| 4 Nov 2024 |
 Bot_wxt1221 | So could we solve the problem of evaluator? | 00:00:34 |
 Tristan Ross | In reply to @bot-wxt1221:matrix.org
Anyway, I don't think we should discuss about a bug which doesn't know if it exist. Afaik, @k900:0upti.me has mentioned impurities but idk if that affects the evaluator | 00:00:43 |
| Bot_wxt1221 | Tristan RossRun a full VM will solve it.
| 00:01:22 |
| Bot_wxt1221 | Even though it is really slow, it is better than nothing. | 00:01:48 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
Tristan RossRun a full VM will solve it.
Of aarch64-darwin? | 00:02:35 |
| Bot_wxt1221 | In reply to @rosscomputerguy:matrix.org
Of aarch64-darwin? I mean evaluate on aarch64-linux | 00:03:02 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
So could we solve the problem of evaluator? Get more aarch64-darwin machines | 00:03:07 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
I mean evaluate on aarch64-linux Ok but if there are impurities problems then that would still cause issues | 00:03:30 |
| Tristan Ross | And emulation isn't the solution | 00:03:44 |
| Bot_wxt1221 | So we can run x86_64-linux VM on aarch64-linux | 00:03:48 |
| Tristan Ross | Too inefficient | 00:03:50 |
| Bot_wxt1221 | It is still better than no. | 00:03:57 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
So we can run x86_64-linux VM on aarch64-linux Why | 00:04:05 |
| Bot_wxt1221 | In reply to @rosscomputerguy:matrix.org
Why Because you are worried about impurities | 00:04:27 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
Because you are worried about impurities That wouldn't solve the problem still | 00:06:17 |
| Tristan Ross | Emulation isn't efficient | 00:06:36 |
| Tristan Ross | The system should match the hardware | 00:06:54 |
| Bot_wxt1221 | However, it is less efficient if aarch64-linux builder do nothing now. | 00:07:42 |
| Tristan Ross | True but doesn't x86_64-linux also have the same thing of not doing much? | 00:12:03 |
| Tristan Ross | It's hard to assign work to machines which inherently don't support the target. | 00:13:00 |
| Bot_wxt1221 | Redacted or Malformed Event | 00:13:26 |
| Bot_wxt1221 | In reply to @rosscomputerguy:matrix.org
It's hard to assign work to machines which inherently don't support the target. Does evaluating have something to do with arch? | 00:18:38 |
| Tristan Ross | In reply to @bot-wxt1221:matrix.org
Does evaluating have something to do with arch? It does affect builtins.currentSystem. | 00:19:26 |
| Tristan Ross | K900 could probably explain things better since they were the one who told me about impurities | 00:21:18 |
 K900 | There shouldn't be any impurities wrt evaluation | 05:41:13 |
| K900 | But it's also not trivial to reassign machines like that for technical reasons AFAIK | 05:41:32 |
 Lily Foster | In reply to @rosscomputerguy:matrix.org
Idk, using aarch64-linux to eval for aarch64-darwin doesn't feel right lol. Ideally it should evaluate the same but what if it doesn't? restricted eval mostly prevents differences. x86_64-linux already evaluates for all 4 arches in ofborg | 13:07:53 |
 ElvishJerricco | hm, ofborg blew up on this pr on x86_64-linux but not aarch64-linux: https://github.com/NixOS/nixpkgs/pull/353543 | 20:38:57 |
| Tristan Ross | In reply to @lily:lily.flowers
restricted eval mostly prevents differences. x86_64-linux already evaluates for all 4 arches in ofborg Interesting, is there a single machine which only does all evals or does it rotate out? | 22:40:17 |
| 5 Nov 2024 |
| Lily Foster | In reply to @rosscomputerguy:matrix.org
Interesting, is there a single machine which only does all evals or does it rotate out? i mean there are 15 evaluators according to https://events.ofborg.org/stats.php | 00:54:59 |
