[darcs-users] [patch371] Implement a test-framework-based shell h... (and 3 more)

Eric Kow kowey at darcs.net
Wed Sep 1 10:32:52 UTC 2010


Thanks for the review Simon!  It's nice to see some more review voices;
I hope we can find more guest reviewers, more regularly (and grow the
Review Team over time)

On Tue, Aug 31, 2010 at 11:15:07 -0700, Simon Michael wrote:
> >hunk ./darcs.cabal 604
> >                      filepath     == 1.1.*,
> >                      QuickCheck   >= 2.1.0.0,
> >                      HUnit        >= 1.0,
> >+                     cmdlib       >= 0.2.1 && < 0.3,
> 
> -1. I know you have your reasons and this is just the test runner,
> not darcs itself, but I think cmdargs will succeed and it would be
> better to use/improve that.

I think it would be nice to have a list outlining the reasons we
currently prefer cmdlib to cmdargs.  Simon's right that we should
default to the most viable candidate unless we have a good reason,
and since we appear to have some good reasons, it'd be nice if they
were documented.

I suggest the most neutral way to do would by for cmdlib to provide
a bullet pointed list of differences between with cmdargs, and for
Darcs to just provide a link plus the things that matter for us.

Anyway, this is a non-blocker.  We can always switch back if we later
learn that actually cmdargs is a better fit.

> >+                     shellish     == 0.1.*,
> 
> The hackage page says "safety is sacrificed". Anything to be
> concerned about ? Will darcs tests built on this be reliable and
> deterministic ?
> 
> >                      test-framework             >= 0.2.2,
> >                      test-framework-hunit       >= 0.2.2,
> >                      test-framework-quickcheck2 >= 0.2.2
 
> +1. However, new packages (the 5 above) bring compatibility
> headaches. It looks like all these "should" work on windows, but
> until tested we should probably assume extra installation hassles
> for darcs developers on that platform. Likewise for folks wanting to
> build tests with GHC 6.10.

I appreciate these reservations.  I think the only thing we can do is
test it out on a build slave.

Meanwhile, I'm going to greenlight the bundle up to the patch that makes
-ftest mandatory.  Once we've determined that darcs-test is easy to
build on Windows, we can flip the switch.
 
> Summary: a good, high-value change, with concerns about the use of
> cmdlib and about unclear platform/compiler compatibility impact.

Another problem is conflicts in the darcs.cabal file with the patch
that makes -ftest mandatory.

If Petr could provide a conflict-fix patch (against mainline for now,
please), and if we knew more about installing this on Windows, then
it can go in at his leisure.

Thanks!

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, try +44 (0)1273 64 2905 or
xmpp:kowey at jabber.fr (Jabber or Google Talk only)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100901/640d9229/attachment.pgp>


More information about the darcs-users mailing list