Hans Hagen
If test.mp draws a square or rectangle instead of an ellipse, then there's no difference (both versions of test.tex work).
looks like a bbox error
I didn't think so, because after mptopdf got hold of the figure, the whitespace around the figure was right and mptopdf should have used the bbox in the figure. So my only guess was that the internal conversion to pdf was getting confused somewhere.
If test.mp draws a square or rectangle instead of an ellipse, then there's no difference (both versions of test.tex work). that's indeed weird, any hypothesis to start from?
My hypothesis was refuted by testing the rectangle. I had hoped (if that's the right word) that the internal conversion was using the width where it should have used the height, which is why the ellipse would cause problems. Which is why I tried the square, and it was fine -- more evidence for the theory. But the rectangle was fine, so my theory died.
in core-fig.tex patch: \def\dogetfiguresizeeps {\dogetEPSboundingbox\wantedfigurefullname\!!widtha\!!heighta\!!widthb\!!heightb \xdef\analyzedfigurewidth {\the\!!widthb}% \xdef\analyzedfigureheight{\the\!!heightb}} somehow i assumed that i was dealing with bbox values
That fixes it here, thanks. I see what happened with my test cases now. The ellipse had its lower left corner away from the origin, so the old analyzedfigureheight was larger than it should have been. But the rectangle and square start at the origin, so you don't see the effect.
btw, a convenient way to visualize such thing sis: \placefigure[force,none]{}{\externalfigure[test.1][frame=on]} \placefigure[force,none]{}{\externalfigure[test-1.pdf][frame=on]}
Good to know. Plus I was wondering how to do that anyway. -Sanjoy `A society of sheep must in time beget a government of wolves.' -- Bertrand de Jouvenal