Re: [gnet-devel] GNet tests

[Tim-Philipp Müller <t i m zen co uk>]

On Tue, 2008-12-23 at 19:01 +0100, Robert Pes wrote:

Hi,

> My name is Robert Pes and I work for Sun Microsystems. I'm just about
> to start to work with my colleague David Fan on integration of your
> GNet library into Java Desktop System (JDS). I'll be responsible in
> particular for testing of the library after its integration into the
> JDS consolidation. The goal is to make sure that the all the library
> features work as expected.

That's great! However, you do realise that GNet is basically in
maintenance-only mode for all practical purposes, right? It will
hopefully be replaced/superseded by a spiffy new network API in GLib
(gnio etc.) in the not too distant future.


> I noticed there were some test suites attached to your source tarball.
> I tried to run them using "make check" but there were no visible
> results. Is there any trick in running them?

You probably need the 'check' library installed (check.sf.net IIRC),
optionally also valgrind + valgrind-headers if you want valgrind support
(and your platform supports it). Same as you need to run 'make check'
for GStreamer basically.


> I was also wondering whether there is any test documentation
> describing the test scope, feature coverage and expected results. I
> didn't find anything either in available docs or the tests/ folder and
> the test source code is not that easy to read (I'm rather Java
> programmer than C one).

'make help' in $srcdir/tests/check will give you an overview of the
options. 'make check' or 'make check-valgrind' in the top-level dir
should be enough for most purposes.

There is no further documentation than that I'm afraid.

The output should look something like:

Running suite(s): GTcpSocket
...........125 attempts
100%: Checks: 1, Failures: 0, Errors: 0
PASS: gnet/gnettcpsocket
Running suite(s): GNetABI
...


> I was also thinking of running one of the applications using GNet,
> like e.g. Mail Notify, to verify functionality of your library. Do you
> think it's a good idea?

How could that possibly not be a good idea? :)


> Which application would in your opinion provide the best coverage of
> the supported features?

I don't think there's one application, as applications tend to use only
a very limited subset of features.

Cheers
 -Tim