On Fri, Feb 01, 2019 at 06:29:22PM +0100, Florian Westphal wrote: > Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote: > > Moreover, this patch adds the unbind step to deliver the event from the > > commit path. This should not be done from the worker thread, since we > > have no guarantees of in-order delivery to the listener. > > Right, they indeed should occur during finalization of commit phase, not > worker. > > > This patch removes the assumption that both activate and deactivate > > callbacks need to be provided. > > The assumption was that deactive has side effects that might need to > be undone in case another rule in the batch triggers an abort. > > AFAICS after this patch .deactivate has no effect if we're still > in the preparation phase so no reactive needed either in case of abort. > > IOW, looks good to me, thanks for fixing this. Thanks Florian, just noticed a small issue with event order (DELSET should come after DELRULE for anonymous sets as it uses to be before this patch) just sent a v3.
