🚀 go-pugleaf

I cannot believe that this is a bug, but I've tried it manually and it
goes wrong, and the more I look at the code, the more it seems wrong.
The flex routines mentioned in the Black Book creates the OtherSubrs
that ends up in a font's Private dictionary. OtherSubrs #1 goes like
this (as defined on page 95, I hope they don't mind me reproducing
this tiny bit, but I don't see how I can explain the problem unless I
do...):

currentpoint grestore gsave currentpoint newpath moveto

As shewn on page 77, it is used at the onset of a piece of flex to
generate the the points on the stack that FlxProc will use to
determine whether or not the segment will be curved or rendered as a
straight line.  Now the act of doing a

currentpoint grestore

after the previous rmoveto will place the first point on the stack,
and then destroy the new point, so when we execute

gsave currentpoint

*this* currentpoint will in fact be the *original* point, so that all
the rmoveto's will work relative to the *original* point and *not* to
eachother.  If this code becomes

currentpoint currentpoint grestore gsave newpath moveto

it seems to work fine.  So what, if anything, is the matter?  I
created a Type One font to test both versions, and the weird thing is
that they *both* seem to work fine, but only during a show or
charpath; if I bung the code in manually, only my version works.  My
guess is that there some sneaky `special case' being used by the
type-one-engine (which seems to live in a world of its own) such that
it remembers the last point independently of gsaves and grestores.

Any ideas???