hip: fix cmake installation directories

[Madouura]

Description of changes

Tracking: #197885
Turns out any ${hip}/hip paths were turning into ${hip}/${hip} paths in hip-config.cmake.
This fixes that, and removed the need for the workaround on my other PRs.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 22.11 Release Notes (or backporting 22.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
  • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
• Fits CONTRIBUTING.md.

[LunNova]

Result of nixpkgs-review pr 197838 run on x86_64-linux 1

2 packages built:

• blender-hip
• hip

[LunNova]

Builds but not sure how to test blender properly so can't really properly review it.

[Madouura]

The only real change is making sure the symlinks in ${hip}/hip correctly point to their corresponding directories in the root folder.
In other words, it's probably fine. It does work I know for sure with my other PRs using hip.

[jtojnar]

This should be ideally fixed upstream, see https://github.com/jtojnar/cmake-snips#assuming-cmake_install_dir-is-relative-path

[Madouura]

Upstream hip or upstream cmake?
This doesn't seem to happen with all roc* projects (rocprim, hipcub), but I have observed 3 others create ${nix_path}/${package_name}/${nix_path}.
For hip, this seems to result in bad symlinks instead.
The problem seems to be nix-specific in that we use /nix/store to store package roots, and amd's cmake projects in particular end up taking the full path instead of a relative path.

[jtojnar]

Hip, they are making incorrect assumptions:

• https://github.com/ROCm-Developer-Tools/hipamd/blob/928684dfed0fb4a8a14c6d5d79014324f3218d82/CMakeLists.txt#L44
• https://github.com/ROCm-Developer-Tools/hipamd/blob/928684dfed0fb4a8a14c6d5d79014324f3218d82/CMakeLists.txt#L280-L282
• https://github.com/ROCm-Developer-Tools/hipamd/blob/928684dfed0fb4a8a14c6d5d79014324f3218d82/CMakeLists.txt#L363-L364
• https://github.com/ROCm-Developer-Tools/hipamd/blob/928684dfed0fb4a8a14c6d5d79014324f3218d82/CMakeLists.txt#L475

[Madouura]

Here's an issue for now. ROCm/hipamd#55

[Flakebi]

LGTM, hip works for me with the patch and the dangling symlinks arer gone.

[LunNova]

Looks good, should add a comment mentioning the upstream issue so we remember to remove those once it's fixed.

[Madouura]

Hold off on merging while I investigate just disabling BUILD_FILE_REORG_BACKWARD_COMPATIBILITY in this and other packages.

[Madouura]

This is okay to merge again, as unfortunately BUILD_FILE_REORG_BACKWARD_COMPATIBILITY and some other things didn't quite work out.
I'll have the rest of the packages in #197885 reference this issue as to why we're defining manual install directories.

[Flakebi]

Still LGTM