| 26 Aug 2021 |
nrdxp | I have that one in the picture there | 16:52:49 |
nrdxp | It works well | 16:52:51 |
@andi:kack.it | What are the specs of build01? | 16:52:51 |
Jonas Chevalier | AMD Ryzen 7 1700X Eight-Core Processor | 16:53:18 |
Jonas Chevalier | 64GB RAM | 16:53:30 |
[0x4A6F] | https://github.com/nix-community/infra#build01- | 16:53:34 |
Jonas Chevalier | maybe we should start our own opencollective | 16:55:13 |
ryantm | Yubikey ordered 5 NFC ordered! | 16:56:01 |
ryantm | * Yubikey 5 NFC ordered! | 16:56:07 |
@andi:kack.it | Is there enough demand for x86_64-linux capacity? For darwin and aarch64 I can understand that there is demand (but I wouldn't donate if the money goes towards apple). | 16:57:09 |
Jonas Chevalier | https://github.com/nix-community/infra/issues/100 | 16:59:53 |
Jonas Chevalier | I think darwin is a bit of a dead end | 17:00:17 |
Jonas Chevalier | for a while it was possible to run darwin in a VM, on a Hetzner box | 17:00:42 |
Jonas Chevalier | it's pretty easy to do so | 17:00:51 |
Jonas Chevalier | the aarch64 and T2 changes are going to make that harder | 17:01:07 |
@andi:kack.it | but not really legal :) | 17:01:13 |
Jonas Chevalier | depends on the country | 17:01:27 |
Jonas Chevalier | are EULA enforceable in Germany? :) | 17:01:45 |
@andi:kack.it | I don't know. I avoid apple as much as possible. They are parasites and don't deserve free work., | 17:02:06 |
Jonas Chevalier | I feel the same for most cloud providers | 17:02:53 |
Sandro | As long as we can ignore Darwin build failures on most PRs I don't care to much | 17:27:13 |
Sandro | Especially for some leaf packages some Darwin user needs to speak up and fix it or provide a patch or a failed test at least | 17:27:49 |
Sandro | And for x86_64-linux: I can't really speak because I have enough build power right now. | 17:28:24 |
ryantm | I think the 60 euro per month server I have is enough for r-ryantm. At least until I do more work to parallelize updates. | 17:29:24 |
ryantm | * I think the 60 euro per month server (build02) I have is enough for r-ryantm. At least until I do more work to parallelize updates. | 17:33:36 |
@andi:kack.it | Some intelligent build scheduling across the machines (self-hostable nixbuild.net?) would be lovely for that work. Of course you can add all the machines to all the other machines as remote builders but that isn't really that great in my experience. | 17:44:54 |
@andi:kack.it | ryantm: hows does r-ryantm work? Does it go one package at a time and waits for the builds to finish? | 17:45:19 |
ryantm | andi-: Currently it has three sources Repology, GitHub releases, PyPi. Whenever it starts a new batch, it queries those and makes a big list of stuff to work on, then it goes through the list one at a time. Builds are usually not the slowest thing. The slowest things are nixpkgs-review and figuring out how many rebuilds a change causes. | 17:58:12 |
ryantm | Recently, I just made it run those three sources in parallel. | 17:58:38 |
ryantm | I have lots of ideas for how to make that better, its just a matter of having the time for it. | 17:59:56 |