still problem with new/old interrupts on PowerPC and CVS ?

[Ralf Corsepius]

On Sat, 2005-08-06 at 07:55 +1000, Chris Johns wrote:
> Ralf Corsepius wrote:
> > 
> > For the moment, I have changed RTEMS-CVS in such a way that memory.h is
> > always being installed.
> 
> This could be the best solution.

Well, to me, it simply is the least intrusive and least troublesome
solution - I don't consider it to be "a good" nor "the best" solution.

> > 
> > But I am considering to remove it, because of 1.)
> > 
> 
> It maybe a BSD specific include that serves no purpose but the BSD 
> maintainers have not made that change. In OpenBSD the file is also just 
> a single line to include <string.h> and has been since 1995.
I.e. even OpenBSD considers it obsolete for 10 years.

> The only issue with removing the file and changing the BSD code 
> currently in RTEMS is it appearing again in the source tree
The RTEMS sources don't use memory.h anymore, for quite a while.

It's just that the "examples" (These are outside of the source tree) had
not been adjusted, because the problem remained invisible as they had
not been built with RTEMS w/o networking (Then memory.h would have not
been available and the problem would have become apparent).

I.e. if RTEMS had not provided memory.h at all, memory.h would have been
removed from the "examples" long ago.

>  else where 
> after someone ports some new BSD code to RTEMS or upgrades the existing 
> code.
Then these people might want to modernize their code. ;)

RTEMS is trying to follow POSIX. It sometimes providing features from
other standards, from my POV is just a random accident and is subject to
change at any time and a feature user code should not rely upon.

Ralf