Use associated constants in std::num::{Zero,One}

[tamird]

This refactor causes the following ICE:

$ /usr/bin/make -j3 check-stage1 RUST_BACKTRACE=1
cfg: version 1.2.0-dev (72e9aaa7a 2015-05-21) (built 2015-05-21)
cfg: build triple x86_64-apple-darwin
cfg: host triples x86_64-apple-darwin
cfg: target triples x86_64-apple-darwin
cfg: disabling rustc optimization (CFG_DISABLE_OPTIMIZE)
cfg: enabling debug assertions (CFG_ENABLE_DEBUG_ASSERTIONS)
cfg: enabling debuginfo (CFG_ENABLE_DEBUGINFO)
cfg: host for x86_64-apple-darwin is x86_64
cfg: os for x86_64-apple-darwin is apple-darwin
cfg: disabling C++ optimization (CFG_DISABLE_OPTIMIZE_CXX)
cfg: good valgrind for x86_64-apple-darwin is
cfg: using CC=clang (CFG_CC)
cfg: using CXX=clang++ (CFG_CXX)
cfg: disabling valgrind run-pass tests
cfg: no xelatex found, disabling LaTeX docs
cfg: no pandoc found, omitting PDF and EPUB docs
cfg: including test rules
cfg: javac not available, skipping lexer test...
rustc: x86_64-apple-darwin/stage0/lib/rustlib/x86_64-apple-darwin/lib/libcore
error: internal compiler error: unexpected panic
note: the compiler unexpectedly panicked. this is a bug.
note: we would appreciate a bug report: https://github.com/rust-lang/rust/blob/master/CONTRIBUTING.md#bug-reports
note: run with `RUST_BACKTRACE=1` for a backtrace
thread 'rustc' panicked at 'path not fully resolved: PathResolution { base_def: DefTyParam(TypeSpace, 0, DefId { krate: 0, node: 115188 }, "A"(326)), last_private: LastMod(AllPublic), depth: 1 }', /Users/rustbuild/src/rust-buildbot/slave/snap3-mac/build/src/librustc/middle/def.rs:81
stack backtrace:
1: 0x1088cb99f - sys::backtrace::write::ha6a0ef69d836f288KVr
2: 0x1088d04b2 - panicking::on_panic::h79a3ac05487064ffNVv
3: 0x1088aca75 - rt::unwind::begin_unwind_inner::h53568b1e4e1d5c49wDv
4: 0x1088ad5cc - rt::unwind::begin_unwind_fmt::h42754663daaabff7CCv
5: 0x1076ff7f8 - middle::ty::resolve_expr::h01be9bdb60bc5d07Tg6
6: 0x10784c2c2 - middle::ty::expr_kind::he9ffd29dd326a14bZi6
7: 0x10784c06d - middle::ty::expr_is_lval::h55b067779b7ace9bWh6
8: 0x1074b5244 - check::check_expr_with_unifier::h11317710005126736159
9: 0x10747bc0d - check::op::check_binop::h63cfeed1e166bc9asYm
10: 0x1074b32ce - check::check_expr_with_unifier::h11317710005126736159
11: 0x1074c965f - check::check_expr_with_unifier::h1631999723096368142
12: 0x107487f3a - check::check_block_with_expected::h959a715da7c43d7bvds
13: 0x10749ca0d - check::check_expr_with_unifier::check_then_else::h5fd509e0bd507d79TCq
14: 0x1074b35b0 - check::check_expr_with_unifier::h11317710005126736159
15: 0x10748829b - check::check_block_with_expected::h959a715da7c43d7bvds
16: 0x10746b828 - check::check_fn::heab4df53bfe14a32pSn
17: 0x1074837e5 - check::check_bare_fn::hc942e99ce3f1c922YHn
18: 0x10748f107 - check::check_method_body::h8bf3cf4630998608Zjo
19: 0x10748168b - check::CheckItemBodiesVisitor<'a, 'tcx>.Visitor<'tcx>::visit_item::h9347c778160154f51En
20: 0x107481aee - check::CheckItemBodiesVisitor<'a, 'tcx>.Visitor<'tcx>::visit_item::h9347c778160154f51En
21: 0x107545c5a - check_crate::closure.38461
22: 0x107541087 - check_crate::h28a26a622d6b3e59tCC
23: 0x1070c7d54 - driver::phase_3_run_analysis_passes::he5b5c42951feb035tGa
24: 0x1070aa378 - driver::compile_input::h78293093b86c5c68Qba
25: 0x107168623 - run_compiler::h1e722a2786102a3765b
26: 0x107165d8a - boxed::F.FnBox<A>::call_box::h15307864961899385171
27: 0x1071652d7 - rt::unwind::try::try_fn::h3375314414886690993
28: 0x1088d2688 - rust_try_inner
29: 0x1088d2675 - rust_try
30: 0x1071655ae - boxed::F.FnBox<A>::call_box::h7666835191292233835
31: 0x1088cefad - sys::thread::Thread::new::thread_start::h41e207b511aa1bfbvYu
32: 0x7fff8ea0c267 - _pthread_body
33: 0x7fff8ea0c1e4 - _pthread_start
make: *** [x86_64-apple-darwin/stage0/lib/rustlib/x86_64-apple-darwin/lib/stamp.core] Error 101

