Looks related to #42658 (I’m not in the position to judge whether it’s better or worse than that).

Bumping for review, please.

@peti @ryantm @roberth like @cocreature mentions this is indeed an alternative to #42658.

Advantages of #42658 compared to this PR:

• haskellPackages.callCabal2nix: accept extraCabal2nixOptions argument #42658 doesn't introduce a new function.

Disadvantages #42658 compared to this PR:

• haskellPackages.callCabal2nix: accept extraCabal2nixOptions argument #42658 convolutes the attribute set for overriding the callPackage arguments with the extraCabal2nixOptions. It's a nifty idea but a bit magical and harder to discover than callCabal2nixWithOptions. However this can be solved by adding some documentation to haskellPackages.callCabal2nix: accept extraCabal2nixOptions argument #42658.

I would go for #42658 if it introduces some documentation. What do you think?

It's fine with me, as long as one of them gets merged soon :)

I have no preference.

I prefer this one, becuase it seems the least path of resistance to actually get moving with this, and this has already sat for quite a while. 😃

@dhess @basvandijk I think we should bite the bullet on this, but also rewrite the function to take an argset instead of multiple arguments, such that next time, we can hide the addition of some field from the user without introducing YetAnotherCallCabal2nixWithFeatureX

@srhb I agree we should bite the bullet and I also think we should backport this to release-18.09.

With regards to your point about the argset argument, do you mean something like:

callCabal2nixWithOptions = name: src: { extraCabal2nixOptions ? "" }: args: ...

@basvandijk I was actually considering

callCabal2nixWithOptions = { name, src, args, extraCabal2nixOptions ? "" }: ...

This way it becomes a uniform invocation no matter what we add in the future. But if it's too radical a change, I don't feel strongly about it. I just thought it could help future-proof a bit.

@srhb I think all callSomething functions in Nix use multiple arguments instead of taking a set. I don't want to be inconsistent here. I've bitten the bullet and merged this.