| 28 Jul 2025 |
 John Ericson | it could just be foo-build-bar, baz-host-quux too, right? | 13:01:04 |
 Alyssa Ross | To start with at least I'd say we don't. | 13:01:05 |
| John Ericson | no need for actual hash | 13:01:08 |
 dramforever | hmm | 13:01:15 |
| Alyssa Ross | In reply to @Ericson2314:matrix.org
it could just be foo-build-bar, baz-host-quux too, right? What do you do when those shift in a later build though? Now they're named wrong. | 13:01:33 |
| John Ericson | ideally they would not persist to a later build | 13:01:59 |
| dramforever | it is what nixpkgs currently calls local and cross though | 13:01:59 |
| John Ericson | build should def not persist | 13:02:17 |
| John Ericson | host.....I am not sure | 13:02:22 |
| Alyssa Ross | In reply to @Ericson2314:matrix.org
ideally they would not persist to a later build Hmm. We can't know that… | 13:02:27 |
| John Ericson | target currently persists | 13:02:29 |
| John Ericson | but that is why everyone should work on GCC NG | 13:02:46 |
| John Ericson | :D | 13:02:56 |
| John Ericson | we can grep | 13:03:31 |
| dramforever | honestly | 13:03:40 |
| Alyssa Ross | TBH I think it would be fine | 13:03:49 |
| Alyssa Ross | We could try it and switch to hashes if it doesn't work | 13:03:53 |
| dramforever | {arch}-unknown-{os}-{abi} and {arch}-cross-{os}-{abi} sounds nice | 13:04:09 |
| Alyssa Ross | (I'm not volunteering to implement this — tbh I'm surprised nobody it's telling me it's too bonkers to even be considering) | 13:04:18 |
| John Ericson | if we do hashes then I don't think we force build != host? | 13:04:26 |
| dramforever | yes i'm bikeshedding | 13:04:27 |
| Alyssa Ross | * | 13:04:28 |
| John Ericson | because they would have the same hash too? | 13:04:34 |
| Alyssa Ross | That's true | 13:04:48 |
| Alyssa Ross | Plus we still don't configurePlatforms by default anyway | 13:05:02 |
 p14 | John Ericson: does gccng pose a solution to https://github.com/NixOS/nixpkgs/issues/265121 ?
A test case is building bashNonInteractive while varying ‘gcc.arch’. | 13:20:56 |
 Tristan Ross | In reply to @p14:matrix.org
John Ericson: does gccng pose a solution to https://github.com/NixOS/nixpkgs/issues/265121 ?
A test case is building bashNonInteractive while varying ‘gcc.arch’. I think I've seen that with gfortran which GCC NG's gfortran works with pkgsLLVM | 17:40:08 |
| Tristan Ross | Ericson and I have been working on this but I've got no bandwidth to spare for the next two weeks. | 17:42:10 |
| p14 | Can you help me parse what you said? You're saying that gcc ng's pkgsLLVM gfortran has this problem and is fixed vs gcc (not-ng)? | 17:44:27 |
| Tristan Ross | In reply to @p14:matrix.org
Can you help me parse what you said? You're saying that gcc ng's pkgsLLVM gfortran has this problem and is fixed vs gcc (not-ng)? No | 17:45:15 |
