| 2 Aug 2023 |
 p14 | raitobezarius: I've been doing some testing of the above on aarch64-darwin. I found that plain pkgsLLVM.stdenv doesn't build on master, I get a silent failure in compiler-rt. NIX_DEBUG exposes the issue that preConfigure is trying to set CMAKE_LIPO=$(command -v aarch64-apple-darwin-lipo) but this goes straight to the exit/failure handler.
Is pkgsLLVM known (meant?) to work on this platform? Is there a darwin platform it works on?
| 09:24:55 |
 sorear | part of why I haven't touched this recently is that I don't have a darwin anything... | 13:17:21 |
| p14 | sorear: Yeah, it's a challenge. I can see from the git blame lens (and also through the diff-the-nix-llvm-recipies-through-the-versions) as I hack on things that people have gone through many iterations of fixing things for different environments. | 13:20:44 |
| sorear | I see you're only fixing this for _16 so if I want pkgsLLVM _15 to work I | 13:21:38 |
| sorear | 'll need to do my own backport | 13:21:49 |
| sorear | _14 works fine with the existing nixpkgs code, i tested a while ago | 13:22:12 |
 raitobezarius | In reply to @p14:matrix.org
raitobezarius: I've been doing some testing of the above on aarch64-darwin. I found that plain pkgsLLVM.stdenv doesn't build on master, I get a silent failure in compiler-rt. NIX_DEBUG exposes the issue that preConfigure is trying to set CMAKE_LIPO=$(command -v aarch64-apple-darwin-lipo) but this goes straight to the exit/failure handler.
Is pkgsLLVM known (meant?) to work on this platform? Is there a darwin platform it works on?
That's a question for the Darwin folks | 13:28:48 |
| raitobezarius | I have no idea personally | 13:28:52 |
| raitobezarius | I know that rrbutani has been doing work to support better Darwin | 13:29:06 |
| raitobezarius | But it's hard to keep track who is working on what | 13:29:19 |
| p14 | Ah, yeah, apols, I was tagging you in part because I saw your name associated with some Darwin things. | 13:29:36 |
| raitobezarius | I don't do Darwin at all | 13:30:08 |
| raitobezarius | (except the basic stuff to avoid breaking it massively) | 13:31:26 |
| raitobezarius | Overall, I'd say it's fine to aim for a pkgsLLVM that works on Linux first | 13:31:59 |
| p14 | Cool. Is there anything I can do to improve the chances of landing #246577 in the reasonably short term? (In that I have more time to fix things right now than I will in the future) | 13:32:11 |
| p14 | Thinking aloud: I suppose that once it's working it might also be reasonable to bump llvmPackages to llvmPackages_16, so that pkgsLLVM uses the more recent compiler? | 13:33:03 |
| raitobezarius | We are getting into #compilers:nixos.org land but | 13:33:25 |
| raitobezarius | (1) We need to finish the unification work between LLVM14/LLVM15/LLVM16/LLVMgit | 13:33:40 |
| raitobezarius | * (1) We need to finish the unification work between LLVM14/LLVM15/LLVM16/LLVMgit (esp for LLVM17 which I believe is in RC now?) | 13:33:49 |
| p14 | Yep, 17rc is out | 13:33:57 |
| p14 | Interesting to hear that there is a unification effort, that's also good to hear. | 13:34:06 |
| raitobezarius | so any changes that will bring a new discrepancy will probably be delayed because it will introduce more work for that unification stuff | 13:34:09 |
| sorear | iiinteresting | 13:34:28 |
| p14 | Is this effort in a PR or other discussion somewhere? Is there a timeline for it? | 13:35:15 |
| p14 | I may also be interested in helping and testing and whatnot if I get spare bandwidth. | 13:35:42 |
| raitobezarius | (2) I think the timing is complicated, if #246577 is the only change you are hoping to land, I can try to take care of it before going into vacations, but, beyond than that, I think most LLVM maintainers are inactive atm (understandably) | 13:36:22 |
| p14 | I think that's a pretty big step forward in terms of the usability of it. | 13:37:17 |
| raitobezarius | In reply to @p14:matrix.org
I may also be interested in helping and testing and whatnot if I get spare bandwidth. https://github.com/NixOS/nixpkgs/issues/213033 | 13:37:19 |
| raitobezarius | (3) pkgsLLVM is probably low priority and too much to chew for me so at some point, you would have to consider stepping in as a maintainer if you want to see continued stability of that package set | 13:38:14 |
| sorear | would it help if I tested 246577 sooner rather than later? | 13:53:12 |
