updated patch for implicit libcall generation

[Jan Hubicka]

> On Fri, 24 Jan 2003, Jan Hubicka wrote:
> 
> > > On Fri, Jan 24, 2003 at 10:28:07PM +0100, Jan Hubicka wrote:
> > > > ... or are there targets in subdirectories including none of these?
> > > 
> > > Yes.  :-(
> > I was affraid of this answer.  I guess I need to produce include graph
> > of all machine headers and find minimal cut in it.  At least it is not
> > NP complette :)
> 
> There's a case for this configuration always going in the OS or CPU/OS
> headers, rather than e.g. declaring that a CPU will have no systems with
> these functions.  Having it in common OS headers allows easy updating to
> condition on which versions of the OS do/don't have the functions across
> all architectures.  The minimum number of headers needn't be the "correct"
> choice.
> 
> Getting the target deprecations / removals done first reduces the work
> involved.  (And this sort of cross-targets work is one reason for doing
> the deprecations.)
OK, when the deprecation is scheduled?
I can delay it for that.  I probably also should swap the default for
TARGET_MEM_FUNCTIONS at the same time, right?

Honza
> 
> -- 
> Joseph S. Myers
> jsm28@cam.ac.uk