gcc: fix building cross-compiler on 32-bit host platforms

[lopsided98]

Motivation for this change

Fixes building cross-compiler on 32-bit host platforms, which was broken by #81844.

Some quirk/bug of moveToOutput seems to sometimes prevent it from creating the output if it does not already exist. Therefore we need to manually create the output. On 64-bit host platforms, the output gets created here: https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/compilers/gcc/builder.sh#L208. Note that the is64bit variable refers to the host platform, although it would probably make more sense for it to refer to the target platform. This is why this bug does not occur for 32-bit ARM cross-compilers.

I'm not sure why #81844 triggered the problem with moveToOutput, as that function's code is hard for me to understand: https://github.com/NixOS/nixpkgs/blob/master/pkgs/build-support/setup-hooks/multiple-outputs.sh#L93

Things done

• Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS linux)
• Built on platform(s)
  • NixOS
  • macOS
  • other Linux distributions
• Tested via one or more NixOS test(s) if existing and applicable for the change (look inside nixos/tests)
• Tested compilation of all pkgs that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Determined the impact on package closure size (by running nix path-info -S before and after)
• Ensured that relevant documentation is up to date
• Fits CONTRIBUTING.md.

cc @matthewbauer @misuzu

[Ericson2314]

Can we just get rid of that is64bit check?, or make it just guard the lib64 part? I rather just delete stupid conditional code than hack around it.

[misuzu]

There is another fix already in master: b6ae819

[lopsided98]

That fix looks pretty broken IMO. It just creates an empty output to satisfy Nix, but the target libraries will be missing.

Once master gets merged into staging, I will revert it in this PR.

[Ericson2314]

@lopsided98 just updated staging for you

[Ericson2314]

Still want [ -z "$enableMultilib" ] though, right?

[lopsided98]

Those directories should always be created. I think the logic for the conditional below is correct (other than is64bit referring to wrong platform). On 32-bit platforms, there should only be a lib directory. On 64-bit platforms, there should be lib and lib64, which should be linked unless multilib is enabled.

It doesn't seem to break anything to have lib64 on 32-bit as well (since it happens unintentionally in the armv7l cross compiler), but that kind of change seems out of scope for this PR.

[Ericson2314]

Well, I'm not sure we ever don't do enableMultilib, but I think without multilib it never prefixes any directories like that.

[lopsided98]

Multilib is disabled for normal builds. In all the 64-bit GCCs I have built, lib64 is a symlink to lib. The only place I can find in nixpkgs where enableMultilib is true is gcc_multi.

[Ericson2314]

Oh oops I forgot the -z and got it backwards.

[vcunat]

I see no issue with this. Thanks.