[darcs-devel] cutting down the test suite

Ganesh Sittampalam ganesh at earth.li
Tue Nov 5 21:12:32 UTC 2013


I've looked into this a bit further:

(1) There are a few tests in the suite that explicitly pass --myers. I
can't see any reason for this as they still pass after removing the flag.

(2) I don't see much value in running any Myers tests as part of the
standard suite, so I plan to just remove it from the default options.

(3) There are some tests that are darcs-1 or darcs-2 specific or are
worth running for both. Here I am inclined to restrict the default to
darcs-2, but introduce the concept of a "pragma" to a test, i.e. some
metadata at the top of the test that controls default configurations it
gets run in.

This is probably quite simple to implement because there's a place in
the code that currently makes an explicit cross-product of the tests to
run, and is in IO - so it could read the file too.

Along with this I would add extra command-line options to allow users to
select more configurations to run.

Shout if there are any objections.

On 31/10/2013 08:06, Eric Kow wrote:
> Sounds like a sensible thing to do.
> Maybe humans should run a limited suite, and bots should run the whole shebang
> 
> 
> On 31 October 2013 07:50, Ganesh Sittampalam <ganesh at earth.li> wrote:
>> Hi,
>>
>> Currently the test suite runs 4 times by default - Darcs1/Darcs2 and
>> Myers/Patience diff.
>>
>> I would like to cut this dow to only be Darcs2 and Patience to reflect
>> our current defaults. Any objections?
>>
>> Any nominations for specific tests that should be run as Darcs1 or Myers?
>>
>> I imagine some of the record tests would be good candidates for Myers.
>>
>> There are a few tests that are Darcs1 specific (grep "darcs-1" tests/*)
>> that certainly should be run, but I'm not sure if there are any other
>> obvious candidates.
>>
>> Cheers,
>>
>> Ganesh
>> _______________________________________________
>> darcs-devel mailing list
>> darcs-devel at darcs.net
>> http://lists.osuosl.org/mailman/listinfo/darcs-devel
> 
> 
> 



More information about the darcs-devel mailing list