| 20 Sep 2025 |
 emily | I mean, you have AwesomeBrower3000 | 18:56:41 |
| emily | it has a copy of Mesa in it and also a copy of some Rust program that uses build.rs | 18:56:41 |
| emily | and it builds both. (put aside that this is horrible.) | 18:56:53 |
| emily | then you want the CLI and the hook. | 18:56:54 |
 dramforever | should "duplicate work" even matter though | 18:56:58 |
| emily | yes because #include_next + duplicate -isystem | 18:57:08 |
| dramforever | like does duplicate include paths actually affect #include_next | 18:57:11 |
| dramforever | oh | 18:58:08 |
| dramforever | it does... | 18:58:12 |
| emily | it's not that weird to me that two copies of the entire flags breaks stuff | 18:58:15 |
| emily | for like multiple potential reasons | 18:58:24 |
| emily | it looks like the only additional smarts the wrapper has is sniffing for -x c++ | 18:58:52 |
| emily | does GCC support -cxx-isystem? | 18:59:19 |
| emily | we could kill off that stuff in our wrappers if just passed the right flags, for Clang at least… | 18:59:29 |
| emily | looks like no. tragic | 18:59:46 |
| emily | OTOH it's not clear why we want to hide C++ headers from plain C compiles. | 19:00:08 |
| emily | given that they're visible on normal systems | 19:00:12 |
| emily | anyway… just dropping the useless dep from Mesa sounds good | 19:00:39 |
 ghpzin | It has duplicates in BINDGEN_EXTRA_CLANG_ARGS on both gcc14 and gcc15, why it fails only on gcc15 not sure (seemingly only hashes/versions change).
https://gist.github.com/ghpzin/e9bd2497c7cebba3d13e409f1f916c6b | 19:00:45 |
| emily | but I don't love the hook and wrapper conflicting in general | 19:00:46 |
| emily | and I hate wrappers :) | 19:00:49 |
 K900 | It did at one point | 19:01:46 |
| K900 | Feel free to remove | 19:01:49 |
| emily | (would anyone object to fixing this and finally making clang -x c++ etc. work) | 19:04:49 |
| emily | (I guess the problem is that we do the linking manually…) | 19:05:10 |
 Lun | anyone looking at lvm2 being broken due to a disallowed ref to bashInteractive? | 19:26:57 |
| Lun | * anyone looking at lvm2 being broken due to a disallowed ref to bash? | 19:27:26 |
| emily | these disallowed requisites seem to have been a mistake | 19:49:59 |
| emily | we should probably just keep it in a. bashless NixOS test or at least make it direct references rather than transitive requisites so that changes don't break stuff further down the tree at random | 19:50:43 |
 Grimmauld (any/all) | Is the preferred cmake 4 fix a patch or adding "-DCMAKE_POLICY_VERSION_MINIMUM=3.10" to cmakeFlags ? | 21:01:36 |
