| 8 Jul 2026 |
 Grimmauld (any/all) | but its a 6 core 12 thread laptop, that makes some sense | 23:36:24 |
| Grimmauld (any/all) | gives 12 | 23:36:33 |
 hexa | hm, does nix pick physical cores? | 23:36:34 |
| Grimmauld (any/all) | i have no idea | 23:36:50 |
| Grimmauld (any/all) | i take more issue with the fact i can't find where these env vars are populated in the first place | 23:37:11 |
| hexa |
The value 0 means that the builder should use all available CPU cores in the system.
| 23:37:13 |
| Grimmauld (any/all) | like, setting these in environment is good and all, but unless we know where these are currently populated we can't be certain what we put in env won't be overridden later | 23:38:08 |
| Grimmauld (any/all) | and finding that will be a goose chase i fear... | 23:38:28 |
| Grimmauld (any/all) | Anyways. I will sleep, pgmpy doing weird things here is more likely caused by not enough thread control for things rather than my hook being wrong. The fix is more than i am willing to do at 2am. | 23:40:55 |
| Grimmauld (any/all) | geese will be chased tomorrow 🪿 | 23:41:22 |
| hexa | I see none of those in pgmpy checkPhase or installCheckPhase | 23:51:59 |
| 9 Jul 2026 |
| Grimmauld (any/all) | How did you look? Abort with a hook, or check /proc/<pid>/environ? | 06:30:46 |
| Grimmauld (any/all) | those env vars are only observable in the latter, suggesting they aren't set in nix | 06:31:14 |
| Grimmauld (any/all) | https://github.com/ryfeus/lambda-packs/blob/74b21b9cddb57790c804ab973e26141eb7e5c912/Sklearn_arm/source/joblib/_parallel_backends.py#L40-L44
pgmpy pulls joblib as a dependency, and that has this list... This looks sus | 06:46:33 |
| Grimmauld (any/all) | https://github.com/ryfeus/lambda-packs/blob/74b21b9cddb57790c804ab973e26141eb7e5c912/Sklearn_arm/source/joblib/_parallel_backends.py#L162-L177# | 06:47:08 |
| Grimmauld (any/all) | will check whether its in our jobliv too, this was the first i found via codesearch | 06:48:07 |
| Grimmauld (any/all) | https://github.com/joblib/joblib/blob/b030e4e1ed6a227c0e587c31b266c03b3b692372/joblib/_parallel_backends.py#L59-L67
pgmpy pulls joblib as a dependency, and that has this list... This looks sus | 06:48:49 |
| Grimmauld (any/all) | * https://github.com/joblib/joblib/blob/b030e4e1ed6a227c0e587c31b266c03b3b692372/joblib/_parallel_backends.py#L221-L248 | 06:49:13 |
| Grimmauld (any/all) | will check whether its in our jobliv too, this was the first i found via codesearch | 06:50:14 |
| Grimmauld (any/all) | And this very much looks like it respects existing parent environment if set, meaning defining those in a hook will work | 06:51:33 |
| Grimmauld (any/all) | I propose https://github.com/NixOS/nixpkgs/pull/539853 as a more aggressive form of what i had before with openmpCheckPhaseHook, that should catch more | 08:03:45 |
| Grimmauld (any/all) | build is still building though, can't yet make a statement about how effective this is yet | 08:04:12 |
