[Gimp-print-devel] Dimension options

[Robert K. <rl...@al...>]

Just to bring everyone up to speed on what's going on, many versions of OS X choke on custom values
for dimension-valued settings for items other than media size.  However, units of full points turn
out to be a bit too coarse for CD printing, particularly for the hub.  It's possible to have a worst
case error of 1/100"(*), which if you're trying to print to the very edge of the hub, has proven to
be visible.

Matt, Steve, and I have looked at various solutions.  Broadly speaking they are:

1) Punt altogether: we know that that's dissatisfying to some people.

2) Accept that OS X won't print with printers that support CDs.  Nope.

3) Treat dimensions as floats, with the fine adjustment.  That proves to be somewhat complex to
handle with cups-genppdupdate.

4) Use a finer adjustment than a full point.  This increases the number of choices, but lets us
reduce the worst error.

My proposal is #4, using units of 0.5 point.  This reduces the worst case error to 1/200", which I
suspect (read: hope) will prove sufficient.  This can be made compatible in the PPD files by
printing the tags as integers for full integer point values, and as 1_0 for 1.0 (replacing the .,
which is not a valid character in PPD tags, with _).  Patch is attached.  I've tested the PPD files
for conformance, and one printer that uses CDs and one that doesn't.

* 1/100" worst case error is calculated as follows: with precision of 1/72", the worst case error in
the X and Y dimensions is 1/144" (the hub being exactly halfway between point boundaries).  If
you're unlucky and both dimensions are off, the error works out to just under 1/100" by the
Pythagorean theorem.