Have you used some means other than PLplot of judging whether the
FreeSerif.ttf font was misaligned?

The reason I ask is there are some rather complicated transformation
formulas that are used to align TrueType fonts in PLplot since our
coordinate system has a different origin than most fonts, and there might be
some deficiency in the PLplot code that is misaligning the results rather
than some deficiency in FreeSerif.ttf itself.

Also, my understanding is there is several generations of FreeSerif.ttf. Is
it possible you are correcting misalignments in an old version that have
already been corrected in later versions?

I was most impressed you went to such lengths to deal with the misalignment
issue, but I thought the above fundamental questions still needed to be
asked. In particular, if you really have found that FreeSerif.ttf was
misaligned by means that are independent of PLplot, that is a result that
greatly relieves my mind since I always assumed before that the misalignment
problem was caused by something we were doing wrong. If it turns
out you have truly found misalignment, then please be sure to feed your
results back to the FreeFont team so that everybody will benefit.

I think the question is what you mean by "misalignment". I don't think that
the FreeSerif font is really misaligned, in the sense that, these glyphs have
no reason to be centered for a "normal" use of a font.
It is is clear, when you edit the font, that the symbols do not lye in the center
of their cell.
It is also clear, after manipulating the font with FontForge,
that "centered" in the sense of plplot means that, in the sense of the font,
the symbol have to be horizontally centered in the cell, and vertically,
must be positionned around the first third of the cell (the same position for
all the symbol of plplot).
So, in my understanding, if there is a deficiency in plplot code, it's no more
than a small vertical bias in the position (to be confirmed).
But I'm really not an expert of font things, and I even hardly know the way that
the glyph are positionned (there is some referentials, like "baseline"
but it remains quite obscure for me).
Anyway, I'll ask the FreeFont team about this "misalignment" things.
To Kay Wiik: obisously, I will send you our version of FreeSerif_plplot.ttf
but it is a raw version that we have not fully validated: take it with care!

Thanks very much for drawing our attention to this bug which I have just
confirmed for one of my private examples that uses -dev png. I hope we will

be able to fix it soon (in the svn version), but no promises.

Ok, it will be great! Knowing this bug, it is still quite easy to filter the data
before using plpoin.