--- Log opened Sat May 13 00:00:59 2023 01:30 -!- jonatack [~jonatack@user/jonatack] has quit [Quit: WeeChat 3.8] 07:01 < andytoshi> the test run on 1312 succeeded btw 07:02 < andytoshi> i may need to PR to nix to put some sort of memory limit on the parallelism of remote builds, because afaict you just can't control it at all. i think both the remote and local think the other one is responsible for controlling parallelism. so i wound up needing to use a shell loop to submit one job at a time, which is pretty hokey 07:03 < andytoshi> seems like 1tb was more than enough .. before i added widemul to the test matrix :P 07:03 < sipa> ha 07:05 < sipa> Most of that memory usage is from high ecmult window sizes, I assume? 07:11 < andytoshi> yep, from running gcc on the precomputed file 07:11 < andytoshi> interesting clang won't run at all when the window is 21 or higher. (though i only use clang with msan, so maybe it's ok without msan....i should check) 07:12 < andytoshi> "won't run at all" meaning it crashes with a link to file a bug report 07:12 < roconnor> I think both ends let you limit parallelism. local can limit how much they send and remote can limit the total that they work on from everyone. 07:12 < roconnor> Other than that I guess you want to ulimit the memory use somehow. 07:12 < andytoshi> well, ulimit will just cause individual compilations to fail i think 07:13 < roconnor> Right, I was thinking of retrying, but maybe that isn't so easy. 07:13 < andytoshi> roconnor: both ends are documented as being able to limit parallelism, but it appears neither work and there are lots of discourse discussions about this and github issues 07:13 < andytoshi> https://discourse.nixos.org/t/developing-a-system-that-replaces-nix-remote-build/22388 is a good overview of what a mess remote building is rn 07:14 < andytoshi> i'd like to jump in and help, but it seems i'm obligated to first learn about other remote build systems, learn about designing concurrent architectures, etc 07:14 < andytoshi> so i'm waiting for somebody to create a well-defined RFC that has broad support that i can just work against 07:16 < roconnor> that discord link is impressive. 07:18 < andytoshi> there's also this very interesting comment from john ericson suggesting that with io_uring we could have a "pull" model rather than "push" model in the kernel itself https://github.com/NixOS/nix/issues/7091#issuecomment-1257084841 and then it wouldn't really be nix's problem to manage resources 07:18 < andytoshi> but it's just a one-off comment, i'm not aware of anybody working toward that, and the linked LWN article is over my head 08:48 -!- jonatack [~jonatack@user/jonatack] has joined #secp256k1 10:38 < sipa> So, all ready for 0.3.2? 10:41 < real_or_random> yep, I'm merging 10:41 < real_or_random> did we post to bitcoin-dev mailing list for 0.3.1 ? I think we did... 10:41 < real_or_random> merged! 10:42 < sipa> Then tag, then open/merge the follow-up? 10:42 < real_or_random> going ahead with tagging 10:43 < real_or_random> yep, that's what the doc says 10:43 < sipa> 👍️ 10:43 < real_or_random> you want to prepare the PR? 10:43 < sipa> Ok. 10:46 < real_or_random> * [new tag] v0.3.2 -> v0.3.2 10:46 < sipa> https://github.com/bitcoin-core/secp256k1/pull/1314 10:46 < hebasto> 🚀 10:48 < sipa> Now the release on github. 10:48 < real_or_random> who? 10:48 < sipa> On it. 10:48 < real_or_random> nice 10:49 < real_or_random> hebasto: wanna review https://github.com/bitcoin-core/secp256k1/pull/1314/files if you're around? 10:49 < hebasto> on it :D 10:50 < sipa> https://github.com/bitcoin-core/secp256k1/releases/tag/v0.3.2 10:52 < real_or_random> nice@ 10:52 < real_or_random> ! 10:52 < sipa> All the links work even. 10:53 < real_or_random> hehe, getting better slowly 10:55 < real_or_random> merged 10:56 < real_or_random> ok. I think the only question remains is if we want to write an email 10:56 < real_or_random> I think we want. it's sad that we *need* to do it again but yeah... +shrug+ 10:56 < sipa> You want to? 10:56 < sipa> I can do it also. 10:57 < sipa> I agree we should. 10:57 < real_or_random> I need to leave for dinner 10:57 < real_or_random> happy to do it but if you're around anyway 10:57 < sipa> Going to unpack a bit more. We can discuss a bit later. 10:58 < real_or_random> ok 10:58 < real_or_random> let's try to do it later 10:59 < real_or_random> (I opened https://github.com/bitcoin-core/secp256k1/pull/1313 so we can already honestly claim we're working on testing improvements for fresh compilers ^^) 11:00 < real_or_random> "Container errored with 'OOMKilled'" ok, so much about that... :D 12:24 < sipa> So, mail? 13:12 -!- midnight [~midnight@user/midnight] has joined #secp256k1 --- Log closed Sun May 14 00:00:00 2023