| 27 Oct 2025 |
 fzakaria | it seems to hang on the download.tar.gz | 18:56:06 |
 toonn | After already having run the build with the correct hash once? | 19:08:55 |
|  Garrett joined the room. | 20:10:09 |
|  backtail joined the room. | 20:29:45 |
| backtail | Hey peeps, I unknowingly ran into the leaveDotGit bug of unpredictable hash generation for nix derivations. I patched the dependency, but now ofborg says that my provided hash is wrong:
´´´bash
error: hash mismatch in fixed-output derivation '/nix/store/q59g9bk2kwh20sr8ir00gz5d5yjkb75r-source.drv':
specified: sha256-HoqFBkxXlmbWaFPOVvO0zUhlDBVzz++T6Fgf0MAuyao=
got: sha256-rcIbtIXTm1sNMCbs6W5DS/9KeMkWWhUdwReW++nW2Qc=
error: 1 dependencies of derivation '/nix/store/q5izjyv5iv1h0zdc06my4dp07wvv9j76-KnobKraft-orm-2.7.1.drv' failed to build
´´´
The point that I don't understand is that every architecture, aarch64, x64_86, etc. all have different hashes. However, I just want to build it for x64_86_linux. So how do I tell ofborg to only build on one architecture and why does the hash that works locally not work with ofborg?
| 20:33:55 |
| backtail | * Hey peeps, I unknowingly ran into the leaveDotGit bug of unpredictable hash generation for nix derivations. I patched the dependency, but now ofborg says that my provided hash is wrong:
´´´
error: hash mismatch in fixed-output derivation '/nix/store/q59g9bk2kwh20sr8ir00gz5d5yjkb75r-source.drv':
specified: sha256-HoqFBkxXlmbWaFPOVvO0zUhlDBVzz++T6Fgf0MAuyao=
got: sha256-rcIbtIXTm1sNMCbs6W5DS/9KeMkWWhUdwReW++nW2Qc=
error: 1 dependencies of derivation '/nix/store/q5izjyv5iv1h0zdc06my4dp07wvv9j76-KnobKraft-orm-2.7.1.drv' failed to build
´´´
The point that I don't understand is that every architecture, aarch64, x64_86, etc. all have different hashes. However, I just want to build it for x64_86_linux. So how do I tell ofborg to only build on one architecture and why does the hash that works locally not work with ofborg?
| 20:35:08 |
| backtail | * Hey peeps, I unknowingly ran into the leaveDotGit bug of unpredictable hash generation for nix derivations. I patched the dependency, but now ofborg says that my provided hash is wrong:
error: hash mismatch in fixed-output derivation '/nix/store/q59g9bk2kwh20sr8ir00gz5d5yjkb75r-source.drv':
specified: sha256-HoqFBkxXlmbWaFPOVvO0zUhlDBVzz++T6Fgf0MAuyao=
got: sha256-rcIbtIXTm1sNMCbs6W5DS/9KeMkWWhUdwReW++nW2Qc=
error: 1 dependencies of derivation '/nix/store/q5izjyv5iv1h0zdc06my4dp07wvv9j76-KnobKraft-orm-2.7.1.drv' failed to build
The point that I don't understand is that every architecture, aarch64, x64_86, etc. all have different hashes. However, I just want to build it for x64_86_linux. So how do I tell ofborg to only build on one architecture and why does the hash that works locally not work with ofborg?
| 20:37:20 |
| backtail | Well... now that I got another review, suddenly ofborg gives me output that it did successfully built the package for linux (aarch64 and x64_86) but failed for darwin. I am honestly confused as to how to interpret ofborg output. I would appreciate a quick explanation! | 21:16:19 |
 Sergei Zimmerman (xokdvium) | More of a nixpkgs question generally: https://matrix.to/#/#users:nixos.org. But this might have to do with case sensitivity and whatnot | 23:44:15 |
| 28 Oct 2025 |
| Sergei Zimmerman (xokdvium) | Eelco: | 00:29:23 |
| Sergei Zimmerman (xokdvium) | * Eelco: as discussed during the meeeting, here's the constant-memory uploads to http binary caches: https://github.com/NixOS/nix/pull/14390 | 00:29:54 |
| backtail | In reply to @xokdvium:matrix.org
More of a nixpkgs question generally: https://matrix.to/#/#users:nixos.org. But this might have to do with case sensitivity and whatnot Thanks for the pointer! | 01:39:08 |
| backtail left the room. | 01:43:26 |
| fzakaria | any good issues I can work on? | 01:58:23 |
| fzakaria | I finished that NAR thing. | 01:58:26 |
 tomberek | @fzakaria:one.ems.host: perhaps a more long term solution to:https://github.com/GrahamDennis/nix/pull/8. We got team approval for "final" to be exposed so that locked fetches don't force a refetch if it already exists. | 02:10:55 |
| fzakaria | expose final to the lock file ? | 02:12:59 |
| fzakaria | or it sounds like assign final to the fetchers if the attrs have whatever flake considers to be final | 02:13:21 |
| fzakaria | (I'm not familiar with this feature so reading in between the lines of the PR)
| 02:13:38 |
|  connor (burnt/out) (UTC-8) changed their profile picture. | 02:19:16 |
| connor (burnt/out) (UTC-8) changed their profile picture. | 02:19:57 |
| tomberek | The idea is that builtins.fetchTree would accept a final = true; parameter and this would signal that the other parameters given (like revCount, lastModified, etc) would be accepted as is when a local store path matches the given narHash. Otherwise what happens is that that source would need to be re-fetched all the time (eg: if the fetcher cache is cleared out) even though the narHas/store-path exists locally, just to fetch those other attributes that are usually provided in a flake.lock/npins/niv/etc situation. In other words, "if final is true, don't bother refetching this source and trust the lockfile's attributes'. | 03:46:23 |
 lovesegfault | Okay, John Ericson Sergei Zimmerman (xokdvium) this is the next step in the grand why-depends-cycle-finding-unification. It's a lot of code, but it's mostly a shim around BGL. You can see the whole work with why-depends here: https://github.com/NixOS/nix/compare/master...lovesegfault:nix:bgl-depgraph | 03:51:56 |
| lovesegfault | I need to look more at why-depends, maybe there's more stuff in there I can reuse the graph for | 03:52:57 |
| lovesegfault | I kinda want to make it always precise, and then non-precise just tosses data out :P | 03:58:02 |
|  lunarizzz joined the room. | 07:49:26 |
| fzakaria | you want to pair on this? | 18:01:18 |
| fzakaria | having a test that shows it download first might be a good step first (capture the regression) | 18:01:32 |
| fzakaria | Redacted or Malformed Event | 18:17:15 |
| fzakaria | what's this new codderrabbit stuff.. | 18:18:49 |
