--- Log opened Wed Dec 22 00:00:08 2021
04:09 < real_or_random> review beg for https://github.com/bitcoin-core/secp256k1/pull/1047 fixing the CI timeouts  will help the other PRs and we should do it before adding more tests 
04:43 < elichai2> real_or_random: partially reviewed, but I don't understand half the things in the yaml file :(
04:45 < real_or_random> elichai2: just replied to your comments
04:50 < elichai2> real_or_random: what's the motivation for not using ECMULT_TABLE_SIZE in the precomputed code?
04:51 < elichai2> just because it's simpler to directly see the window size itself?
04:51 < real_or_random> It's simpler
04:51 < real_or_random> Yeah, it's a monotonic function, so I felt checking the inputs is simpler.
04:52 < real_or_random> It's also more direct. ECMULT_WINDOW_SIZE is the parameter, ECMULT_TABLE_SIZE is just a consequence of it (that happens to be a monotonic function)
04:58 < real_or_random> thanks for the review :)
06:19 < elichai2> Done :)
09:23 -!- elsirion_ [~quassel@gateway/tor-sasl/elsirion] has joined #secp256k1
09:23 -!- elsirion [~quassel@gateway/tor-sasl/elsirion] has quit [Remote host closed the connection]
09:33 < robot-dreams> Silly question, is there some easy way to pay for more hardware for the CI (like some $10 / month plan that gives twice as many CPUs)?
10:00 < BlueMatt[m]> Probably easier to host runners instead if that’s a problem
10:00 < BlueMatt[m]> There’s plenty of places that can host stuff
10:01 < sipa> We really only have issues with macOS CI, which seems much harder to get scheduled on on cirrus
10:11 < BlueMatt[m]> does cirrus support self-hosted runners? not that expensive to go buy a mac mini or two.
10:14 < sipa> Apparently, yes: https://cirrus-ci.org/guide/persistent-workers/
10:15 < sipa> though, we do have lower-hanging fruit; just reducing the number of combinations to test on macOS is not a bad start (https://github.com/bitcoin-core/secp256k1/pull/1048)
10:16 < BlueMatt[m]> seems to me like its probably quicker to put a mac in a back room somewhere than spend eng time rewriting CI, but, 🤷‍♂️
10:17 < BlueMatt[m]> you can also run macos pretty reliably in kvm. its a copyright violation, of course, but....works well
10:54 < real_or_random> robot-dreams:  we could pay 10$/month and get twice the bandwidth but not sure if macos builds will have a lower latency until they get scheduled 
10:56 < real_or_random> https://github.com/bitcoin-core/secp256k1/pull/1047 should bring down the running time on macos from 30 min down to 10 min again 
10:57 < real_or_random> BlueMatt[m]: yeah, a host runner would be nice but setting it up needs eng time, too?
10:58 < BlueMatt[m]> chaincode does have a part-time sysadmin who we can get to do $THINGS, but presumably "plug in mac, run bash script provided by cirrus" isn't all that much work?
10:59 < real_or_random> I agree, I just don't think that reducing the number of combinations to test on macOS is more work ^^
10:59 < real_or_random> a stupid thing with macos on cirrus is that you always get a full VM with 12 CPUs and we only need 1 or 2. We could of course try to run all macOS tests in parallel but that seems to need more eng work
11:03 < halosghost> knowing nothing about cirrus really, if you get a full VM, can you run containers or qemu VMs inside it (essentially pretend like you're given 12 1-core VMs instead of 1 12-core VM?
11:03 < halosghost> (I suspect unless that's something you can setup to happen by-default, there'd be a lot of runner time that would be taken just by the process to bring all those up though)
11:05 < halosghost> that also dramatically increases the complexity of running the CI suite
11:05 < halosghost> just ignore that
11:05 < halosghost> ☺
12:08 -!- lukedashjr [~luke-jr@user/luke-jr] has joined #secp256k1
12:10 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Ping timeout: 240 seconds]
12:11 -!- lukedashjr is now known as luke-jr
12:44 -!- lukedashjr [~luke-jr@user/luke-jr] has joined #secp256k1
12:46 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Ping timeout: 256 seconds]
12:46 -!- lukedashjr is now known as luke-jr
13:16 -!- darosior [~darosior@194.36.189.246] has quit [Quit: darosior]
13:18 -!- darosior [~darosior@194.36.189.246] has joined #secp256k1
13:35 -!- lukedashjr [~luke-jr@user/luke-jr] has joined #secp256k1
13:37 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Ping timeout: 240 seconds]
13:38 -!- lukedashjr is now known as luke-jr
--- Log closed Thu Dec 23 00:00:09 2021