30 Oct 2022 |
Madoura | has nothing to do with the gpu | 08:07:28 |
pie_ | though technically you can check the toolchain mixing just by looking at the drvs | 08:07:29 |
Madoura | I put it there for rapid testing rather than all, which compiles for all gpus | 08:07:41 |
pie_ | aha | 08:07:50 |
pie_ | so yeah i guess its faster to check if there are still references to gcc (other than libc I guess?) in the closure of the deriation | 08:08:13 |
Madoura | Yup, that's the plan. | 08:08:24 |
pie_ | * so yeah i guess its faster to check if there are still references to gcc (other than libc I guess?) in the closure of the derivation, as opposed to yoloing the builds a few more times | 08:08:27 |
Madoura | https://github.com/NixOS/nixpkgs/blob/3928cfa27d9925f9fbd1d211cf2549f723546a81/pkgs/development/compilers/llvm/rocm/default.nix#L21
https://github.com/NixOS/nixpkgs/blob/3928cfa27d9925f9fbd1d211cf2549f723546a81/pkgs/development/compilers/llvm/rocm/default.nix#L40
| 12:17:52 |
Madoura | I suspect the GCC include is coming here, from llvm. | 12:18:04 |
Madoura | * I suspect the GCC include is coming here, from rocm-llvm. | 12:18:14 |
Madoura | * I suspect the GCC include is here, from rocm-llvm. | 12:19:09 |
pie_ | 0.o | 13:00:39 |
1 Nov 2022 |
| @emilytrau:matrix.org joined the room. | 21:43:17 |
3 Nov 2022 |
Madoura | Looks like it's faulty includes. I have rocBLAS correctly compiling though with a patch! | 02:31:50 |
Madoura | https://github.com/ROCmSoftwarePlatform/rocBLAS/issues/1277 | 02:32:02 |
5 Nov 2022 |
| markuskowa joined the room. | 17:14:15 |
7 Nov 2022 |
Madoura | So I don't have to copy/paste the entire chat log, please check: https://matrix.to/#/!kjdutkOsheZdjqYmqp:nixos.org/$HKubKzcJR12uwnZB7A1ymr8jeY-iE31wstzwfMOLC3Q?via=nixos.org&via=matrix.org&via=nixos.dev Starting from:
Is this the right place to have a discussion about something that could possibly be put on the official hydra, or a hydra-adjacent server? It concerns MIOpen.
| 15:03:30 |
Madoura | * So I don't have to copy/paste the entire chat log, please check: https://matrix.to/#/!kjdutkOsheZdjqYmqp:nixos.org/$HKubKzcJR12uwnZB7A1ymr8jeY-iE31wstzwfMOLC3Q?via=nixos.org&via=matrix.org&via=nixos.dev Starting in #dev:nixos.org from:
Is this the right place to have a discussion about something that could possibly be put on the official hydra, or a hydra-adjacent server? It concerns MIOpen.
| 15:04:08 |
10 Nov 2022 |
| @rdg:matrix.org joined the room. | 11:27:05 |
14 Nov 2022 |
pie_ | I'm about to try using proot to get nix on a 10 year old kernel | 12:25:07 |
pie_ | waiting for the static build to hopefully work right now... | 12:25:20 |
pie_ | The idea is to use nix to get steam-run to hopefully be able to run a headless unity application on an hpc node. | 12:30:29 |
pie_ | In reply to @jcie74:matrix.org I'm about to try using proot to get nix on a 10 year old kernel :/ proot build breaks... | 13:10:59 |
pie_ | I'm not sure why. It's breaking somewhere in util-linux I think but I have to check. | 13:11:19 |
pie_ | Oh, no, actually it's python :/ | 13:12:45 |
pie_ | Ok, so it is util-linux, but I'm pretty sure it's also rebuilding python every time...
