Change timeout kill signals to SIGQUIT from SIGTERM

We've been having a lot of timeouts on CI recently.  Our coredumps might
be helpful in tracking these down, but when we send `SIGTERM` or
`SIGKILL` we don't get coredumps.  This changes timeout signals to
generally use `SIGQUIT` instead of `SIGTERM`, when sending the first kill signal.

stdlib/Distributed/src/managers.jl

@@ -735,8 +735,12 @@ function kill(manager::LocalManager, pid::Int, config::WorkerConfig; exit_timeou
 
         # Check to see if our child exited, and if not, send an actual kill signal
         if !process_exited(config.process)
-            @warn("Failed to gracefully kill worker $(pid), sending SIGTERM")
-            kill(config.process, Base.SIGTERM)
+            @warn("Failed to gracefully kill worker $(pid), sending SIGQUIT")
+
+            # Support overriding the termination signal for Base.runtests() on CI.  We often want to
+            # get a coredump from a worker that times out, and we need to send some other signal than
+            # `Base.SIGTERM`, so allow that to be overridden by an environment variable.
+            kill(config.process, Base.SIGQUIT)
 
             sleep(term_timeout)
             if !process_exited(config.process)

stdlib/LibGit2/test/libgit2-tests.jl

@@ -62,7 +62,7 @@ function challenge_prompt(cmd::Cmd, challenges)
             end
 
             if process_running(p)
-                kill(p)
+                kill(p, Base.SIGQUIT)
                 put!(timer, :timeout)
             elseif success(p)
                 put!(timer, :success)

stdlib/Profile/test/runtests.jl

@@ -182,7 +182,7 @@ let cmd = Base.julia_cmd()
     t = Timer(120) do t
         # should be under 10 seconds, so give it 2 minutes then report failure
         println("KILLING BY PROFILE TEST WATCHDOG\n")
-        kill(p, Base.SIGTERM)
+        kill(p, Base.SIGQUIT)
         sleep(10)
         kill(p, Base.SIGKILL)
     end
@@ -210,7 +210,7 @@ if Sys.isbsd() || Sys.islinux()
             t = Timer(120) do t
                 # should be under 10 seconds, so give it 2 minutes then report failure
                 println("KILLING BY PROFILE TEST WATCHDOG\n")
-                kill(p, Base.SIGTERM)
+                kill(p, Base.SIGQUIT)
                 sleep(10)
                 kill(p, Base.SIGKILL)
                 close(iob)

stdlib/Sockets/test/runtests.jl

@@ -18,7 +18,7 @@ function killjob(d)
         ccall(:uv_kill, Cint, (Cint, Cint), getpid(), SIGINFO)
         sleep(5) # Allow time for profile to collect and print before killing
     end
-    ccall(:uv_kill, Cint, (Cint, Cint), getpid(), Base.SIGTERM)
+    ccall(:uv_kill, Cint, (Cint, Cint), getpid(), Base.SIGQUIT)
     nothing
 end
 sockets_watchdog_timer = Timer(t -> killjob("KILLING BY SOCKETS TEST WATCHDOG\n"), 600)

test/threads.jl

@@ -255,7 +255,7 @@ cmd = """
     sleep(100)
     isopen(stdin) || exit()
     println(stderr, "ERROR: Killing threads test due to watchdog expiry")
-    ccall(:uv_kill, Cint, (Cint, Cint), $(getpid()), Base.SIGTERM)
+    ccall(:uv_kill, Cint, (Cint, Cint), $(getpid()), Base.SIGQUIT)
 """
 proc = open(pipeline(`$(Base.julia_cmd()) -e $cmd`; stderr=stderr); write=true)