| 30 Nov 2025 |
 aleksi | * | 11:38:03 |
 helle (just a stray cat girl) | yep, as the quality of tinycc uh, leaves some to be desired | 11:41:16 |
 Alyssa Ross | I think there certainly would be | 12:11:51 |
 dish [Fox/It/She] | i would definitely be interested in that | 15:19:24 |
| dish [Fox/It/She] | also looking for any additional reviews/feedback on https://github.com/NixOS/nixpkgs/pull/448701 before I merge it | 15:26:04 |
|  Ingo Blechschmidt joined the room. | 21:28:52 |
| 3 Dec 2025 |
| aleksi | dish [Fox/It/She]Nice, the bootstrap chain for x86_64 and riscv64 is basically working on my branch of Aux foundation: https://git.auxolotl.org/aleksi/foundation/src/branch/riscv64-bootstrap
Compared to that PR (LGTM but I'm not a nixpkgs expert), the difference is mostly bugfixes to mes & tinycc-bootstrappable, upgraded tinycc-mob and getting gcc 4.6 from ekaitz-zarraga for a RV64 backend | 19:14:31 |
| aleksi | There are some assorted little patches besides that | 19:14:38 |
| dish [Fox/It/She] | those are some nice patches, will look at that soon | 19:18:49 |
| dish [Fox/It/She] | definitely appreciate the work that aux foundation has done on bootstrapping properly | 19:19:03 |
| aleksi | Thanks, no problem | 19:23:25 |
| aleksi | I see live-bootstrap people have also bootstrapped autotools to prevent relying on autogenerated configure scripts. I don't know if nixpkgs wants or needs that though | 19:24:13 |
 Grimmauld (any/all) | i wish we had proper hooks to remove all autogenerated files ahead of builds. Like, test binaries (xz backdoor), but also autotools artifacts. We currently don't really do that... | 19:26:58 |
| helle (just a stray cat girl) | I mean I complain already about downloading binaries when we can build from source for things | 19:27:44 |
| helle (just a stray cat girl) | "but there is no build system in the source" and it turns out Debian has already made a perfectly adequate build system for it | 19:28:20 |
 Qyriad | heck autoreconfHook isn't in default stdenv but running ./configure is | 19:30:01 |
| Grimmauld (any/all) | yes | 19:30:19 |
| Grimmauld (any/all) | its a mess | 19:30:22 |
| Grimmauld (any/all) | I would like find . -type f -executable -exec rm {} \; as an early build step... | 19:32:18 |
| Grimmauld (any/all) | (more care would be needed, because some projects have the permission bits on files they later install. But still, nothing of the source should start out executable.) | 19:33:27 |
| helle (just a stray cat girl) | one of my reasons for annoyance is, we know that what is in nixpkgs is consistent across builders, but the optimisation of knowing what is in it in the first place would be nicer still | 19:34:25 |
| Grimmauld (any/all) | we do have some dumb impurities in nix though | 19:35:23 |
| Grimmauld (any/all) | like, it makes a difference whether builder is running hardened kernel or not for things like asan and valgrind | 19:35:44 |
| Grimmauld (any/all) | * like, it makes a difference whether builder is running hardened kernel or not for things like asan and valgrind being invoked in tests | 19:35:51 |
 Tristan Ross | Or page sizes of your host can affect what binaries can run | 19:45:46 |
| Tristan Ross | I think if nix had a built in microVM, it would help with a lot of the impurities | 19:46:22 |
| helle (just a stray cat girl) | (can someone figure out a Haskell bootstrap or am I going to have to do it myself) | 20:01:36 |
 Puna | you mean from stdenv to modern GHC without fetching prebuilt GHCs? | 20:03:01 |
| helle (just a stray cat girl) | mhmhm | 20:03:08 |
| helle (just a stray cat girl) | btw, we have investigated the underlying issue and we know EXACTLY the potential starting points, but ooof, it is one chore | 20:03:32 |
