!lheuhImcToQZYTQTuI:nixos.org

Nix on macOS

1190 Members
“There are still many issues with the Darwin platform but most of it is quite usable.” — http://yves.gnu-darwin.org199 Servers

Load older messages


SenderMessageTime
26 Feb 2026
@reckenrode:matrix.orgRandy EckenrodeWe don’t have a package set at that point, and we can’t get the version in the stage for obvious reasons.03:14:59
@reckenrode:matrix.orgRandy Eckenrode llvm is llvmPackages.llvm, which is llvmPackages_XX.llvm. Trying to overlay llvmPackages_${lib.versions.major llvm.version} will cause an infinite recursion. 03:17:07
@reckenrode:matrix.orgRandy Eckenrode It worked before because the stdenv did it backwards. It defined the overlay for the unversioned llvmPackages then added a versioned one at the end. 03:18:16
@reckenrode:matrix.orgRandy EckenrodeBut that’s not how the package sets are defined normally.03:18:58
@alexfmpe:matrix.orgalexfmpewelp, 300 rebuilds just for hello03:19:13
@alexfmpe:matrix.orgalexfmpealso doing ispc and ghc03:19:19
@alexfmpe:matrix.orgalexfmpegoing to leave it overnight, anything else interesting to try?03:19:25
@reckenrode:matrix.orgRandy EckenrodeSomething using Boost? Chromium or Qt Web Engine.03:19:56
@reckenrode:matrix.orgRandy EckenrodeMaybe Node.03:20:05
@alexfmpe:matrix.orgalexfmpehuh I see 2 llvm 22 in the dep chain, that expected?03:20:29
@reckenrode:matrix.orgRandy EckenrodeYes.03:20:41
@reckenrode:matrix.orgRandy EckenrodeFirst one is built with the bootstrap tools. Second one is rebuilt with Clang 22.03:21:10
@reckenrode:matrix.orgRandy EckenrodeAt least the LLVM tests are only tun once.03:22:04
@reckenrode:matrix.orgRandy Eckenrode * 03:22:13
@reckenrode:matrix.orgRandy EckenrodeSecond one also has dependencies built by Clang 22 and not the bootstrap tools.03:23:27
@alexfmpe:matrix.orgalexfmpehmm building llvm_22 with 21 as default yields https://github.com/NixOS/nixpkgs/pull/493742#issuecomment-396067342403:23:31
@alexfmpe:matrix.orgalexfmpeshould I try 22.1.0-rc3 with 21, or 22 with 2203:23:56
@alexfmpe:matrix.orgalexfmpeguess the later03:24:07
@alexfmpe:matrix.orgalexfmpehope it doesn't hit the same test failure03:24:22
@reckenrode:matrix.orgRandy Eckenrode IIRC some new tests aren’t getting DYLD_LIBRARY_PATH set up to find the required dylibs in the build folder instead of in $lib. 03:24:52
@reckenrode:matrix.orgRandy EckenrodeI ran into something similar with Swift’s LLVM fork, but I just disabled tests for now because I’d rather prioritize getting Swift updated.03:29:30
@reckenrode:matrix.orgRandy Eckenrode * 03:29:41
@lt1379:matrix.orgLuni think we just want to turn off the one failing test there (llvm 22 pr), commented that. it's a regression test for specifically empty env array so if we patch it to pass DYLD_LIBRARY_PATH it isn't even testing the thing it wants to test any more, it kinda just doesn't work for us.04:23:40
@qyliss:fairydust.spaceAlyssa RossWe already fixed this https://github.com/NixOS/nixpkgs/pull/48740708:22:08
@alexfmpe:matrix.orgalexfmpe fails on darwin.adv_cmds 13:24:14
@alexfmpe:matrix.orgalexfmpe
../localedef/parser.y:232:18: error: incompatible pointer types passing 'wchar_t[2]' (aka 'int[2]') to parameter of type 'char *' [-Wincompatible-pointer-types]
  232 |                         copy_category(w);
      |                                       ^
13:24:20
@alexfmpe:matrix.orgalexfmpegot through 200 rebuilds, last 100 depend on this one13:25:06
@telent:matrix.orgtelent left the room.13:40:53
@reckenrode:matrix.orgRandy EckenrodeAny details on what’s failing?13:44:53
@reckenrode:matrix.orgRandy Eckenrode .y suggests the Bison step is failing for some reason. 13:46:40

Show newer messages


Back to Room ListRoom Version: 6