Regression?: Melismata disturb note head distances in 2.13.x

[Alexander Kobel]

Hi, all,


I think I stumbled across a regression regarding note head distances 
with melismata in 2.13.24, compared to 2.12.3, which is not yet in the 
tracker.  Anybody to confirm and/or add to the tracker?  (PNGs attached, 
the PDFs are too large for the -bug list.  64KB is darn small these days...)
A workaround can be seen in the snippet, but it requires quite a bit 
more writing when you actually want to use the induced melismata by 
manual beaming.

I also tried to find the exact commit causing the regression.  Looks 
like it's some commit for 2.13.5: hit 
<http://lilypond.org/test/v2.13.5-0/compare-v2.13.3-9/index.html> and 
scroll down for an output distance of 10.998043 for the regtest 
lyric-ignore-melisma-alignment.ly, or 10.875328 for 
lyrics-melisma-beam.ly.  You can see there that the spacing between the 
quavers gets uneven, probably caused by the "Whee!" syllable.
I looked at the git log (only commit descriptions, not the code, yet), 
and could find that there were some changes to the automatic and manual 
beam behaviour going on, mainly for the beam direction shortcuts in 
manual beams.  I'm not sure if this is the correct place to investigate 
this, though, because spacing also changes with ties and manual 
melismata, not just manual beams, as shown in the attached snippet.

On a side note, we seem to be short of a regtest for melismata where 
syllables have a relevant length to check those issues.  The additional 
space in both snippets is visible, but not quite too present for a 
regtest, and Ky -- ri -- e might not be the best text in 
lyric-melisma-manual.ly.


Cheers,
Alexander

melismata.ly
Description: Text Data

melismata-2.12.3.png
Description: PNG image

melismata-2.13.24.png
Description: PNG image