29 May 2025 |
Leonardo Santiago | I didn't remove nor add any new things, just nix flake update nix && nix build | 13:26:20 |
Leonardo Santiago | Robert Hensing (roberth): did anything change in the libs side? I tried looking this error up and all I could find was a gcc version mismatch on different libraries, not sure that's the case here | 13:27:17 |
Leonardo Santiago | * I didn't remove nor add any new packages, just nix flake update nix && nix build | 13:28:34 |
Leonardo Santiago | is this a problem with libstd++? maybe a version mismatch? | 13:36:39 |
Leonardo Santiago | nix/2.28.0 seems to be fine, nix/2.29.0 does reproduce the issue | 13:42:58 |
Leonardo Santiago | sadly I'm looking for libflake-c upgrades, 2.28.0 doesn't seem to contain, time to do some git bisect | 13:44:31 |
Leonardo Santiago | c212035d94ee4407cd19927ba33e3246a07a54d0 is the first bad commit
commit c212035d94ee4407cd19927ba33e3246a07a54d0
Author: Robert Hensing <robert@roberthensing.nl>
Date: Tue Apr 1 15:19:41 2025 +0200
flake: nixpkgs: 24.11 -> nixos-unstable
Flake lock file updates:
• Updated input 'nixpkgs':
'github:NixOS/nixpkgs/48d12d5e70ee91fe8481378e540433a7303dbf6a?narHash=sha256-1Noao/H%2BN8nFB4Beoy8fgwrcOQLVm9o4zKW1ODaqK9E%3D' (2024-12-16)
→ 'github:NixOS/nixpkgs/52faf482a3889b7619003c0daec593a1912fddc1?narHash=sha256-6hl6L/tRnwubHcA4pfUUtk542wn2Om%2BD4UnDhlDW9BE%3D' (2025-03-30)
flake.lock | 8 ++++----
flake.nix | 2 +-
2 files changed, 5 insertions(+), 5 deletions(-)
bisect found first bad commit
| 14:14:20 |
Leonardo Santiago | * c212035d94ee4407cd19927ba33e3246a07a54d0 is the first bad commit
commit c212035d94ee4407cd19927ba33e3246a07a54d0
Author: Robert Hensing <robert@roberthensing.nl>
Date: Tue Apr 1 15:19:41 2025 +0200
flake: nixpkgs: 24.11 -> nixos-unstable
Flake lock file updates:
• Updated input 'nixpkgs':
'github:NixOS/nixpkgs/48d12d5e70ee91fe8481378e540433a7303dbf6a?narHash=sha256-1Noao/H%2BN8nFB4Beoy8fgwrcOQLVm9o4zKW1ODaqK9E%3D' (2024-12-16)
→ 'github:NixOS/nixpkgs/52faf482a3889b7619003c0daec593a1912fddc1?narHash=sha256-6hl6L/tRnwubHcA4pfUUtk542wn2Om%2BD4UnDhlDW9BE%3D' (2025-03-30)
flake.lock | 8 ++++----
flake.nix | 2 +-
2 files changed, 5 insertions(+), 5 deletions(-)
bisect found first bad commit
| 14:14:36 |
Robert Hensing (roberth) | Hmm yeah. Was going to say no major changes on our side | 14:14:43 |
Leonardo Santiago | it seems that it might be related to the nixpkgs upgrade in nix | 14:15:12 |
Leonardo Santiago | though I'm not sure how, maybe it changed the major version of cc (is it using clang?) and the one I'm using is incompatible with it? | 14:15:44 |
Leonardo Santiago | I looked up your nixops4 setup and it seemed pretty similar (almost exactly the same) as mine so I don't know what I'm doing wrongly | 14:16:13 |
Robert Hensing (roberth) |
clang?
Not on linux
| 14:16:18 |
Leonardo Santiago | oh, so it's on gcc ? that may be it, you have some gcc measures that I'm not using | 14:16:45 |
Robert Hensing (roberth) | I do use clang on linux in my dev shell when hacking on the thing | 14:17:38 |
Leonardo Santiago | I'm not a super c++ guy but the only instance I found someone commenting this was because of a gcc13 -> 14 version mismatch and so I think that'd be the issue here too | 14:18:03 |
Robert Hensing (roberth) | May be possible to build it that way with pkgsLLVM | 14:18:07 |
Leonardo Santiago | especially given that it is a nixpkgs upgrade | 14:18:08 |
Leonardo Santiago | wdym? | 14:18:20 |
Robert Hensing (roberth) | Something like nix build nixpkgs#pkgsLLVM.nixVersions.git | 14:19:03 |
Robert Hensing (roberth) | It's not the most reliably buildable package set, but if it works... | 14:19:30 |
Leonardo Santiago | oh, it was building fine with gcc before, I'll try to add your gcc countermeasures and see if that fixes it | 14:19:45 |
Robert Hensing (roberth) | ok | 14:20:09 |
Leonardo Santiago | yep, apparently it worked | 14:30:58 |
Leonardo Santiago | nope, false positive, tested on old nix version | 14:35:18 |
Leonardo Santiago | just to confirm, indeed the bump made the jump from gcc13 to gcc14 and it seems to be inline with this gentoo forum error report: https://forums.gentoo.org/viewtopic-t-1172390-start-0.html | 15:23:17 |
Leonardo Santiago | * just to confirm, indeed the nixpkgs bump made the jump from gcc13 to gcc14 and it seems to be inline with this gentoo forum error report: https://forums.gentoo.org/viewtopic-t-1172390-start-0.html | 15:23:52 |
Leonardo Santiago | though I'm not entirely sure where this mismatch comes from. maybe the libclang I'm using in bindgen is equivalent to gcc13? I tried updating nixpkgs to latest but nothing worked | 15:25:09 |
Leonardo Santiago | https://github.com/llvm/llvm-project/issues/96610 | 15:54:41 |
Leonardo Santiago | this means that you're using bindgen with gcc, is that correct? | 15:55:12 |