 | DevOS | 33 Members |
| Seeking help and geeking out together on https://github.com/divnix/devos & https://github.com/divnix/digga | 10 Servers |
| DevOS | 33 Members |
| Seeking help and geeking out together on https://github.com/divnix/devos & https://github.com/divnix/digga | 10 Servers |
| Sender | Message | Time |
|---|---|---|
| 10 Jul 2021 | ||
 David Arnold (blaggacao) | I'm currently transitioning to set up my hm, so this is might be a good time window to share ideas / findings. 🙂 | 00:54:12 |
| David Arnold (blaggacao) | I'm still not 100% settled on a conlusion, but right now after some discussions with Pacman99 I'm thinking of
Does that +- make sense? | 01:01:00 |
| 11 Jul 2021 | ||
 fufexan | In reply to @blaggacao:matrix.orgsorry, I wasn't online yesterday, but yes, I'll review it :) | 12:06:06 |
| fufexan | David Arnold: is it necessary to use homeConfigurations in the form user@host as described here? https://github.com/divnix/digga/pull/84/files#diff-34527206e391faf315867402419cb48d33a1351f2b245863dd78a7d521cfd9c1R4-R10 | 12:16:39 |
| fufexan | maybe I don't really understand how it works | 12:17:15 |
| David Arnold (blaggacao) | | 12:18:39 |
| David Arnold (blaggacao) | (hm docs) | 12:18:43 |
| David Arnold (blaggacao) | So this means that the home-manager utility has knowledge of the current user and hostname. Btw. that's quite similar to divnix/bud here, but I digress. | 12:20:05 |
| fufexan | oh, right. this makes sense now | 12:20:30 |
| fufexan | in that case, it looks fine | 12:20:38 |
| David Arnold (blaggacao) | I think the key take away is, that we still want to have a "portable" user variant that uses $HOME/.nix-porfile and is completely evaluated without any access to any host config. | 12:21:50 |
| David Arnold (blaggacao) | For example, without such host related stuff | 12:22:47 |
| fufexan | true, that would be nice to have | 12:23:27 |
 rosariopulella joined the room. | 15:46:00 | |
 @timdeh:matrix.org | I dunno if I like that. I like how the current setup can share configuration between home-manager and nixos, even if using hm standalone. Why would we give that up? | 17:35:07 |
| David Arnold (blaggacao) | I don't see us giving that up: a In fact and as far as I can tell right now, the only requirement for a hm config to work portably is to set | 21:03:10 |
| David Arnold (blaggacao) | * I don't see us giving that up: a `hm` config can still be created independently of any specific host, and the same config then can still be deployed to any nixos host. In fact and as far as I can tell right now, the only requirement for a hm config to work portably on a non-nixos host is to set `useUserPackages = false` while setting it to true on a nixos host seems to be a good idea. | 21:03:49 |
| David Arnold (blaggacao) | * I don't see us giving that up: a `hm` config can still be created independently of any specific host, and the same config then can still be deployed to any nixos host. In fact and as far as I can tell right now, the only requirement for a hm config to work portably on a non-nixos host is to set `useUserPackages = false` while on the other hand setting it to true on a nixos host seems to be a good idea. | 21:04:10 |
| David Arnold (blaggacao) | * I don't see us giving that up: a `hm` config can still be created independently of any specific host, and the same config then can still be deployed to any nixos host. In fact, and as far as I can tell right now, the only requirement for a hm config to work portably on a non-nixos host is to set `useUserPackages = false` while on the other hand setting it to true on a nixos host seems to be a good idea. | 21:04:32 |
| David Arnold (blaggacao) | I also think ot might be prohibitive for a "portable" user to depend on nixos host config values of any particular host. However, currently this seems to be our thought model which the PRs propose to rectify. | 21:05:50 |
| David Arnold (blaggacao) | * I also think it is prohibitive for a "portable" user to depend on nixos host `config` values of any particular host. However, currently this seems to be our thought model which the PRs propose to rectify. | 21:06:06 |
| David Arnold (blaggacao) | * I also think it is prohibitive for a "portable" user to depend on nixos host `config` values of any particular host. However, currently this seems to be our thought model which the PRs seeks to rectify. | 21:06:41 |
| David Arnold (blaggacao) | In fact, I think our current implementation simply ignores the difference between a host-specific application of a hm config and a "portable" variant of the same config and treats them as the same which might result in intricate problems in either use case. | 21:08:04 |
| David Arnold (blaggacao) | * In fact, I think our current implementation simply ignores the difference between a host-specific application of a `hm` config and a "portable" variant of the same config and treats them as the same which might result in intricate problems in either use case. Such problems manifest when a `useUserPackes = false` - nixos hm module suddenly is used in conjunction with a `userUserPackages = true` config. However, this is currently the case if you would use the `home-manager` cli in good faith with the current inplementation | 21:10:45 |
| David Arnold (blaggacao) | * In fact, I think our current implementation simply ignores the difference between a host-specific application of a `hm` config and a "portable" variant of the same config and treats them as the same which might result in intricate problems in either use case. Such problems manifest when a `useUserPackes = false` - nixos hm module suddenly is used in conjunction with a `userUserPackages = true` config. However, this is currently the case if you would use the `home-manager` cli in good faith with the current implementation. | 21:10:57 |
| David Arnold (blaggacao) | * In fact, I think our current implementation simply ignores the difference between a host-specific application of a `hm` config and a "portable" variant of the same config and treats them as the same which might result in intricate problems in either use case. Such problems could manifest when a `useUserPackes = false` - nixos hm module suddenly is used in conjunction with a `userUserPackages = true` config. However, this is currently the case if you would use the `home-manager` cli in good faith with the current implementation. | 21:11:24 |
| David Arnold (blaggacao) | Hence I think it is better to provide a fully nixos-compliant hm config as user@host so that good faith use of home-manager cli at least does not do any potential harm. And render a separate user with useUserPackages = mkForce false so that this config is guaranteed to work on non-nixos hosts. | 21:13:10 |
| David Arnold (blaggacao) | * Hence I think it is better to provide a fully nixos-compliant `hm` config as `user@host` so that good faith use of `home-manager` cli at least does _not_ do any potential harm. And render a separate `user` with `useUserPackages = mkForce false` so that this config is guaranteed to work on non-nixos hosts. | 21:13:34 |
| David Arnold (blaggacao) | * Hence I think it is better to provide a fully nixos-compliant `hm` config as `user@host` so that good faith use of `home-manager` cli at least does _not_ do any potential harm. And render a separate `user` with `useUserPackages = mkForce false` so that such config is guaranteed to work "portably" on non-nixos hosts. | 21:14:02 |
| David Arnold (blaggacao) | Lastly, one might want to deploy a user (normally used on 86_64-linux) on an aarch64 server. user@host would not help, since that would have 86_64-linux. To successfully "portably" use that user on an aarch64 machine, I'd have to either create a dummy aarch64 host or render a user-acrch64 output for homeConfigurations. | 21:17:24 |