Dear all,
I've advanced somewhat the DPS extension.
I've implemented swapped dispatch, and I've been able to get DPS
clients from my Solaris/Sparc workstation to display on my Linux/Intel
laptop. This should be convenient for testing.
I've finally found the long-standing rendering bug: it was an overflow
in the fixed_to_int_rnd family of macros (my bug, not Deutsch's).
Dpsclock now displays correctly in all depths. Wonderland runs
correctly (although a little slowly), except that there is a slight
drawing artifact in the second demo (the Bezier curve). Another
rendering bug?
I don't know whether to make a distribution with this code now, or
wait for GhostScript 6.0. Mr Deutsch, any firm dates yet?
The main remaining problems are therefore the lack of time-slicing,
and the single-context limitation.
(By the way: contrary to what I said in the previous message, the
usage of xdps is well documented. For some reason, though, the
GNUstep demos don't run on my extension; they just hang.)
J.
|
