PATCH: don't warn about types defined in casts for sys headers

[Nathan Sidwell]

Brendan Kehoe wrote:

> a system package from a random vendor, they chose to implement something by
> way of a type defined inside a cast.  Since those headers were not available
> for quick fixing, it seemed instead to carry forward into this check the
> idea
> that we shouldn't bitch about what systems (not users) have chosen to do.
I disagree. I'm getting fed up with working around broken system headers. 
The Right Thing is to warn with -fpermissive. The only way system headers are
going to get fixed is if users file bug reports with system vendors - and they
ain't going to do that if they never get a diagnostic.

For instance, the following won't compile on some systems without #defining OK,
	#ifdef OK
	#include <sys/time.h>
	#include <sys/time_impl.h>
	#endif

	#include <pthread.h>

	void foo ()
	{
	  struct timespec timeout;
  
	  pthread_cond_timedwait (0, 0, &timeout);
	}
because struct timespec is declared in a parameter list and its definition
is in sys/time_impl.h which is only included when KERNEL is true - ugh!

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org