GSoC idea: Frescobaldi extension
Now that GNU has been accepted for GSoC I've pushed to staging
Remove LO extension from GSoC ideas list
I've pushed this and set the "Fixed" label, but I'm not sure if I should have also changed the "Patch" to "Push".
Move Guile-style modules from scm to scm-modules
That's strsnge because I had pulled master before. However after discussion I wanted to withdraw the patch anyway.
\relative { \tuplet 3/2 2 { c'4 d8 e f4 } g2 } shows what I mean: The middle triplet is broken (not beamed), obviously because LilyPond does not consider the group as a "virtual" 3/4 measure with a 2 2 2 grouping, but as 2/4 of a 4/4 measure where the possible break point is at the "real" beat.
Attached is the work I have spent analyzing the issue. I want to stress that this is also related to wrong handling of tuplets, whose fix is sort of a prerequisite to get beam subdivisions right.