| 24 Aug 2023 |
 flokli | so YMMV | 10:15:41 |
 @linus:schreibt.jetzt | oh right, yeah, if it's just a CDN without extra smarts that makes sense | 10:15:56 |
| @linus:schreibt.jetzt | but depending on how costly requests are, it could make sense to have caching both behind and in front of the assembly bit | 10:16:28 |
| flokli | I still think you should optimize for disk storage on S3 / $backingStore, so small chunk sizes, because everything in the front is cache-able. | 10:16:33 |
| @linus:schreibt.jetzt | at a large scale at least | 10:16:34 |
|  BMG joined the room. | 14:55:34 |
| BMG | Hey, I've been looking into the binary cache protocol today and have noticed that once you do a copy, you can never update the narinfo again. If you sign the path locally with a new key and want to push it, well you can't. | 15:02:47 |
| BMG | Found these related issues https://github.com/NixOS/nix/issues/4221 https://github.com/NixOS/nix/issues/7562 | 15:02:56 |
| BMG | Am I right? | 15:03:03 |
| @linus:schreibt.jetzt | BMG: there's a dedicated nix store copy-sigs command, I wonder if that works? | 15:08:54 |
| BMG | not that i've been able to make work yet | 15:11:03 |
| @linus:schreibt.jetzt | ok, then I'm not sure. But yeah there are a lot of weird things about narinfos in flat-file binary caches | 15:11:45 |
| @linus:schreibt.jetzt | (also paths only being able to have one deriver is weird in general) | 15:12:22 |
| BMG | In reply to @linus:schreibt.jetzt
BMG: there's a dedicated nix store copy-sigs command, I wonder if that works? It seems to be geared towards copying signatures from a remote store into your local store. Can't see a way of copying from local to remote | 15:12:45 |
| @linus:schreibt.jetzt | might work if you pass --store file:///... | 15:13:18 |
| @linus:schreibt.jetzt | (or s3:/// or whatever, as appropriate) | 15:13:27 |
| BMG | That means copying from that store into your local. I'm looking at updating a remote cache after i've signed something again locally | 15:13:55 |
| @linus:schreibt.jetzt | no, --store is the "destination" store | 15:14:19 |
| BMG | Well I don't have a use case, just wanted to confirm that uploading a narinfo is a one and done action. You have to remove it remotely in order to upload again | 15:14:20 |
| @linus:schreibt.jetzt | --substituter is where it's copied from | 15:14:33 |
| BMG | 
Download image.png | 15:14:40 |
| @linus:schreibt.jetzt | to be clear, I'm not sure if it actually works, but using --store should be the right way to express what you want | 15:15:39 |
| BMG | just playing around with it to see | 15:15:52 |
| BMG | nix store copy-sigs -s http://localhost:3000 nixpkgs#hello --refresh --debug is showing the following | 15:17:23 |
| BMG | downloading 'http://localhost:3000/ibpsas4imhv84qmdk5ffh51y0ayrqa94.narinfo'...
starting download of http://localhost:3000/ibpsas4imhv84qmdk5ffh51y0ayrqa94.narinfo
finished download of 'http://localhost:3000/ibpsas4imhv84qmdk5ffh51y0ayrqa94.narinfo'; curl status = 0, HTTP status = 200, body = 1344 bytes
imported 0 signatures
download thread shutting down
| 15:17:35 |
| @linus:schreibt.jetzt | --store, not --substituter | 15:17:42 |
| @linus:schreibt.jetzt | -s is --substituter | 15:17:50 |
| BMG | ah ... fuck me | 15:17:51 |
| BMG | nix sign-paths --store <store> -k <private key> <path> from the first issue was the one that worked. Copy didn't seem to | 15:22:11 |
| @linus:schreibt.jetzt | yeah, I'm not surprised at nix copy not changing the narinfo | 15:23:53 |
