On Fri, Jan 24, 2003 at 01:01:20AM +0100, Klaus Klein wrote:

 > Actually... these should not be conditional on a kernel configuration
 > option at all, or the _POSIX_SEMAPHORES symbol ought to be defined as
 > zero in <sys/unistd.h> (and applications check run-time variables).
 > Given the low overhead of the module (unlike SysV IPC), the former
 > might make more sense.

Hm, okay.

FWIW, the current _ksem_*() stuff is totally broken.  It was clearly
not tested before it went in.  I'm currently shaking the remaining
bugs out of a near-rewrite.  Sadly, it adds a touch more overhead
in order to provide the POSIX semantics.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>