Don't do bare using Foo

src/ElasticClusterManager.jl

@@ -1,15 +1,19 @@
 module ElasticClusterManager
 
-using Distributed
-using Sockets
+# We don't do `using Foo`
+# We either do `using Foo: bar`, or we do `import Foo`
+# https://github.com/JuliaLang/julia/pull/42080
+
+import Distributed
+import Sockets
 import Pkg
 
-export launch, manage, kill, init_worker, connect
+# Bring some names into scope, just for convenience:
+using Distributed: launch, manage, kill, init_worker, connect
 
+export launch, manage, kill, init_worker, connect
 export ElasticManager, elastic_worker
 
-import Distributed: launch, manage, kill, init_worker, connect
-
 function worker_cookie()
   Distributed.init_multi()
   return Distributed.cluster_cookie()

src/elastic.jl

@@ -11,9 +11,9 @@ else
     my_errormonitor(t) = nothing
 end
 
-struct ElasticManager <: ClusterManager
-    active::Dict{Int, WorkerConfig}        # active workers
-    pending::Channel{TCPSocket}          # to be added workers
+struct ElasticManager <: Distributed.ClusterManager
+    active::Dict{Int, Distributed.WorkerConfig}        # active workers
+    pending::Channel{Sockets.TCPSocket}          # to be added workers
     terminated::Set{Int}             # terminated worker ids
     topology::Symbol
     sockname
@@ -38,7 +38,7 @@ struct ElasticManager <: ClusterManager
 
         t1 = @async begin
             while true
-                let s = accept(l_sock)
+                let s = Sockets.accept(l_sock)
                     t2 = @async process_worker_conn(lman, s)
                     my_errormonitor(t2)
                 end
@@ -83,7 +83,7 @@ function process_pending_connections(mgr::ElasticManager)
     while true
         wait(mgr.pending)
         try
-            addprocs(mgr; topology=mgr.topology)
+            Distributed.addprocs(mgr; topology=mgr.topology)
         catch e
             showerror(stderr, e)
             Base.show_backtrace(stderr, Base.catch_backtrace())
@@ -144,7 +144,7 @@ function elastic_worker(cookie, addr="127.0.0.1", port=9009; stdout_to_master=tr
     c = connect(addr, port)
     write(c, rpad(cookie, HDR_COOKIE_LEN)[1:HDR_COOKIE_LEN])
     stdout_to_master && redirect_stdout(c)
-    start_worker(c, cookie)
+    Distributed.start_worker(c, cookie)
 end
 
 function get_connect_cmd(em::ElasticManager; absolute_exename=true, same_project=true)

test/elastic.jl

@@ -7,7 +7,8 @@
     # launch worker
     old_cmd = ElasticClusterManager.get_connect_cmd(em)
     new_cmd = `$(old_cmd) --coverage=user`
-    run(`sh -c $(new_cmd)`, wait=false)
+    # run(`sh -c $(new_cmd)`) # comment out this line when you are finished debugging
+    run(`sh -c $(new_cmd)`; wait=false) # uncomment this line when you are finished debugging
 
     # wait at most TIMEOUT seconds for it to connect
     @test :ok == timedwait(TIMEOUT) do