[Bioperl-l] Clean handling of SQLite choice when running ./Build test

[Chris Fields]

On Apr 12, 2011, at 8:38 AM, Scott Cain wrote:

> Hi Lincoln and Chris,
> 
> I'm wondering what we should do about the SQLite option for testing
> when running ./Build test.  Currently (for me at least :-) when I
> select SQLite as the testing database when running perl Build.PL, the
> Bio::DB::GFF tests get "dubious" (which is fairly amazing, since there
> isn't a SQLite adaptor for Bio::DB::GFF), but it runs a different
> number of tests that planned, thus the dubious.  My inclination would
> be to not include SQLite as a testing option to avoid this problem,
> otherwise I suppose the Bio::DB::GFF test could be modified to use the
> memory adaptor when SQLite is the chosen testing database.
> 
> Scott

I think the latter option (use memory adaptor if SQLite is used) is best, it at least runs the Bio::DB::GFF tests.

This is also uncovering a general problem we should address down the line, mainly when running tests, if a resource isn't available there should be at least be a visible note about the point of failure instead of silently passing.  This set of tests was silently passing for me with SQLite selected, and I've been bitten by this with network tests quietly passing when the remote resource is no longer accessible.

chris