| 8 Apr 2026 |
 Randy Eckenrode | I didn’t change anything. Or do you mean newly broken? | 11:36:50 |
| Randy Eckenrode | I can look into it a little this morning, but I won’t really have time until this evening. | 11:37:10 |
| Randy Eckenrode | It’s expecting _libiconv_version if you’re using libiconv, but that’s specific to GNU libiconv. The check in https://github.com/libarchive/libarchive/blame/2bb6f3461c413c0461f0b377431062358e5902b2/libarchive/archive_version_details.c#L476 should be disabled on Darwin. | 11:42:58 |
| Randy Eckenrode | Though https://github.com/apple-oss-distributions/libiconv/blob/593134fb16998e214952a13ea0826a50eff4dfbf/citrus/iconv.c#L36 suggests it should be present in Apple’s build, so 🤔 | 11:44:53 |
 whispers [& it/fae] | the odd thing is that the same code and version passed on darwin for 25.11 | 11:45:22 |
| whispers [& it/fae] | * the odd thing is that the same code and libarchive version passed on darwin for 25.11 | 11:45:27 |
| whispers [& it/fae] | hence why i was worrying in that comment that there's something Odd™ going on there with the build system | 11:45:43 |
| whispers [& it/fae] | if it is just that simple failure though, i'm happy that that's the case :) i don't know anything darwin | 11:46:22 |
| whispers [& it/fae] | * if it is just that simple issue though, i'm happy that that's the case :) i don't know anything darwin | 11:46:33 |
| whispers [& it/fae] | * hence why i was worrying in that github comment that there's something Odd™ going on there with the build system | 11:46:45 |
| whispers [& it/fae] | * if it is just that simple issue though, i'm happy that that's the case :) i don't know anything darwin and i could be making it more of a deal than it is | 11:47:51 |
| Randy Eckenrode | The symbol should be present, so it’s seemingly an odd case. | 11:48:18 |
 sampointon | openblas is failing tests on aarch64-linux, not just musl (though there can always be fun extra musl-only errors). https://github.com/OpenMathLib/OpenBLAS/issues/5708 is the upstream issue and https://github.com/OpenMathLib/OpenBLAS/pull/5710 a patch that makes the tests pass, but it's not clear what's going on or if there are more issues lurking | 13:34:52 |
| whispers [& it/fae] | openblas (/nix/store/4g2wialyszpsy3py3bmqgyj0sd0cszqp-openblas-0.3.32, same as hydra) built locally for me on aarch64-linux just fine fwiw. so it seems like a flaky thing mby | 13:38:50 |
| whispers [& it/fae] | * "regular" non-musl openblas (/nix/store/4g2wialyszpsy3py3bmqgyj0sd0cszqp-openblas-0.3.32, same as hydra) built locally for me on aarch64-linux just fine fwiw. so it seems like a flaky thing mby | 13:39:12 |
| whispers [& it/fae] | * "regular" non-musl openblas (/nix/store/4g2wialyszpsy3py3bmqgyj0sd0cszqp-openblas-0.3.32, same as hydra) built locally for me on aarch64-linux just fine fwiw. so it seems like a flaky/environment-dependent thing mby | 13:40:14 |
 emily | uh dropping a volatile fixing something is cursed beyond all comprehension | 13:40:50 |
| whispers [& it/fae] | * "regular" non-musl openblas built locally for me on aarch64-linux just fine fwiw. so it seems like a flaky/ environment-dependent thing mby (/nix/store/4g2wialyszpsy3py3bmqgyj0sd0cszqp-openblas-0.3.32, same as hydra) | 13:41:06 |
| whispers [& it/fae] | * "regular" non-musl openblas built locally for me on aarch64-linux just fine fwiw. so it seems like a flaky/ environment-dependent thing mby (/nix/store/4g2wialyszpsy3py3bmqgyj0sd0cszqp-openblas-0.3.32, same as what's failing on hydra) | 13:41:50 |
 K900 | Gotta be a compiler bug | 13:43:20 |
| emily | well, could also be code with UB that volatile or not happens to kick into a different state space | 14:44:33 |
| emily | using volatile in the first place is highly suggestive of UB going on | 14:44:44 |
| emily | usually those perturbations are things getting fixed by volatile though... | 14:45:05 |
 Sapii | If a package is marked broken, but it depends on a still open pr to pass, do I wait? this is in the case of https://github.com/NixOS/nixpkgs/pull/471310 stopping https://github.com/NixOS/nixpkgs/pull/507464 | 16:48:36 |
| whispers [& it/fae] | fwiw I'm wondering if this might be an autoconf moment. I'm seeing some discussion about iconv on the mailing list, but i can't find the origin of that discussion to see if it's the same issue or something else https://lists.gnu.org/archive/html/autoconf/2026-04/msg00000.html | 17:16:41 |
| whispers [& it/fae] | * | 17:17:14 |
| whispers [& it/fae] | * | 17:18:47 |
| 9 Apr 2026 |
| Randy Eckenrode | I just hit the libarchive problem while working on a Darwin source release cleanup. Our libiconv has that symbol. Is libarchive not linking libiconv? 🤔 | 00:01:24 |
| Randy Eckenrode | It’s the autoconf libiconv issue. | 00:05:18 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/pull/508131 fix for libarchive | 00:29:51 |
