On Fri, May 22, 2009 at 10:04:58AM -0700, Niels Provos wrote:
> On Fri, May 22, 2009 at 9:15 AM, Thor Lancelot Simon <t...@panix.com> wrote:
> > If I look at events[8] and events[9], not just the udata is 0 -- all
> > members of the structure are 0. ?Should the kernel ever return events
> > like these?
> That does seem odd.   You would expect at least ident and filter to be
> non-zero.
> How easy is this to reproduce?   Could you provide us with a setup
> where we can reproduce the problem?

I _think_ I can pin it on a bug in NetBSD, but I can't investigate more
fully until Monday as I'm about to fall over tired.

The same executable run "native" on a netbsd-5 system appears to work
properly, while run in a chroot on a netbsd-current system with netbsd-5
libraries it fails as I described (I use the chroot for most of my
testing because the system it's on is the fastest I have).

That makes the most likely possibilities, I think:

        1) The compat code for handling netbsd-5 kevent() syscalls
           under netbsd-current (which has to translate back and
           forth for 64-bit time_t).  But a number of us just looked
           at it and it seems correct.

        2) Some more general bug in kevent in NetBSD-current.  This
           seems more likely to me, and I'm going to try to put
           together a test case over the weekend.

Libevent-users mailing list

Reply via email to