[PATCH] alpha bits/mman.h vs mman-linux.h

[Joseph S. Myers]

On Sat, 7 Dec 2013, Richard Henderson wrote:

> Joseph mentions in the wiki that alpha isn't using <bits/mman-linux.h>.  I'd
> been avoiding this because mman-linux.h isn't quite as configurable as it might
> be, but here's a first cut.
> 
> Is this reasonable, with the undefs?  Ought I play more with __FOO as we
> already do for __MAP_ANONYMOUS?

My preference is use of __FOO in such cases, at least if whether the macro 
should be defined depends on feature test macros, so that the 
architecture-specific files don't need to duplicate the logic about which 
feature test macros enable which header definitions.  (Of course, if the 
symbol, not just its value, is architecture-specific, then you can't avoid 
such tests in the architecture-specific files.)

-- 
Joseph S. Myers
joseph@codesourcery.com