wcfs/pintimeout_kill: Fix killed testing if successful

In case 'test_wcfs_pintimeout_kill' is successful, WCFS kills the
client process. In order to prevent WCFS from killing the full
testing process, we therefore need to encapsulate the test code into
a subprocess, so that only this subprocess is killed in case of a
success.

wcfs/wcfs_test.py

@@ -44,6 +44,7 @@ import sys, os, os.path, subprocess
 from thread import get_ident as gettid
 from time import gmtime
 from errno import EINVAL, ENOTCONN
+from multiprocessing import Process, Value
 from resource import setrlimit, getrlimit, RLIMIT_MEMLOCK
 from golang import go, chan, select, func, defer, error, b
 from golang import context, errors, sync, time
@@ -1463,15 +1464,22 @@ def test_wcfs_pintimeout_kill():
     f = t.open(zf)
     f.assertData(['','','c2'])
 
-    # XXX move into subprocess not to kill whole testing
-    ctx, _ = context.with_timeout(context.background(), 2*tkill)
+    # move into subprocess to avoid killing testing process, in
+    # case test is successful
+    def test(state):
+        try:
+            _test(state)
+        except Exception:
+            state.value = 3
 
+    def _test(state):
+        ctx, _ = context.with_timeout(context.background(), 2*tkill)
         wl = t.openwatch()
         wg = sync.WorkGroup(ctx)
         def _(ctx):
             # send watch. The pin handler won't be replying -> we should never get reply here.
             wl.sendReq(ctx, b"watch %s @%s" % (h(zf._p_oid), h(at1)))
-        fail("watch request completed (should not as pin handler is stuck)")
+            state.value = 1
         wg.go(_)
         def _(ctx):
             req = wl.recvReq(ctx)
@@ -1485,10 +1493,22 @@ def test_wcfs_pintimeout_kill():
             )
             if _ == 0:
                 raise ctx.err()
-        fail("wcfs did not killed stuck client")
+            state.value = 2
         wg.go(_)
         wg.wait()
 
+    state = Value("b", 0) 
+    p = Process(target=test, args=(state,))
+    p.start()
+    p.join()
+    if state.value != 0:
+        state = (
+            "watch request completed (should not as pin handler is stuck)",
+            "wcfs did not killed stuck client",
+            "error in test code",
+        )[state.value - 1]
+        fail(state)
+
 
 # watch with @at > head - must wait for head to become >= at.
 # TODO(?) too far ahead - reject?