Trace_macros!() is not effectively gated

[bluss]

Since the trace_macros feature has effect before compilation stops, it's not really effectively gated.

#![feature(trace_macros)]
trace_macros!(true);

fn main() {
    println!("{}", 1);
}

output:

println! { "{}" , 1 }
print! { concat ! ( "{}" , "\n" ) , 1 }
<anon>:1:1: 1:26 error: #[feature] may not be used on the stable release channel
<anon>:1 #![feature(trace_macros)]
         ^~~~~~~~~~~~~~~~~~~~~~~~~
error: aborting due to previous error

[durka]

Compilation doesn't complete, so is it really a stability hazard?

[pnkfelix]

I'd say that gate is pretty effective, but the compilation is indeed not effect-free ... ha ha ha...

[nikomatsakis]

Discussed in @rust-lang/compiler triage mtg. Since compilation doesn't succeed, it's not really possible to publish crates that rely on this. If we take it away, some people could get upset, but it won't cause widescale breakage (and people can always get upset when you take something away, feature-gate or no). Therefore, P-low.

triage: P-low