On Fri, Aug 07, 2009 at 01:57:41AM -0400, Gwern Branwen wrote:
> What is there to follow? There's no need to call bash for a trivial
> executable invocation - I'm not even using conditionals or anything
> reasonably fancy.

It wouldn't surprise me if, for example, there was no sh(1) in some
Cygwin environment.  Basing even a trivial script on the template will
improve homogeneity between scripts, avoid inexplicable failures
resulting from that kind of weirdness.

The template also encourages authors to explain what the test is
testing, and cite relevant BTS tickets for further reading.

>> Do the docs tell testers to install hlint ≥ <version>?
>
> No; what docs would I edit? I grepped in docs/ and tests/ for
> 'haskell_policy' and found nothing.

Unfortunately, I'm not sure.  Perhaps HACKING and/or the wiki;
wherever it tells developers they should run "cabal test".

>>>  we don't edit haskell_policy.sh directly because if there's a single
>>>  warning or error it errors out and hlint generates>1k items
>>
>> I'm not sure what you mean by this.  How about redirecting hlint's
>> output to a file?  This way the output within the test harness is
>> minimal, but users can see the exact error by consulting the report
>> file.
>>
>> hlint --hint=darcs-custom.hint ../src --report hlint.html>/dev/null ||
>> {
>>   echo>&2 "Errors found by hlint!  See $(dirname "$0")/hlint.html for 
>> details."
>>   exit 1
>> }
>
> All the warnings & errors should be fixed, or we ought to discuss
> why a class of warnings & errors is unreliable and ought to be
> disabled. Hiding them up is not a good idea, I [...] think.

Good point.

>> Incidentally, looking only at ../src/ causes Setup.lhs and
>> Distribution/*.lhs to be skipped.  Any reason just ../ wouldn't
>> work?
>
> No. Should those files be looked at?

I think they should (after all, hlint reports problems with them), but
I'm willing to be convinced otherwise.
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to