Not a bug, but probably of interest here. For some reason I compared the regression tests between 2.16 and 2.18, up through lyrics-tenor-clef.ly before I got tired. Many look a bit nicer. For most changes I recognized the cause.
Some non-obvious changes: ‘ambitus-with-ligature.ly’ lost barlines, which looks correct ‘chord-tremolo-accidental.ly’ obviously fails. The patch that introduced this feature didn't work well, so we removed it, but David wanted to keep the regression test in place. ‘font-kern.ly’ still broken on Windows, but maybe some progress thanks to help from Pango people ‘fret-diagrams-dots.ly’ ‘fret-diagrams-fingering.ly’ I happen to think the 'skyline-integral' feature of fitting things together based on their outlines, might better be turned off for fret charts. 'lesNereides.ly' collides again. The collision is with a cross-staff slur, which LilyPond has never so far included in collision-avoidance between systems. However, we did have 'horizontal-padding' settings so that the 'm.g.' from the lower system stayed clear of the low notes and 'Ped.' in the upper system. Maybe these went awry. I suspect that the automatic tester skips over full-page regression tests, at least in the image-comparison step, so developers should take a look at these independently, when a change might affect them. Then I quit before test beginning with 'm'. _______________________________________________ bug-lilypond mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-lilypond
