[m-rev.] for review: Fix failing tabled_typeclass test case.

Zoltan Somogyi zoltan.somogyi at runbox.com
Tue Aug 7 18:46:26 AEST 2018



On Tue, 7 Aug 2018 17:59:40 +1000, Peter Wang <novalazy at gmail.com> wrote:
> > One more general point: for test cases where there are multiple expected
> > outputs I think there needs to be more specific documentation about what
> > each expected output corresponds.  While it can sometimes be found
> > in the log messsages, occasionally it's a bit opaque.  My suggestion
> > is that the header comment of each test with multiple expected outputs
> > should list what they all are.  Opinions?
> 
> I can't think of a simpler solution.

Agreed. We should then require any diff that changes any of the .exp files
of a test to modify all the others as well, if necessary.

> We should be able to modify the test scripts to write out, for each
> passing test, which .exp file was matched in that test grade.

That is also a very good idea, though I believe that in some cases,
the differences between expected output files depend not on grade,
but on things such as which optimizations are applied.

One minor modification: I would modify the test script to write out
the name of the matched .exp file *only* if there is more than one.

Zoltan.


More information about the reviews mailing list