Re: No test changes required?

Norm Tovey-Walsh writes:

[Case by case discussion elided.  You may be right.]

> I don’t think there are any tests that say “parse this input against the
> current version of the ixml grammar”. I’m not sure how you’d write such
> a test.

The catalog syntax/catalog-as-grammar-tests.xml shows one way.

At least, as documented the meaning of a grammar test with ixml input
is: parse the supplied grammar using the grammar built in to your
processor and check it for conformance.  When created, the test results
were intended to reflect the then current grammar, and I think that
should be maintained as an invariant.

(There may be other ways; the relative reference in
syntax/catalog-as-instance-tests.xml will always use whatever is in
tests/reference/ixml.ixml, which I have been assuming should be kept
current.)

But everything in syntax/*.ixml illustrates errors in grammars.

If we don't have a test whose effective meaning is "parse the current
ixml.ixml against the current ixml.ixml", then we should.  

> I think what really needs to be done is that the input grammars
> tests/ixml/ixml.ixml and tests/refrence/ixml.ixml need to be updated.
> That *will* break a bunch of tests. Until then, the tests are still
> testing the old grammar.

Any test that provides an explicit grammar, yes.  Grammar tests should
always be using the current grammar. 

> Given the pending change to the namefollower rule and issue #57, I’m not
> inclined to do that until after we’ve rev’d the grammar again.

I don't have a strong view on timing.

-- 
C. M. Sperberg-McQueen
Black Mesa Technologies LLC
http://blackmesatech.com

Received on Thursday, 24 March 2022 13:56:42 UTC