runtime: Remove last slab dependency

[bdonlan]

This removes the last slab dependency by replacing the current slab-based
JoinHandle tracking with one based on HashMap instead.

Motivation

Cleaning up dependencies

Solution

Slab was only used to track JoinHandles to ensure that all threads are fully cleaned up before returning from shutdown (this, in turn, is intended to avoid valgrind false-positives). Since this path isn't particularly performance-critical, this change replaces this use case with a HashMap instead.

[bdonlan]

CI blocked on tokio-rs/loom#175

[carllerche]

loom 0.3.6 is out.

[carllerche]

LGTM 👍 thanks.

[Darksonn]

Loom claims to be unhappy about a deadlock in shutdown code.

[bdonlan]

Hmm?

thread 'runtime::tests::loom_pool::group_b::pool_shutdown' panicked at 'assertion failed: state.notified', /home/runner/.cargo/registry/src/github.com-1ecc6299db9ec823/loom-0.3.6/src/rt/notify.rs:112:13
stack backtrace:
   0: std::panicking::begin_panic
   1: scoped_tls::ScopedKey<T>::with
   2: loom::rt::notify::Notify::wait
   3: loom::thread::JoinHandle<T>::join
   4: <tokio::runtime::blocking::pool::BlockingPool as core::ops::drop::Drop>::drop
   5: core::ptr::drop_in_place
   6: core::ops::function::FnOnce::call_once{{vtable.shim}}
   7: loom::rt::scheduler::spawn_threads::{{closure}}::{{closure}}
   8: generator::stack::StackBox<F>::call_once
   9: generator::gen_impl::gen_init
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
thread 'runtime::tests::loom_pool::group_b::pool_shutdown' panicked at 'deadlock; threads = [(Id(0), Blocked), (Id(1), Terminated), (Id(2), Terminated), (Id(3), Terminated)]', /home/runner/.cargo/registry/src/github.com-1ecc6299db9ec823/loom-0.3.6/src/rt/execution.rs:207:13

We're deadlocked on JoinHandle.wait() but the blocked thread is the only one alive. There shouldn't be any JoinHandles for the initial thread, so it can't wait on itself. Is this a bug in loom? @carllerche

[bdonlan]

I'm unable to reproduce this failure locally. Since a rebase is needed anyway, rebasing and seeing if it fails again.

[bdonlan]

As far as I can tell, this seems like it might be a loom bug - filed tokio-rs/loom#177

[bdonlan]

Found the loom bug - we're waiting on joinhandles in indeterminate order. Fixing.

[carllerche]

Thanks 👍