> > > We're all building those grass runways on the island.
> >
> > Does that mean the current structure is necessary?
>
> I based it on the other code I'd seen. I'll admit that I don't really
> fully understand the private contract interfaces.
>
> I spent some time testing this, and I found that I don't need to do
> that at all. Both the contract_latest() and the contract_abandon_id()
> work fine if done after the child has exited. I used 'ps -o ctid' to
> show that I still had a separate contract for dhcpagent, and 'ctstat'
> to show that the contract parameters were all the same.
Great, the new code is what I had in mind.
> My update for this review comment (simplifying the code) is:
>
> diff -r 14c9daa6faaf usr/src/lib/libdhcpagent/common/dhcpagent_util.c
> --- a/usr/src/lib/libdhcpagent/common/dhcpagent_util.c Fri May 15
> 08:41:49 2009 -0400
> +++ b/usr/src/lib/libdhcpagent/common/dhcpagent_util.c Fri May 15
> 09:52:38 2009 -0400
> @@ -152,10 +152,6 @@
>
> childpid = fork();
>
> - /* parent needs contract id */
> - if (childpid != 0 && childpid != (pid_t)-1 &&
> - contract_latest(&ct) == -1)
> - ct = -1;
> (void) ct_tmpl_clear(ctfd);
> (void) close(ctfd);
>
> @@ -173,7 +169,8 @@
>
> /* wait for the daemon to run and then abandon the contract */
> (void) waitpid(childpid, NULL, 0);
> - if (ct != -1)
> +
> + if (contract_latest(&ct) != -1)
> (void) contract_abandon_id(ct);
>
> while ((timeout != -1) && (time(NULL) - start_time < timeout)) {
--
meem
