Hello all,

Given the precedent that the test framework uses tool_conf.xml.sample rather than tool_conf.xml, I would expect it to use example.loc.sample rather than example.loc when running unit tests with parameters whose values come from a loc file. This is not the case.

When writing unit tests of the Tool Shed, we can assume that as part of the automated installation example.loc will be generated from the supplied example.loc.sample file - and so for things like the nightly automated testing on the Tool Shed, the files will be the same.

However, on real Galaxy installations, the administrator will likely edit example.loc and it will not match example.loc.sample any more. This means that for writing a unit test, I cannot make any assumptions about the contents of example.loc, but it would be reasonable to expect example.loc.sample setup will still match the unit tests.

Is it sensible to make the test framework load *.loc.sample rather than *.loc for this reason?

If not, how exactly are unit tests dependent on values in a loc file meant to be written?

This is a problem for me writing tests for Blast2GO (property files in blast2go.loc), BLAST+ (databases in blastdb.loc etc). It hasn't yet been a problem for Effective T3 as here my effectiveT3.loc file still matches the sample file.

Thanks,

Peter