Re: Regtest for issue 5181

[Thomas Morley]

2017-09-23 9:31 GMT+02:00 David Kastrup <address@hidden>:
> Thomas Morley <address@hidden> writes:
>
>> 2017-09-22 12:22 GMT+02:00 David Kastrup <address@hidden>:
>>>
>>> Now things become _really_ embarrassing.  Trying to come up with
>>> improvements on your regtest convinced me that this kind of
>>> syntactical feature is not served well by a graphical test featuring
>>> the typeset results.
>>>
>>> So I stole a bunch of code from the display-lily-tests.ly regtest.

Up to now I've only a vage impression what
'parse-lily-and-compute-lily-string' and #(read-hash-extend ...) does.
I'll currently not dive deeper into it until forced. ;)

>>> I am not convinced this is good either: only outputting stuff that is
>>> less than perfect, and only to the terminal (namely without any
>>> typeset output) is not nice for the printed variant of the regtests
>>> either.

Yeah, that's my main concern as well.

>>> But if there is a better choice, it would probably also
>>> warrant implementing in display-lily-tests.ly .
>>
>> for now only very few notes. I will need some more time to make my
>> mind about it, not in the middle of the night...
>>
>>
>> (define (lily-string->markup str) ...)
>> seems to be unused (same in display-lily-tests.ly) any reason not to
>> delete it?
>
> Oh.  That's the problem with copy&paste jobs.  I'll take a look.
>
>> I'd like to see an example for LyricText/Hyphen. Like:
>>
>> lyr = \lyricmode { foo }
>> \test "limitation" ##[ \lyr -- #]
>
> More like \test ##[ \lyricmode { \lyr -- } #] I guess?

Ofcourse. Though, this does not cause any printed output (like the
other test-cases without settings for the optional 'harmless').
Wouldn't it make sense to print something for fully working cases as well?

Doing
\test "equal" ##[ \lyricmode { \lyr -- } #]
returns:
warning: Test equal: unexpected.
[...]

Throwing a warning, because something works is a little strange, though.
Why not put out something at the lines of
(ly:message (*location*) "Test equal: already working. <formated-args>")
?

>
>> This wish is triggered by an ongoing thread at the german forum:
>> https://lilypondforum.de/index.php/topic,150.msg982.html#msg982
>
> Perfectly reasonable, and it does show that the changed test framework
> not bothering to typeset the actual examples makes it easy to add
> relevant test cases without wracking one's brain.

Indeed.

Cheers,
  Harm