Re: regression: point-and-click no longer includes absolute path

[David Kastrup]

Mark Knoop <address@hidden> writes:

>>> The point-and-click link in the resulting pdf is relative:
>>> textedit://testpaths.ly:9:2:3 - and therefore doesn't work.
>>
>>Hm, the PDF you attached *does* give an absolute link, but:
>
> Yes, sorry, testing with different versions and attached the wrong one.
> See attached pdf now.
>
>>Indeed, this is right, so obviously my patch *did* change the
>>behaviour of point-and-click. (I have to repeat that nobody noticed
>>that during review).
>
> It could be that most point-and-click users are also users of
> Frescobaldi or other front-ends that pass the full pathname to
> lilypond.

I would think that often potential reviewers entertain the naive notion
that a patch submitter has tested the most basic functionality of his
patches as stated in the issue description.

So the reviewers do not even try what is declared as the basic idea of
the patch (that is, if they try anything at all).

But realistically, we don't get a whole lot of pre-acceptance testing
anyway.  So it's always a good idea to check out stuff oneself,
optimally also submitting regtests that are able to certify the change
in behavior.

Yes, this happens less often than desirable.  But "I have to repeat that
nobody noticed that during review" is a bit silly, given how few of our
developers indulge in full-scale manual reviews and testing.  Our
automated tests (performed and evaluated using manual labour, though)
_do_ catch quite a few things.  But if they are supposed to demonstrate
the validity of a change, there must be a regtest created for the
change.

-- 
David Kastrup