
> take their OSX source and compile it unchanged.

That motivates which of "always pass" and "always fail" we choose, I think. I 
believe it'd be safer to always pick the path corresponding to the earliest 
release, so for a given version 10.x:

#if …_MIN_DEFINED > 10.x_CONSTANT
would always be false, and
#if …_MAX_ALLOWED < 10.x_CONSTANT
would always be true.

In that situation, GNUstep would sometimes report unavailability of API it 
actually has, but would more often default to a path where support is more 
complete. Developers could (and perhaps should be encouraged to) investigate 
what APIs they actually need and are available, and tweak their tests to get 
the best out of GNUstep.

Some people might have written:

#if /*test for API version*/
#error Your Mac OS is too old.
#endif

Those people would get build failures if it works as I outlined above, but I 
still think it's better to be more conservative in how we report compatibility. 
Ideally developers would be testing for availability of specific classes or 
selectors, in which case none of this would be an issue :-).

Graham.