assertion failure in trans_alt when combining if, nested patterns

[nikomatsakis]

The following test:

enum side { mayo, catsup, vinegar }
enum order { hamburger, fries(side), shake }
enum meal { to_go(order), for_here(order) }

fn foo(m: @meal, cond: bool) {
    match *m {
      to_go(_) => { }
      for_here(_) if cond => {}
      for_here(hamburger) => {}
      for_here(fries(s)) => {}
      for_here(shake) => {}
    }
}

fn main() {
    foo(@for_here(hamburger), true)
}

results in an assertion failure in trans:

;RUST_LOG=rustc=0 rustc ~/tmp/pat-bug-in-trans.rs 
rust: task failed at 'Assertion m.len() > 0u || is_some(chk) failed', /Users/nmatsakis/versioned/rust-gold/src/rustc/middle/trans/alt.rs:431
error: internal compiler error: unexpected failure
note: the compiler hit an unexpected failure path. this is a bug
note: try running with RUST_LOG=rustc=0,::rt::backtrace to get further details and report the results to github.com/mozilla/rust/issues
rust: task failed at 'explicit failure', /Users/nmatsakis/versioned/rust-gold/src/rustc/driver/rustc.rs:264
rust: domain main @0x7f8d99800010 root task failed
rust: task failed at 'killed', /Users/nmatsakis/versioned/rust-gold/src/libcore/task.rs:520

The problem has to do with the if, something in the logic there gets a bit confused.

[catamorphism]

This looks very similar to #2869

[catamorphism]

I have a fix for this; running tests. Basically, enter_default needs to know whether a pattern has a guard, and not consider a guarded pattern to be a default.

[catamorphism]

Oops, it's not as simple as that. Still working on it.

[graydon]

Bumping this since it can always be worked around. Should fix but non blocking.

[pnkfelix]

Not critical for 0.6; de-milestoning

[catamorphism]

Nominating for milestone 5, production-ready. I have an old branch that partially fixes it, but it needs more work.

[graydon]

reproduced, and people run into it often, sure

[graydon]

accepted for production-ready milestone

[jasom]

Is this fixed in 0.7? I think I've run into it recently. Is there a workaround?

[pnkfelix]

@jasom: combine the caes with the same patterns, moving their ifs to the right-hand side of the pattern?

[msullivan]

I think I have a pretty good idea of how to fix this, although it seems a little subtle/hacky.