Re: [frogs] patch for issue 708

[Carl D. Sorensen]

On 5/26/09 3:55 PM, "Neil Puttock" <address@hidden> wrote:

> 2009/5/24 Carl D. Sorensen <address@hidden>:
> 
>> On a more general note, do you have any suggestions for how to check
>> convert-ly rules?  For code, we have regression tests.  For convert-ly, as
>> far as I know, we have nothing.  Should we be establishing convert-ly
>> regression tests?
> 
> I'm not sure how that would work.  Convert rules are unlikely to break
> unless there are drastic changes between Python versions, so I think
> the best option is to keep each rule as specific as possible and test
> it thoroughly before it's committed.

I guess the question then becomes what "thoroughly" means.  In this
particular example, we had a demonstration that the rule worked on a
particular set of input.

With your experience, you were able to identify a couple of cases where it
wouldn't work.  I obviously missed both of those cases, and hence committed
a broken patch.  I'd like to avoid that in the future.

I guess committing a patch that only partially solves a convert-ly problem
is not completely devastating to the project.  It doesn't break lilypond.
And it improves the convert-ly behavior for some cases.  For the ones it
didn't work for, a new bug report would eventually be filed, and then
hopefully fixed.

Anyway, I'd like to see if there's some systematic way we could avoid having
this problem in the future that doesn't depend on having Neil's eagle eye.

Thanks,

Carl