Irrefutable bindings matching code is broken

[bblum]

fn main() {
    let x = "hello";
    let ref y = x;
    // let y = match x { ref y => y };  // this way works fine
    error!("%?", *y);
}

produces:

rustc: /home/bblum/rust/src/llvm/lib/VMCore/Instructions.cpp:280:
void llvm::CallInst::init(llvm::Value*, llvm::ArrayRef<llvm::Value*>,
const llvm::Twine&): Assertion `(i >= FTy->getNumParams() ||
FTy->getParamType(i) == Args[i]->getType()) && "Calling a function
with a bad signature!"' failed.
Aborted (core dumped)

Happens with structs, tuples, etc too.

[nikomatsakis]

for some reason that is fairly unclear to me, trans/alt has a completely distinct code path for irrefutable patterns. that path should be deleted and they should all use the normal alt path, imo. anyway, that particular path is hard-coded to copy and doesn't know about ref bindings.

[nikomatsakis]

I changed the title to my preferred solution.

[catamorphism]

Ugh, this has been a pain in the you-know-what. Not going to be ready for 0.5

[Aatch]

Also causes this strange behaviour:

fn main() -> () {
    let a = ~(1, 2);
    test(a);
}

fn test<T,U>(a : &(T, U)) {
    let (ref _x, ref _y) = *a;
}

Returns error: moving out of dereference of immutable & pointer

However:

fn main() -> () {
    let a = ~(1, 2);
    test(a);
}

fn test<T,U>(a : &(T, U)) {
    let (_a, _b) = match *a {
        (ref a, ref b) => (a, b)
    };
}

Is fine.

[alexcrichton]

Nominating for the well-defined milestone.

This makes working with non-copyable tuples laborious, and actually has serious performance implications for destructuring assignment (at least I think that's related to this bug, copies are generated when they shouldn't be). Failing the well-defined milestone, I think this would fit well into the feature-complete milestone.

[catamorphism]

I don't feel like this fits under anything except production-ready, but I agree with nominating it.

[nikomatsakis]

For what it's worth, I don't think that this is necessarily the right fix anymore. But the bug is real.

[nikomatsakis]

Update the title to reflect the general bug.

[nikomatsakis]

Very close to fixing this. I've got a branch that's been slowly progress towards stability.

[graydon]

accepted for backwards-compatible milestone

[catamorphism]

a48ca32 fixes