On Wed, 2011-12-14 at 13:03 -0500, Stephen Gallagher wrote:
On Wed, 2011-12-14 at 09:15 -0500, Stephen Gallagher wrote:
>
> On Dec 14, 2011, at 9:09 AM, Jakub Hrozek <jhrozek(a)redhat.com> wrote:
> >
> > Are you going to fix the two comments you had originally in the later
> > patch? Those were don't consider it fatal if the PID isn't found
> > in the hash in sss_child_handler() and reordering sss_child_invoke_cb()
> >
> > If so, then ack
>
>
> The first part of that I already described above as being an incorrect
> reading of the source. I missed the second part and I'll fix that
> shortly.
New patch attached fixes the missing reordering of the
sssd_child_invoke_cb() and adds more explicit handling of unknown PIDs
(with a new DEBUG log message).
I located one very serious bug while working on some dependent patches.
The waitpid() in this patch was using a bitwise AND with the WNOHANG
instead of a bitwise OR (resulting in the WNOHANG being disabled and
causing waitpid() to spin forever in some situations). New patch
attached.
I also moved the child_common routines out of the provider directory and
into util. I did this in a second patch so it's easier to follow the
changes.