remove ref-counting Gc<T>

[pnkfelix]

Summary

Remove the reference-counting based Gc<T> type from the standard
library and its associated support infrastructure from rustc.

Doing so lays a cleaner foundation upon which to prototype a proper
tracing GC, and will avoid people getting incorrect impressions of
Rust based on the current reference-counting implementation.

(rendered)

[steveklabnik]

Big 👍 here.

[CloudiDust]

+1, and we should make the plan known to as many in the community as possible, so people will not assume that Rust is never going to have a proper tracing GC.

[Ericson2314]

Sounds good!

[nrc]

Take 'em out - keep things clean as possible. Chances are the future GC will end up having different lang items, and in any case it is no bother to add more lang items.

[nikomatsakis]

👍

[pnkfelix]

(hmm, the sample program I just added to the RFC seems to consistently cause the playpen timeout to trigger. AFAICT it seems to be triggered by my call to timer::sleep on a non-main task; you can remove that to run it to completion in the playpen if you want.)

[thehydroimpulse]

👍

[jsanders]

@pnkfelix: Would it be possible to just include output from a run of the sample program, so that we don't need to run it in the playpen (or compile it locally) to see what it's demonstrating?

[brson]

Accepted as RFC 67. Discussion. Tracking.