On Oct 1, 2011, at 7:22 PM, Neil Puttock wrote: > On 1 October 2011 18:19, Peekay Ex <pkx1...@gmail.com> wrote: > >> Neil what command do you run when you do a test-baseline so that I >> might get something like you do? > > I don't. :) I couldn't work out which file was failing, so I did what > I usually do: run all the regression tests until it crashes. > > The output I've posted is from debugging mozart-hrn-3.ly on its own via GDB. > > Cheers, > Neil >
Scheiße. I still can't reproduce the problem on my VBox with the unoptimized binary, so all of my suggestions to try to help make this go away will be from thought experiments. It seems as if robust_scm2interval is getting the interval from the property lookup instead of using the default Interval (0.0,0.0), as otherwise this crash wouldn't happen because there would be two valid values. This means that in memory-land, somehow, somewhere, the interval that is accessed for "positions" is somehow being garbled such that Interval::center can't do its thing. A one-line solution would just be to check in calc_y_offset that the interval actually has values before calling center, but this would be painting over rust (as Reinhold would say). Could someone revert the tuplet patch and see if it compiles clean? Again, sorry I can't be of much help, but I can't reproduce the error :( Cheers, MS _______________________________________________ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel