Improve failed test diagnostics.

src/ptvsd/server/cli.py

@@ -337,7 +337,7 @@ attach_pid_injected.attach(port={port}, host=host, client={client}, log_dir=log_
 
 
 def main():
-    original_argv = sys.argv
+    original_argv = list(sys.argv)
     try:
         sys.argv[1:] = parse(sys.argv[1:])
     except Exception as ex:

tests/debug/session.py

@@ -561,8 +561,9 @@ class Session(object):
         )
 
         if start_request.response is not None:
-            # It was an immediate response - configuration is not possible. Just get
-            # the "process" event, and return to caller.
+            # It was an immediate response - either the request failed, or there is
+            # no configuration stage for this debug session.
+            start_request.response.result  # raise exception if failed
             return self.wait_for_process()
 
         # We got "initialized" - now we need to yield to the caller, so that it can

tests/net.py

@@ -63,11 +63,11 @@ def wait_until_port_is_listening(port, interval=1, max_attempts=1000):
         try:
             log.info("Probing localhost:{0} (attempt {1})...", port, i)
             sock.connect(("localhost", port))
-        except socket.error:
+        except socket.error as exc:
             # The first attempt will almost always fail, because the port isn't
             # open yet. But if it keeps failing after that, we want to know why.
             if i > 1:
-                log.exception()
+                log.warning("Failed to connect to localhost:{0}:\n{1}", port, exc)
             time.sleep(interval)
         else:
             log.info("localhost:{0} is listening - server is up!", port)