On Fri, 2011-10-14 at 17:02 +0200, Jan Zelený wrote:
> On Fri, 2011-10-14 at 16:10 +0200, Jan Zelený wrote:
> >
https://fedorahosted.org/sssd/ticket/1038
> >
> > With the change I also implemented what Stephen suggested. I schedule all
> > callbacks to fire at tevent_timeval_current() directly, instead of
> > scheduling them one by one.
>
> Nack.
>
> Don't remove the debug message in sss_dp_req_destructor() if
> hash_delete() fails.
>
>
>
> There's no point to the block:
> /* Free any remaining callback */
> if (sdp_req->err_maj == DP_ERR_OK) {
> sdp_req->err_maj = DP_ERR_FATAL;
> sdp_req->err_min = EIO;
> sdp_req->err_msg = discard_const_p(char, "Internal Error");
> }
> Because we're no longer passing this info into callbacks (and sdp_req is
> being freed).
>
>
> Otherwise it looks fine. No obvious regressions while testing, either.
Thanks for that tip, I'm sending updated patch.
Simo made an astute observation off-list. We shouldn't be stealing the
timer event onto NULL. It should move to the parent of sdp_req (which is
rctx). I'm going to make this minor change before committing. Otherwise,
ack.