| 9 Jun 2021 |
 LnL | btw, this is not unexpected?
/nix/store/0v8zly62czgwbql79950whxzc9s9lw2z-bootstrap-stage0-Libsystem/include/dlfcn.h:40:10: fatal error: 'stdbool.h' file not found | 18:59:23 |
 toonn | LnL: Every time I get that error it's because I'm not using the proper bootstrap-tools. | 19:08:32 |
| LnL | euh | 19:09:27 |
| LnL | don't tell me I uploaded the wrong thing :p | 19:09:51 |
| toonn | Oh, it also happened when using that bootstrap-tools with a non-updated stdenv. | 19:10:47 |
| toonn | No, nvm, that's not true. | 19:11:01 |
| LnL | that's what I was asking I just tried to build stage1 of that branch, nothing else included | 19:11:33 |
| toonn | It's the reason I added the bootstrapFiles argument to make-bootstrap-tools.nix. Because it was so hard to pass the right bootstrap-tools to that. | 19:11:42 |
| LnL | nix-shell is acting weird but clang --version in stage0 returns 11.1.0 | 19:14:17 |
| LnL | the store path is clang-wrapper-7.1.0 but that's just weird because the stdenv isn't updated yet and we assume the version there when unpacking | 19:15:20 |
| toonn | Yeah, and then stdenv ends up looking at tarball/lib/clang/7.1.0 | 19:18:03 |
| toonn | Which doesn't exist anymore. | 19:18:14 |
| toonn | Hmm, so I guess my nvm was wrong. | 19:18:45 |
| toonn | LnL: Btw, how did you get the build products uploaded? | 19:30:05 |
| toonn | Do you have privileged access to hydra? | 19:30:16 |
| LnL | I have a hydra account but only have limited permissions, tarballs.nixos.org is completely separate | 19:31:51 |
| LnL | I got access to the s3 bucket at some point because we need to update bootstrap tools semi regularly in comparison to eg. linux | 19:32:50 |
| toonn | Oh, right. I saw the hash from the range and assumed this was on hydra. | 19:32:55 |
| LnL | that's just for traceability | 19:35:23 |
| LnL | I could upload a random locally built tarball but this way anybody can validate whether the mirror is valid and if the bootstrap build is reproducible the hydra result can even be verified | 19:40:03 |
| toonn | LnL: Oh, should I still base the stdenv LLVM bump on that same commit? Just with your commit cherry-picked in between? | 19:56:48 |
| LnL | we can pick another one | 19:57:51 |
| LnL | what I usually do is pick an eval that looks good, eg. https://hydra.nixos.org/eval/1677022 | 19:58:53 |
| LnL | and then switch to the inputs tab, you can see the nixpkgs commit for that eval there | 19:59:07 |
| LnL | and then later you can use that eval number to compare https://hydra.nixos.org/eval/:id?compare=1677022 | 20:00:01 |
| LnL | this matters a lot more if you base against staging tho master changes but not that much as long as there's no staging merge inbetween | 20:01:39 |
| toonn | LnL: What's your reasoning for picking 1677022? Most passing jobs and recent but not too recent so things have mostly built? | 20:12:30 |
| LnL | it's the first one with the new bootstrap tools which I had still open in a tab :D | 20:13:50 |
| 10 Jun 2021 |
 Alyssa Ross | reading https://repository.tudelft.nl/islandora/object/uuid:2f66fe0c-4080-4148-a01c-acd530160797/datastream/OBJ/download really makes me want to try building a pure darwin system with Nix | 08:53:46 |
 Domen Kožar | is that even possible? I always thought that you can't avoid linking to libSystem | 08:55:51 |
