SIGTERM test leaks stderr interrupt trace about distributed.jl

[IanButterworth]

Actually, this was #93

---

I can't understand why we're seeing

Warning: rmprocs: process 1 not removed

because it's coming from

Distributed.jl/src/cluster.jl

Line 1299 in 8c03305

atexit(terminate_all_workers)

Distributed.jl/src/cluster.jl

Lines 1246 to 1251 in 8c03305

	function terminate_all_workers()
	myid() != 1 && return
	if nprocs() > 1
	try
	rmprocs(workers(); waitfor=5.0)

Distributed.jl/src/cluster.jl

Lines 982 to 989 in 8c03305

	function workers()
	allp = procs()
	if length(allp) == 1
	allp
	else
	filter(x -> x != 1, allp)
	end
	end

Distributed.jl/src/cluster.jl

Lines 1043 to 1049 in 8c03305

	function _rmprocs(pids, waitfor)
	lock(worker_lock)
	try
	rmprocset = Union{LocalProcess, Worker}[]
	for p in pids
	if p == 1
	@warn "rmprocs: process 1 not removed"

And I don't understand how it's possible for nprocs() > 1 and workers() to contain 1

[IanButterworth]

I dont think this is happening because of

Distributed.jl/src/cluster.jl

Line 1299 in 8c03305

atexit(terminate_all_workers)

because we don't see this log

Distributed.jl/src/cluster.jl

Line 1253 in 2b23ae4

@warn "Forcibly interrupting busy workers" exception=_ex

[IanButterworth]

I believe this is coming from this test which swallows the log

Distributed.jl/test/distributed_exec.jl

Lines 1919 to 1925 in 2b23ae4

	@test_logs (:warn, r"sending SIGTERM") begin
	remote_do(w) do
	# Cause the 'exit()' message that `rmprocs()` sends to do nothing
	Core.eval(Base, :(exit() = nothing))
	end
	wait(rmprocs([w]))
	end