Re: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?

• Subject: Re: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
• From: Gregor Schmid <Gregor.Schmid@?.de>
• Date: Sun, 19 Jul 2015 21:15:36 +0200

Hi Michael,

unfortunately there are many common cases where the process started by
QF-Test is not the same as the one that connects to QF-Test. In some
cases even the parent/child process relationship becomes impossible to
track. The most notable example is Java WebStart which always starts a
second process and terminates the first one.

Thus, whether a process terminates or not is irrelevant, QF-Test has
to wait for an SUT connection until the entire timeout period has
elapsed.

In addition to that, you need to be able to access already terminated
clients, e.g. when checking their exit code or retrieving their
output.

Consequently a NoSuchClientException is only thrown if no client be
the given name was ever started (or is no longer among the recently
terminated processes).

That said, your short cut implementation for the case where the client
terminates immediately is basically fine. I'd modify it slightly, but
I guess that's just a matter of taste:

+ Start SUT client: $(client)
+ Try
  + Wait for client to terminate (timeout 2000)
  + call qfs.utils.testrun.stop.stopTestRun  // terminated quickly
  + Catch ClientNotTerminatedException
  + Try
    + Wait for client to connect
    + Catch ClientNotConnectedException
      + ... deal with it somehow ...

Best regards,
    Greg

Michael Pruemm <eso@?.org> writes:

> Hi all,
>
> I just had the following interesting scenario: My tests start the
> client, which immediately fails because the JVM cannot allocate
> sufficient memory. Right after the "Start SUT client", I wait for the
> client to connect. This fails of course after the timeout with a
> "ClientNotConnectedException".
>
> This is of course technically correct, since the client did not connect.
>
> However, since the client failed to start immediately, it was not even
> running when the "Wait for client to connect" started to execute. In
> that case, I would expect a "NoSuchClientException".
>
> The description for ClientNotConnectedException seems to back me up here:
>
>> It differs from a NoSuchClientException in that there is an active
>> process for that name but no RMI connection.
>
> I just confirmed that behavior by creating a new "Wait for client to
> connect" node with a time-out of 2s and a client name of "foo". When I
> execute this, it leads to a ClientNotConnectedException instead of a
> NoSuchClientException.
>
>
> To work around this problem, I am using this:
>
> + Start SUT client: $(client), delay after 2000 -- client starts slowly
> + If "${qftest:client.exitcode.$(client)}" != ""
>   + call qfs.utils.testrun.stop.stopTestRun
> + Try
>   + Wait for client to connect
>   + Catch ClientNotConnectedException
>     + ... deal with it somehow ...
>
> Is this the best approach?
>
>
> Here a few more details about my setup: QFTest 4.0.5; the client is a
> Java application that gets started via a shell script.
>
> - Michael
> _______________________________________________
> qftest-list mailing list
> qftest-list@?.de
> http://www.qfs.de/mailman/listinfo/qftest-list

--
Gregor Schmid

E: gregor.schmid@?.de
T: +49 8171 38648-11
F: +49 8171 38648-16

Quality First Software GmbH | www.qfs.de
Tulpenstr. 41 | 82538 Geretsried | Germany
GF Gregor Schmid, Dr. Martina Schmid, Karlheinz Kellerer
HRB München 140833

• Follow-Ups:
  • Re: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
    • From: Michael Pruemm

• References:
  • [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
    • From: Michael Pruemm

• Prev by Date: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
• Next by Date: Re: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
• Previous by thread: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
• Next by thread: Re: [QF-Test] Shouldn't "Wait for client to connect" check if the client is (still) running?
• Index(es):
  • Date
  • Thread