Just a followup or two to what is a very interesting and relevant
discussion. WordNet-Similarity and WordNet-SenseRelate-* are actually
projects that are independent of QueryData at least in terms of who is
doing the developing and looking at these bug reports, so I can of ran
into this discussion by accident, but I'm glad I did.
Let me pose a question from the point of view of WordNet-Similarity and
then WordNet-SenseRelate-* as there is very much a chain reaction here,
WordNet-Similarity relies on WordNet-QueryData, and then
WordNet-SenseRelate relies on WordNet-Similarity.
We *think* that we are now (as of our newest release for
WordNet-Similarity) handling the case where a tester does not have
QueryData installed, and we now fail more gracefully. We think. :) But
we don't seem to be getting many test cases run against us these days,
and so we aren't sure.
Did we get "blacklisted" in general, and are we out of the testing
cycle? How can we get back into the cycle, if for no other reason than
to see that we are failing gracefully if QueryData isn't there? I am not
sure if the SenseRelate modules fail gracefully if WordNet-Similarity is
not there, they might not, but we'd like to work on it, and the testing
is quite helpful.
Is there a way to find out if a module is "out of the testing" cycle -
we have a few modules that get tested a lot (Text-Similarity or
Algorithm::Munkres, for example) but others that rarely if ever do. I
don't mind if we get a few failure reports piling up, we'll try and fix
them. It doesn't seem like a failing module is too bad a deal for a
tester, so I'm not sure exactly why testing would be disabled...
Cordially,
Ted
On Thu Mar 27 15:37:38 2008, ANDK wrote:
Show quoted text> I am a cpan tester and I have already disabled testing for
> WordNet-QueryData, WordNet-Similarity, WordNet-SenseRelate-AllWords.
>
> Sorry, this does not work out.
>
> No modules on CPAN I know of demands that the user set two environment
> variables, one for build time (WNHOME) and one for run time
> (WNSEARCHDIR). Or both for runtime? Sorry, I forget. I had to read the
> source code to find out and in the end I gave up.
>
> How about thinking of ways to configure wordnet during the PL phase,
> installing config data along with the module so that it will be able to
> run without an environment variable? At runtime you can still decide to
> die when things are not fitting together. Think of modules downstream
> that depend on yours. If everybody demands two environment variables to
> get a module running, this will get ridiculous quickly.
>
> It would also be a great idea to actually verify if wordnet is installed
> at the PL phase. Maybe Devel::CheckLib can help you with that or you can
> certainly write a little perl glue code. If WN is not there, just exit
> the Makefile.PL without writing a Makefile and without returning an error.
>
> Tell the user it did not work out and why. Then no make and no test
> phase will happen and you will never receive a test report from those
> who have wordnet not installed.
>
> It's a pity it happens that software producers and users annoy each
> other. Sorry to have played my part in the game.
>
> Imagine, we also only realize after testing what kind of module we just
> tested:-(