rocksdb: fix cross-compiling on darwin

[yihuang]

install_name_tool not found when crossing compiling on darwin

Closes: #216431

Description of changes

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/)
• 23.05 Release Notes (or backporting 22.11 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
• Fits CONTRIBUTING.md.

[winterqt]

Nit with regards to the commit message (if you think it's fine, feel free to ignore this): maybe mention "fix cross-compilation on darwin" in the first line? It doesn't really matter since you mention it in the description though :)

[yihuang]

Nit with regards to the commit message (if you think it's fine, feel free to ignore this): maybe mention "fix cross-compilation on darwin" in the first line? It doesn't really matter since you mention it in the description though :)

done, fixed.

[yihuang]

@winterqt may I ask why some derivations don't need to include this prefix but still build succesfully?

[winterqt]

@yihuang Can you remove the Update ... from the commit message (and optionally the Co-authored-by trailer, I don't really think I deserve it for just fixing a small thing)?

may I ask why some derivations don't need to include this prefix but still build succesfully?

Do you have an example in mind? I assume you mean when cross-compiling -- that does sound weird.

[yihuang]

Do you have an example in mind? I assume you mean when cross-compiling -- that does sound weird.

zstd cross-build fine without this prefix.

[yihuang]

@yihuang Can you remove the Update ... from the commit message (and optionally the Co-authored-by trailer, I don't really think I deserve it for just fixing a small thing)?

done

[winterqt]

zstd cross-build fine without this prefix.

Huh, that is weird, as even the hook it uses specifies the prefix.

I assume I can reproduce this by just trying to cross build from x86_64-darwin to aarch64-darwin?

[yihuang]

zstd cross-build fine without this prefix.

Huh, that is weird, as even the hook it uses specifies the prefix.

I assume I can reproduce this by just trying to cross build from x86_64-darwin to aarch64-darwin?

yes, that's what I'm doing.

[stephank]

Do you have an example in mind? I assume you mean when cross-compiling -- that does sound weird.

zstd cross-build fine without this prefix.

Are you referring to this install_name_tool invocation?

nixpkgs/pkgs/tools/compression/zstd/default.nix

Line 93 in 1ad0905

install_name_tool -change @rpath/libzstd.1.dylib $out/lib/libzstd.1.dylib $bin/bin/pzstd

That is dependant on the buildContrib flag, which is disabled when cross-compiling:

nixpkgs/pkgs/tools/compression/zstd/default.nix

Line 8 in 1ad0905

, buildContrib ? stdenv.hostPlatform == stdenv.buildPlatform

[yihuang]

Do you have an example in mind? I assume you mean when cross-compiling -- that does sound weird.

zstd cross-build fine without this prefix.

Are you referring to this install_name_tool invocation?

nixpkgs/pkgs/tools/compression/zstd/default.nix

Line 93 in 1ad0905

install_name_tool -change @rpath/libzstd.1.dylib $out/lib/libzstd.1.dylib $bin/bin/pzstd

That is dependant on the buildContrib flag, which is disabled when cross-compiling:

nixpkgs/pkgs/tools/compression/zstd/default.nix

Line 8 in 1ad0905

, buildContrib ? stdenv.hostPlatform == stdenv.buildPlatform

I see, that makes sense, thanks

[winterqt]

Was the second commit meant to go in a different PR? I thought rocksdb worked with just the install_name_tool reference fix?

[yihuang]

Was the second commit meant to go in a different PR? I thought rocksdb worked with just the install_name_tool reference fix?

I just found out, the cross-compiled library is like this:

$ otool -L ./result/lib/librocksdb.7.dylib
./result/lib/librocksdb.7.dylib:
	@rpath/librocksdb.7.dylib (compatibility version 7.0.0, current version 7.9.2)
	/nix/store/13aj4w3gzw1xf0clx81nkd5jvl63fmnq-jemalloc-aarch64-apple-darwin-5.3.0/lib/libjemalloc.2.dylib (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libsnappy.1.dylib (compatibility version 1.0.0, current version 1.1.9)
...

there are two link path not replaced with absolute path, it doesn't happen when doing native build. I don't know where is the best place to fix this.

And the @rpath/librocksdb.7.dylib I can't replace it even with the install_name_tool command.