| 28 Oct 2025 |
 Alex Nicolaou | well I see technically it says with the dterminate flag. but that is the default now, so perhaps you're saying use some other flag to make the determinate installer install nix instead. | 03:58:17 |
 emily | should be fine | 04:16:06 |
| emily | yes, Determinate Nix is an opt-out now | 04:16:41 |
| emily | per "To install the recommended vanilla upstream Nix, you will need to explicitly say no when prompted to install Determinate Nix." in our readm | 04:16:44 |
| emily | * per "To install the recommended vanilla upstream Nix, you will need to explicitly say no when prompted to install Determinate Nix." in our readme | 04:16:44 |
| emily | we should maybe remove the wording about the --determinate flag, but I guess it won't be relevant for long | 04:16:58 |
 samasaur | we could/should change it to --prefer-upstream-nix i think | 04:17:33 |
| samasaur | i guess we should probably check that that flag actually exists | 04:18:43 |
| samasaur | i know they said it would in the initial announcement | 04:18:49 |
| samasaur | but iirc that flag will make it install upstream nix for the two months | 04:19:15 |
| samasaur | and then honestly i think our recommendation should probably be to install lix and use that to bootstrap? | 04:19:50 |
| emily | probably we should just recommend the experimental installer… if it actually ships… | 04:20:22 |
| samasaur | because i'm figuring that most people just use the default of "letting nix-darwin manage your nix install with the default options" (as i do) and if they aren't doing that they should know what they're doing anyway | 04:20:43 |
| emily | afaik the main blocker has just been some infra stuff for months | 04:20:45 |
| emily | but it sounded like that was getting resolved | 04:20:49 |
| samasaur | yes sorry. that should be "experimental installer if it ships; otherwise lix installer" | 04:21:04 |
| samasaur | prioritizing the "install something that immediately gets overritten by nix-darwin managing nix" workflow | 04:21:27 |
| samasaur | trying to convince myself to do some nix-darwin work next which will include looking at this situation | 04:23:56 |
| samasaur | but i keep getting distracted by other things | 04:24:02 |
| Alex Nicolaou | Thanks to both of you for all yoru help. I am going to go the lix installer way for now. | 05:08:10 |
| Alex Nicolaou | If it was me, I would remove references to the Determinate installer as the flow is confusing, but OTOH anyone going down this path has a steep hill to climb anyway, so perhaps it doesn't matter much. | 05:08:47 |
| emily | Ihar Hrachyshka: were you saying something about builds being really slow lately? | 16:52:15 |
 cujanovic.filip | Why is the aerospace service not reading the default config file but instead is creating a new one inside the nix store? | 17:01:38 |
 Ihar Hrachyshka | @emilazy:matrix.org: yes, coreutils cp / tar were taking a long time to copy into store; but I later found had some general issues with the builder (it was hanging / crashing even without builds) so I decided to upgrade to 26 and hope it magically fixes it. So far no issues after the upgrade, but who knows... | 17:50:46 |
| Ihar Hrachyshka | * @emilazy:matrix.org: yes, coreutils cp / tar were taking a long time to copy into store; but I later found I had some general issues with the builder (it was hanging / crashing even without builds) so I decided to upgrade to 26 and hope it magically fixes it. So far no issues after the upgrade, but who knows... | 17:51:14 |
| 29 Oct 2025 |
 niklaskorz | nixpkgs documentation suggests libiconv is included in the Linux stdenv and only needs to be added explicitly for darwin builds, but cross compiling for Linux on a Darwin build machine has iconv missing. Is that expected behavior? | 17:33:12 |
| niklaskorz | nix-build -A pkgsCross.gnu64.protonup-rs on an aarch64-darwin host gives
> = note: ld: library not found for -liconv
> clang: error: linker command failed with exit code 1 (use -v to see invocation)
| 17:34:04 |
| niklaskorz | Hm, it's also missing if I explicitly add iconv to buildInputs, so I guess it's something else | 17:35:13 |
| niklaskorz | * Hm, it's also missing if I explicitly add libiconv to buildInputs, so I guess it's something else | 17:35:23 |
| niklaskorz | * nixpkgs documentation suggests libiconv is included in the Linux stdenv and only needs to be added explicitly for darwin builds, but cross compiling to Linux on a Darwin build machine has iconv missing. Is that expected behavior? | 17:35:33 |
