On Thu, Jul 11, 2013 at 09:13:15AM -0400, Ed Maste wrote: > On 10 July 2013 19:03, Ben Pfaff <b...@nicira.com> wrote: > > (The usual reason for ppoll() is to ensure atomicity of changing the signal > > mask and blocking, but time_poll() does not need atomicity. Therefore, > > even a userspace emulated ppoll() would work here in a race-free manner, if > > the kernel does not support ppoll().) > > And we'll need that userspace emulated ppoll() for non-Linux systems; > seem reasonable to just implement it in timeval.c inside an #ifndef > HAVE_PPOLL from autoconf?
Hmm, I had this notion that ppoll() had been standard for a long time now, but I see that it's not even in SUSv4. I'm dropping this patch for now since the rationale is based entirely on speculation, not a good reason to reduce portability. Thanks, Ben. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
