| 29 Jan 2026 |
 Randy Eckenrode | The idea that updating the libc++ headers would break the Swift compiler hurts my head. | 17:54:05 |
| Randy Eckenrode | Or maybe it’s that plus the hardening flags interacting badly for our 5.10.1 build. | 17:54:35 |
 Ihar Hrachyshka | I think folks on github refuted the hardening theory | 17:54:58 |
| Randy Eckenrode | Though that wouldn’t explain Linux. | 17:55:23 |
| Ihar Hrachyshka | is linux crashing the same way | 17:55:49 |
| Ihar Hrachyshka | thought it's darwin only (with swift-frontend burning) | 17:56:22 |
 samasaur | Linux was already broken in a different way; unsure if this new issue affects it too | 18:12:56 |
| Ihar Hrachyshka | if it's darwin.libcxx it just can't | 18:16:23 |
 eveeifyeve | I am just building your update-branch. WIth nix build .#swift | 18:16:28 |
| samasaur | In reply to @ihar.hrachyshka:matrix.org
if it's darwin.libcxx it just can't yea. from a "solving this" standpoint I hope it is libcxx. from an "our shared sanity" standpoint I hope it isn't | 18:22:59 |
| Ihar Hrachyshka | what would be the approach forward if we confirm the bump for libcxx is the problem? is it realistic to revert it while this is being fixed? | 18:23:34 |
| samasaur | vendor the old version inside the swift directory? | 18:23:50 |
| Ihar Hrachyshka | ah right. the power of Nix :p | 18:24:20 |
| samasaur | actually that wouldn't work I think... | 18:26:18 |
| samasaur | bc swift doesn't reference darwin.libcxx directly | 18:26:27 |
| samasaur | it builds the LLVM one | 18:26:39 |
| samasaur | so if libcxx broke something it's due to transitive behavior I think | 18:26:50 |
| Ihar Hrachyshka | "just" build a second stdenv tree for swift? | 18:26:52 |
| samasaur | dear god I hope not | 18:27:37 |
| samasaur | maybe try and track down which specific package is causing the transitive error? | 18:27:57 |
| samasaur | a dependency could have actually different behavior due to availability checking, if I understand correctly | 18:28:38 |
| samasaur | and knowing what the actual difference is might help | 18:28:54 |
| samasaur | In reply to @samasaur:matrix.org
maybe try and track down which specific package is causing the transitive error? not really sure how we'd do that though, since libcxx is in the stdenv so it affects ~all Darwin programs | 18:30:23 |
| Ihar Hrachyshka | In reply to @eveeifyeve:matrix.org
Known issue: https://github.com/NixOS/nix/issues/11676 Fwiw this is not the same issue. I also hit this issue on another machine that is forced to run Microsoft Defender crap (also mentioned in that issue). But the machine where Swift was unpacking for 30m is different and the symptoms are not during eval. | 18:40:51 |
| Ihar Hrachyshka | * Fwiw this is not the same issue. I also hit this linked issue on another machine that is forced to run Microsoft Defender crap (also mentioned in that issue). But the machine where Swift was unpacking for 30m is different and the symptoms are not during eval. | 18:42:13 |
| Randy Eckenrode | If it really is, we can override the one in the stdenv to use the libc++ from the 15.5 SDK. | 19:01:45 |
 crushing-smite | Are there any solutions for declarative network management on mac with nix? | 21:55:59 |
| samasaur | In reply to @crushing-smite:matrix.org
Are there any solutions for declarative network management on mac with nix? nix-darwin can do some network management. Depends on what exactly you're trying to do | 22:05:58 |
| Ihar Hrachyshka | there's also https://github.com/nix-darwin/nix-darwin/pull/1657 | 22:06:38 |
| samasaur | ack more for me to review 😅 | 22:08:53 |
