Odd regtest behavior

[address@hidden]

Hey all,

When I try to run the regtests on my patch @ http://codereview.appspot.com/4290069/, I get a really weird result - several of the files in input/regression look substantially different in the PNGs than they do when I compile them normally.  The PNGs look bad, whereas normal compilation yields the expected results.

Could someone please run the regtests on this patch and then compile from the command line some of the following files:

beam-collision-opposite-stem.ly

beam-collision-beamcount.ly

slur-scoring.ly

beam-collision-basic.ly

beam-extreme.ly

beam-collision-prefatory-matter.ly

beam-collision-classic.ly

beam-collision-feasible-region.ly

collision-dots-move.ly

grace-start.ly

I'd like to know if you get divergent results as well.

Cheers,

Mike