Re: regtests about very small differencies

[Janek Warchoł]

Hi Reinhold & all,

2011/9/12 Reinhold Kainhofer <address@hidden>:
> Am Monday, 12. September 2011, 12:01:25 schrieb Janek Warchoł:
>>
>> I'm going to fix an issue where a note is misplaced by about 0.07
>> staffspace.  I'll add a regtest for this, but how will we make sure
>> that it won't be overlooked in the future?  When we watch a regtest
>> comparison, it shows us the output in a quite low-resolution
>> rasterized form; it will be impossible to spot the difference.  I can
>> also think of some more issues that i'm planning to fix which involve
>> such small changes.
>
> Look at how beam-quanting (input/regression/beam-quant-standard.ly) does it:
> It prints the position of the beams, so that all changes in the position will
> lead to a change in the displayed text, which will be detectable in the
> regtests.

Interesting idea.  However, somehow i don't like it...  I didn't
respond so long because i didn't know how to justify my dislike, but
now i have a nice example.  This solution is of no use when there's a
change in shape rather than position.  See images attached to comment
#2 here: http://code.google.com/p/lilypond/issues/detail?id=1898#c2 -
it's impossible to spot the difference in that case unless the
resolution is really big.

My idea of solving this problem: add information about the resolution
at which a regtest should be compiled (and displayed to the user in
the comparison) to the ly file.  For example add a header field:
resolution = #1000
Only a handful regtests would need those, every other would be
compiled at default resolution.
How do you like it?

cheers,
Janek