| 8 Oct 2021 |
 Jonas Chevalier | nix-build --store is not reliable, I think it needs to have write access to the store to work as expected, which wouldn't work with cachix. | 07:53:14 |
 colemickens π³οΈβπ | yeah, :P I really don't quite understand how it's meant to be used for that scenario anyway. I guess my assumption was that they've made it so that it will notice the store being readonly and fallback to the "auto" worker? seems like the "auto" keyword has been getting some work? | 07:56:09 |
| colemickens π³οΈβπ | (otherwise I'm not sure why it was at all suggested as a solution? not to be dismissive of the work, I assume I just don't understand) | 07:56:47 |
| colemickens π³οΈβπ | * (otherwise I'm not sure why it was at all suggested as a solution on that Issue? not to be dismissive of the work, I assume I just don't understand) | 07:56:56 |
 Mic92 | Personally I moved on to hydra, in cases I would have need nix-build-uncached. I would pick it up, I would go for https://github.com/nix-community/nix-eval-jobs which is faster due to use of threads. | 09:15:18 |
| Mic92 | I only use it for my nur-packages | 09:15:54 |
| Mic92 | I am actually trying to make hydra more accessible: https://github.com/NixOS/hydra/pull/1033 | 09:16:11 |
| Mic92 | Once you have documentation on how to declarative add jobsets it's probably the fastest way to evaluate nix projects. | 09:17:05 |
| Mic92 | And than there is hercules if you want better github integration | 09:17:44 |
|  fedx joined the room. | 15:20:21 |
| 9 Oct 2021 |
| colemickens π³οΈβπ | I like to dream about something better. | 00:28:30 |
| colemickens π³οΈβπ | a CI system that knew about flake inputs, knew about upstreams for the flake inputs, knew about "upgrade groups" of flake inputs -> it could list outputs, update all inputs in an "upgrade group", relist outputs, build the new outputs, move on to next upgrade group. You could have rich data about your flake input upgrades over time, you could have a smarter flake upgrade process than I've seen anyone implement so far... | 00:30:13 |
| colemickens π³οΈβπ | that doc pr is great too | 00:30:18 |
| colemickens π³οΈβπ | I had actually done all this once, I even have scripts that wipe/resetup hydra/rePOST my initial declarative job, etc. | 00:30:42 |
| colemickens π³οΈβπ | But I don't like hydra. | 00:30:45 |
| Jonas Chevalier | In reply to @colemickens:matrix.org
anyway, I think I figured out the issue, I need to be more forceful about having cachix push drvs up to get the behavior I want I just remembered, you might be hitting the allowSubstitutes = false issue. See https://github.com/Mic92/nix-build-uncached#packages-with-allowsubstitutes--false | 06:26:32 |
| colemickens π³οΈβπ | !! that definitely looks relevant, I was wondering about that, because a lot of the paths it needs to build are the service/text file type of things. | 06:27:25 |
| colemickens π³οΈβπ | thank you! | 06:27:26 |
| colemickens π³οΈβπ | I'm not even sure what I'm looking at in this lib -_- | 06:53:39 |
 Domen KoΕΎar | allowSubstitues are the worst, we should fix the efficiency of binary caches instead of adding exceptions manually | 17:17:36 |
| colemickens π³οΈβπ | does someone have an example of using this workaround in their actual nix(os)-config? https://github.com/Mic92/nix-build-uncached#packages-with-allowsubstitutes--false | 20:08:04 |
| colemickens π³οΈβπ | just to fully wrap my head around it? | 20:08:15 |
| Mic92 | AFAIK zimbatm in some internal repo | 20:13:43 |
| Jonas Chevalier | there is an example in the repo: https://github.com/Mic92/nix-build-uncached/blob/master/scripts/force_cached_example.nix | 20:15:44 |
| Jonas Chevalier | the core idea is that nix ignored allowSubstitutes = false if the derivation is a dependency of another one. So we find all these in the passed attrset and recursively put a little derivation in front. | 20:17:41 |
| Jonas Chevalier | * the core idea is that nix ignores allowSubstitutes = false if the derivation is a dependency of another one. So we find all these in the passed attrset and recursively put a little derivation in front. | 20:17:51 |
| 10 Oct 2021 |
| colemickens π³οΈβπ | does it have to be a symlink? | 02:43:45 |
| colemickens π³οΈβπ | it'd be nice to not need coreutils for that, just for ease/readability of having to get a sys/pkgs around to get coreutils | 02:44:20 |
 nix-community-bot | [nix-community/infra] Mic92 pushed to update/nixpkgs-6fa4e75f7622ccedee8cb58a8d2ecaa658667a10: morph: substitute on remote hosts - https://github.com/nix-community/infra/commit/cf51a421385e083f0002dc8ea093553796309ea6 | 08:53:45 |
| nix-community-bot | [nix-community/infra] web-flow pushed to update/nixpkgs-6fa4e75f7622ccedee8cb58a8d2ecaa658667a10: niv nixpkgs: update 6fa4e75f -> 107a51dd - https://github.com/nix-community/infra/commit/85b70c6fab3e9600622597297212f64120784b3a | 08:53:46 |
