On Wed, Nov 27, 2019 at 06:21:01PM +0200, Lauri Tirkkonen wrote: > All that said I do understand if there is reluctance to merge the > jobserver stuff since it doesn't actually help the current situation in > most cases. Nevertheless it has been personally beneficial to me in > identifying areas of improvement, even if those are nothing new to > OpenBSD developers.
Oh, I'm not opposed to merging a jobserver once it's roughly as stable as what we have. My concerns are: - it should work with all of src and xenocara (once that's done, I can do ports) - it should not interfere with other jobservers in other makes. - it does not replace the expensive jobs heuristics. - it does not render make harder to maintain and evolve. Especially since, as I've said before, there's no actual guarantee anywhere in POSIX that shells will cooperate with it. I don't have enough time to do everything I want, but it's well known that there are still issues with make -j compared to sequential make. If you look at cvs history, I've completely rewritten job.c a few times. And I had a presentation at eurobsdcon about what's still missing a few years ago. I've been swamped with changes in dpb/bsd.port.mk/pkg_add... recently, but it's still something I want to get back to eventually ! One reason that work goes slowly is because make was a seriously entwined series of things that all depend on each other... It still is, but slightly less so, and thus anything that's not squeaky clean is unlikely to get in. (hence me wanting the jobserver stuff to be reasonably self-contained and not interfering with the rest of make)
