| 14 Sep 2025 |
 emily | ah, 6.4 will be required for CMake 4 anyway it looks like? I am happy to merge it if all looks good then | 22:28:22 |
 Lun | I need to learn jj 😅 | 22:29:00 |
| Lun | Yeah it will be.
I think I want to drop one part of that PR I'm less confident about to handle in a followup and then I'd appreciate a merge.
Haven't really had enough people test on different systems to confirm a change to arch selection to use -generic when possible works right so probably safer to delay. | 22:32:09 |
| emily | left a review, so maybe you can throw in my nitpicks with the core-hours spent on checking msgpack-cxx :p | 22:33:47 |
| Lun | ty! | 22:34:38 |
| emily | (literally scrolled down the entire diff too quickly to read anything and only stopped when my instinctive pattern matching told me it noticed something) | 22:34:39 |
| Lun | understandable | 22:37:15 |
| Lun | nooo the messed up date you pointed out means rebuilding CK | 22:39:29 |
| emily | I mean I'm okay not rechecking the build after that. | 22:53:11 |
| emily | if you're doing anything cursed enough to break depending on that version field then you deserve it. | 22:53:28 |
| 15 Sep 2025 |
| emily | @lt1379:matrix.org FWIW after doing some digging I think that if we get pkgsStatic GCC producing static PIEs by default then all we need is GO_EXTLINK_ENABLED for static ELF platforms | 00:38:31 |
| emily | and the right thing will magically happen | 00:38:42 |
| emily | https://discourse.llvm.org/t/default-setting-for-clang-default-pie-on-linux/61688/2 "I cannot name a distribution that doesn’t use default PIE. I’d like to know if this even exists:)" 🫣 | 00:44:43 |
| emily | lol Fedora has to do the PIE flag injection stuff because of GCC maintainer politics | 00:47:45 |
| emily | poor guys | 00:47:47 |
| emily | Lun: looks like quite literally ROCm stuff is the only thing unhappy with the new CMake stuff 😅 | 01:22:22 |
| emily | though i remember we had to patch FoundationDB for it (before I realized we have two copies of the library even) | 01:22:33 |
| emily | okay no neovim-qt also is | 01:22:44 |
| emily | okay and some other stuff just fails to find it | 01:25:56 |
| Lun | can https://github.com/LunNova/nixpkgs/blob/62a423d771655ae43b581a35984060e3e4b52bb9/pkgs/development/rocm-modules/6/migraphx/msgpack-6-compat.patch be simpler? | 01:28:49 |
| Lun | * can https://github.com/LunNova/nixpkgs/blob/62a423d771655ae43b581a35984060e3e4b52bb9/pkgs/development/rocm-modules/6/migraphx/msgpack-6-compat.patch be made simpler? | 01:29:05 |
| emily | well | 01:39:44 |
| emily | are you planning to upstream it? | 01:39:46 |
| emily | because if you don't want to do free labour for AMD I think you can just find_package(msgpack REQUIRED) → find_package(msgpack-cxx REQUIRED), msgpackc-cxx → msgpack-cxx | 01:40:21 |
| emily | it also probably makes no sense to add support for a half-decade-old version that they don't even support currently? | 01:40:56 |
| emily | (have they adapted some libs upstream but not all or something?) | 01:41:52 |
| Lun | Planning to upstream it. It looks like before it was using the msgpack target and just happily working via FHS paths / -isystem wrapper magic despite not setting the target right | 01:42:17 |
| Lun | yeah | 01:43:24 |
| emily | I would skip the support for bare msgpack | 01:43:41 |
| emily | if they're already on msgpackc-cxx | 01:43:45 |
