I've been having similar problems, but I have some trouble recreating
them now. I'm attaching three files to demonstrate: first, an eps file
created with matplotlib on about December 7th, with a version that
probably was nearly current at that time; second, the result of
epstopdf on that file; third, the result of printing this pdf to a ps
file in Adobe Reader 7 on Linux (exact version is "7.0.1 07/27/2005").
Note that the bounding box in the third file is wrong, so you have to
select e.g. A4 paper in gv to see anything.
The pdf file displays in Adobe Reader just fine, but when printed, the
text is replaced by various characters. There seems to be method in
the madness in that zeros are replaced by #s, decimal points by $s,
etc., so I am suspecting some kind of an encoding problem. Xpdf seems
to produce a completely good ps file from the pdf, so this could be a
bug in Adobe Reader and not matplotlib. It would be nice to find a
workaround anyway, since Adobe products are so widely used for
handling pdf files.
The problem I'm having in recreating this on a current matplotlib is
that gv chokes on the output of Adobe Reader. It draws the plot but
then shows an empty stack error instead of drawing the text.
teht3b-2.eps.gz (51.4 KB)
teht3b-2.pdf (7.9 KB)
teht3b-2.ps.gz (17.2 KB)