error: builder for '/nix/store/4f54gwg5jhknxdpgfcayp4k76nyv3w64-util-linux-static-x86_64-unknown-linux-musl-2.37.4.drv' failed with exit code 2;
last 10 log lines:
> /nix/store/0r52lb67wqwbh6awlagnpbhciaakcqxn-x86_64-unknown-linux-musl-binutils-2.35.2/bin/x86_64-unknown-linux-musl-ld: /nix/store/bxdkvd3djqf9z3wigjcsbx6yshnc9vfn-linux-pam-static-x86_64-unknown-linux-musl-1.5.1/lib/libpam.a(pam_audit.o): in function `_pam_auditlog':
> (.text+0x337): undefined reference to `audit_close'
> /nix/store/0r52lb67wqwbh6awlagnpbhciaakcqxn-x86_64-unknown-linux-musl-binutils-2.35.2/bin/x86_64-unknown-linux-musl-ld: /nix/store/bxdkvd3djqf9z3wigjcsbx6yshnc9vfn-linux-pam-static-x86_64-unknown-linux-musl-1.5.1/lib/libpam.a(pam_audit.o): in function `pam_modutil_audit_write':
> (.text+0x541): undefined reference to `audit_close'
> collect2: error: ld returned 1 exit status
> make[2]: *** [Makefile:7396: chsh] Error 1
> make[2]: Leaving directory '/tmp/nix-build-util-linux-static-x86_64-unknown-linux-musl-2.37.4.drv-0/util-linux-2.37.4'
> make[1]: *** [Makefile:14611: all-recursive] Error 1
> make[1]: Leaving directory '/tmp/nix-build-util-linux-static-x86_64-unknown-linux-musl-2.37.4.drv-0/util-linux-2.37.4'
> make: *** [Makefile:6321: all] Error 2
For full logs, run 'nix log /nix/store/4f54gwg5jhknxdpgfcayp4k76nyv3w64-util-linux-static-x86_64-unknown-linux-musl-2.37.4.drv'.
error: 1 dependencies of derivation '/nix/store/nam1fgzycp7viggn8xphdac43f9ana0j-e2fsprogs-static-x86_64-unknown-linux-musl-1.46.5.drv' failed to build
error: 1 dependencies of derivation '/nix/store/y5f3dr38vmzn3d24ialzrn09w88yrqhj-libarchive-static-x86_64-unknown-linux-musl-3.5.3.drv' failed to build
building '/nix/store/c6xs1gzmnikhnfk9yxviq5gswlg9hh8j-talloc-static-x86_64-unknown-linux-musl-2.3.3.drv'...
error: 1 dependencies of derivation '/nix/store/gzc1xw5h2q5a686wd4jafrpmlk8hs1yl-proot-static-x86_64-unknown-linux-musl-5.2.0.drv' failed to build
| 13:32:40 |
pie_ | Oh hm, proot has an enablePython, and it seems like util-linux isn't even a direct dependency of proot. I'm not sure what's pulling it in. The error above is pam related, and util-linux also has pamSupport. | 13:47:36 |
pie_ | It would be nice if why-depends had a mode that didn't require a successful build... | 13:47:55 |
pie_ | Though I imagine it's probably related to one of these.
these 5 derivations will be built:
/nix/store/2spfjgh2vpn9r6yk4yw082943rci8jjx-zstd-static-x86_64-unknown-linux-musl-1.5.2.drv
/nix/store/4f54gwg5jhknxdpgfcayp4k76nyv3w64-util-linux-static-x86_64-unknown-linux-musl-2.37.4.drv
/nix/store/nam1fgzycp7viggn8xphdac43f9ana0j-e2fsprogs-static-x86_64-unknown-linux-musl-1.46.5.drv
/nix/store/y5f3dr38vmzn3d24ialzrn09w88yrqhj-libarchive-static-x86_64-unknown-linux-musl-3.5.3.drv
/nix/store/32l2s1szabqzgij3g9yx0za2zmlriyna-proot-static-x86_64-unknown-linux-musl-5.2.0.drv
| 13:49:29 |
pie_ | Though those could also just be fails somewhere in the deptree :/ | 13:50:05 |