apr/test/testatomic.c trunk doesn't compile due to implicit, as apr_atomic_xchgptr is not defined in os390/atomic.c,
doesn't have anything to do with apr_atomic_casptr.

So next step would be implement xchgptr in os390/atomic.c, or ifdef it out for __MVS__ where ever it exists (should only be trunk of apr/test/testatomic.c afaik).

On 10/16/07, Jeff Trawick <trawick@gmail.com> wrote:
On 10/16/07, David Jones <oscaremma@gmail.com> wrote:
> Create a z/OS implementation of apr_atomic_casptr.

> Passes apr/test/testatomic at trunk, at 554294 level, which does not include
> the following test of apr_atomic_xchgptr (which does not yet have a zOS
> implementation, but is also not in httpd 2.2.6):
>           Revision 554995 - (view) (download) (annotate) - [select for
> diffs]
>  Modified Tue Jul 10 16:35:45 2007 UTC (2 months, 4 weeks ago) by davi
> File length: 13674 byte(s)
> Diff to previous 554294 (colored) Introduce apr_atomic_xchgptr, which
> atomically exchanges a pair of pointer
> values. Missing OS/390 implementation.

I'm confused about why you're telling us this ;)

Are you saying that the current apr/test/testatomic from trunk doesn't
work on z/OS?

Does it fail even without your patch to add apr_atomic_casptr(), or
does it start failing when you add apr_atomic_casptr() either because
of a testcase problem or a restriction that apr can't implement one
function without implementing the other, or ???
Born in Roswell... married an alien...