stdenv: throw from ghcjs.cc.outputs rather than ghcjs.cc

[ghost]

Description of changes

pkgsCross.ghcjs.stdenv.cc currently throws, and it is (a good) part of the Nix semantics that exceptions transmit no information to an enclosing tryEval. This makes it impossible to express the fact that the answer to "is the C compiler from GNU?" should be false when no compiler exists.

The reason why pkgsCross.ghcjs.stdenv.cc is a throw is explained here:

nixpkgs/pkgs/stdenv/cross/default.nix

Lines 66 to 71 in aa37158

	else if targetPlatform.isGhcjs
	# Need to use `throw` so tryEval for splicing works, ugh. Using
	# `null` or skipping the attribute would cause an eval failure
	# `tryEval` wouldn't catch, wrecking accessing previous stages
	# when there is a C compiler and everything should be fine.
	then throw "no C compiler provided for this platform"

which refers to this check in splice.nix:

nixpkgs/pkgs/top-level/splice.nix

Lines 58 to 63 in aa37158

	tryGetOutputs = value0: let
	inherit (builtins.tryEval value0) success value;
	in getOutputs (lib.optionalAttrs success value);
	getOutputs = value: lib.genAttrs
	(value.outputs or (lib.optional (value ? out) "out"))
	(output: value.${output});

Since the result of that check is always passed to getOutputs, which is strict in its argument's outputs attribute, we can simply move the throw deeper into the stdenv.cc attrset. This leaves room to provide a useful value for attributes like isGNU and isClang when the compiler is "none".

Hopefully this removes the need for 1ee0f4c from #208947.

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
  • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
• Fits CONTRIBUTING.md.

[Ericson2314]

This might still be a good idea, but note both of you that I made a stdenv.hasCC which can be used to guard accesses to e.g. stdenv.cc.isClang.

[ghost]

This might still be a good idea, but note both of you that I made a stdenv.hasCC which can be used to guard accesses to e.g. stdenv.cc.isClang.

Yeah, I used that in an earlier revision of this PR, but it really uglifies the caller because (for some reason I didn't track down) stdenv is sometimes missing the hasCC attribute. I think only cross-compiling stdenvs have that attribute. So callers need to do (stdenv.hasCC or true) && stdenv.cc.isGNU.

And of course, callers have to remember to do that. Which I'm sure they won't, unless they are using ghcjs.

So I went looking for a more automatic way to accomplish the goal and came up with this.

[ghost]

@ofborg eval

The Darwin builders appear to be stuck.