On Wed, Aug 31, 2022 at 02:13:51PM +0200, Ævar Arnfjörð Bjarmason wrote: > Removing coccinelle rules because we're seeing slowness somewhere seems > particularly short-sighted to me. > > Maybe we do run into intractable problems somewhere with it being slow, > and we'd also like to cater to more "interactive" use. Agreed. I'm not wild about how long it takes to run either, but if it's producing useful results, it seems worth it to pay the CPU (and I think unused.cocci did find some useful results already). There's a point at which the CPU use becomes intractable, but I don't think we're there yet. > There's nothing in unused.cocci that we either aren't running into > elsewhere, or wouldn't run into if we had 10x the coccinelle rules we > have now (which I think would be a good direction, we should rely on it > more heavily). >From past experience, I suspect the "<... ...>" operator is what's expensive. I don't see an easy way of avoiding it here, though. I'm more skeptical on more coccinelle in general, just because I've spent so many hours fighting with it for both output and performance reasons. But if somebody else is willing to do that work, I'm OK with it. I have often wondered if our rules are sufficiently simple that libclang plus some light scripting might get us similar results with less hassle. But maybe that's a rabbit hole. -Peff
