Indeed Mickael, that was the sort of setup I was
imagining, although you take a shared disk or so as a
prerequisite, while I was thinking of sockets on the network. A
disk might do the trick though.... good idea.
I have to think this over in detail. I will inform the list
when things get moving...
Anybody any remarks on Jubula? (or should I ask the jubula
mailing list ;-))
My point of view, is that it's much easier to write and maintain
tests with SWTBot than with Jubula.
The quality of the tooling for Java (Thanks to JDT) and the fluent
API of SWTBot make that it's really easy to write a test.
However if your user/functionnal tester hate programming or hate
having to deal with a textual langague, I understand they may prefer
Jubula.
Maybe a way to achieve that is to write 1 TestCase per
user, and write in your test some synchronization points.
You could use a file to write when a user performed a set
of action, and other user tests will wait for the file to
contain a specific token to go ahead.
For example
TestUser1() {
// Do Step 1
foo();
bar();
// Step 1 done
markStepAsDone(1);
waitForStepDone(3);
// do Step 4
foo();
bar2();
}
TestUser3() {
waitForStepDone(2);
// Do Step 3
...
}
Then you can start one instance of your app for each user
with a different TestCase, and the implementation of
waitForStepDone ensure the sequence of actions is well
performed.
However, I admit it is not so easy to maintain...
Please keep us informed if you make progress on this
topic, that's something that could be useful for my
product too ;)
Regards.
Le 13/04/2011 16:23, Tom Brus a écrit :
Thanks for the pointer Ketan, I will
have a look, but our schedule will probably not
allow us to maintain our own fork...
Brittle is indeed the word ;-) but then the
main aspect of our application is
multi user so we need to know if problems occur
there before shipping.... alas.
On Wed, Apr 13, 2011 at 6:26 AM, Tom Brus
<tombrus@xxxxxxxxx>
wrote:
> I am specially interested in these
issues because our application needs
> multi user testing and I need
to orchestrate parallel starts of our
> application on multiple machines. If
every instance is running its own test
> in its VM it is just harder to
synchronise.
Someone had forked SWTBot to add RMI support
in order to do just this.
They were building some plugins for pair
programming and needed to
orchestrate 2 eclipse instances from within
one test.
The fork is available at: https://github.com/szuecs/swtbot.
It is
slightly out of date, but you should be able
to apply those patches to
current master.
From prior testing, this form of testing is
very very brittle given
the number of moving parts. I'd suggest
keeping it to a bare minimum.
This message and any attachment
(the "message") is intended solely for the
addressees and is confidential. If you receive
this message by mistake, please delete it and
notify the sender immediately. Any use not in
accordance with its purpose, any out-spread or
disclosure, either as a whole or partially, is
prohibited except with formal approval. Internet
cannot guarantee the integrity of this message,
therefore BonitaSoft will not be liable for the
message if modified.