[darcs-users] patch: add makefile target named "disttest" which does the test currently written in _darcs/prefs/prefs

David Roundy daveroundy at gmail.com
Thu Sep 4 14:11:59 UTC 2008


On Tue, Sep 2, 2008 at 5:16 AM, Eric Y. Kow <eric.kow at gmail.com> wrote:
> On Mon, Sep 01, 2008 at 17:55:03 -0400, David Roundy wrote:
>> The maintainers can certainly use different settings for the test, but
>> I don't think there's a test that's much faster that will really help
>> significantly--except that we could skip the unit tests, which involve
>> code that is almost never touched.  In fact, with some work, we could
>> potentially only rerun the unit tests when code that affects them is
>> modified.  That could save a lot of time--but would be hard to code.
>
> I realise this may be a bit extreme, but maybe we could do away with the
> testpref altogether and just ask people to use a send prehook?

No, I think it's important to have some documentation in the code that
indicates how one can test darcs.  I know most people don't want to do
so, but it shouldn't be hard for them to do so.

> That puts extra burden on us to test, but we're running the suite
> anyway.
>
> No strong feelings on this.  I think my hope was just to lower the
> barrier to entry for one-off contributors who may be suprised by the
> test suite being run.  "I just wanted to fix a typo!" or in Zooko's
> case, "make a ChangeLog entry!" (Zooko was not aware of --no-test, and I
> think the same may be true for many potential contributors)

I think it'd be more reasonable to change the default behavior of
record.  Not that I'm advocating this change, but it'd be more
reasonable than failing to use best-practice on darcs itself.  This is
what the setpref test functionality is for, running the test suite,
and if it really is too onerous for contributors to darcs to learn to
use this aspect of darcs' functionality, then we should fix the
problem for other darcs-using projects as well.

David


More information about the darcs-users mailing list