On Wed, Sep 30, 2009 at 2:49 PM, Hyrum K. Wright <hyrum_at_hyrumwright.org> wrote:
> We currently have a couple of tests which conditionally run if the
> sqlite3 python module is found. As we move more working copy stuff
> into sqlite, the number of tests that will be needed to be skipped
> will increase, and the transparency of *why* they need to be skipped
> will decrease. For this reason, I'd like to propose that we require
> the python sqlite3 module be installed when running the tests.
On a related topic, I think it's wrong to put tests relying on the
internals of subversion (white box tests) together with the acceptance
tests (black box tests).
Together, or at least, not clearly marked as being dependent on the
internals of Subversion.
Tests are there not only now to help write the code, but they remain
in the suite to test for regressions in the years to come.
Specifically for this regression testing, I'm wary of changing
existing tests in the suite, as that typically indicates changed
functionality or cmdline output that people rely on. OTOH, it should
be clear that these tests can be will have to be changed when
internals change.
I'm not suggesting to move those tests out of the python test suite,
but just to mark them clearly, with a comment or so.
If people don't mind (or can't be bothered with it), I'll do it
myself, later tonight.
Lieven
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2402249
Received on 2009-09-30 21:27:59 CEST