[rust-highfive]

r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

[Gankra]

What's the status of this in the face of #25687 ?

[tamird]

It's a superset - I'll rebase this when #25687 lands.

[Gankra]

The extra bits just ICE, though, right?

[tamird]

Yeah - I'd file an issue but I wasn't able to reduce the ICE down

[quantheory]

We are in the unhappy position where trait-associated constants are not really useful, because you can't actually use them from type parameters.

For use of trait-associated constants in non-constant expressions, we need something similar to the existing type projections, so that the compiler isn't confused about constants that can't be resolved. For use in constant expressions, see: rust-lang/rfcs#1062.

[eddyb]

Like I told @tamird on IRC, this particular ICE is a mishandling of unresolved paths to associated items.
They can all be treated as if they were RvalueDatum because both methods and associated constants are that.

[quantheory]

@eddyb Ah, I see. We are asking if it is an lvalue before the type is checked, which is where resolve_ufcs is called. However, if that is addressed, I still don't think that this code will make it past check_const, due to #25046. In fact, #25091 puts in a check to simply reject such code, until we have something better that doesn't just hit an ICE.

[bluss]

This makes the trait less general (only available to types that can produce a zero in constants). I suppose that's useful and ok, it just needs to be mentioned.

[tamird]

@bluss this is true, but not such a big problem given https://github.com/rust-lang/rust/pull/25609/files#diff-9e764fdb4b37fceaf4fdf1ddb027dbcaR15

[bluss]

BigNum are allocating and can never impl the new trait (?) I don't think it's important.

[eddyb]

@bluss if the BigInt implementation uses a variant/tags for the non-allocating cases (fitting in u64, usually), you may be able to put that in a constant. It's not a given, though.

[alexcrichton]

I agree with @bluss that this probably excludes bignum from implementing the trait, but I'm somewhat hesitant on casting it aside as not important. The std::num module got by at 1.0 by moving all methods on primitives to inherent implementations, so there are actually 0 stable traits in std::num. One reason we didn't stabilize anything is that we couldn't strike the right balance between supporting primitives while allowing for other implementors.

These existence of these traits is necessitated by the iterators that use them, and I believe that's needed for inference to work out ok, so these need to stick around. I think it's fine to experiment with them having associated constants instead of methods, but we'd likely want to give everything a rethink still before stabilizing them.

[tamird]

As a temporary measure, #25091 rejects this code, so this PR is blocked behind a future change that allows associated constants to depend on Self.