Re: [Groff] Problems with arcs and angles

[Ralph Corderoy]

Hi John,

> > In the groff source, adjust_arc_center() seems the culprit that many
> > of the output devices call to mess with arc's calculations.
>
> Well... right and wrong way aside, I'm simply glad to have achieved
> parity with grops/gropdf.

CSTR 54 documents the DIT format, `Da' included.  Plan 9, Heirloom
troff, groff, should all implement it the same way.  If groff doesn't,
that's a groff bug AFAICS.

I've run

    .PS
    move down
    A: box
    arc at A.n
    arc; arc; arc; arc
    line right
    line from A.n to A.n - 0, A.ht * 2
    line from A.n to A.n + arcrad * 2, (-A.ht - arcrad) * 2
    .PE

through heirloom-doctools 160217-1's pic|troff|dpost, groff 1.22.3-7's
pic, troff, and grops, and then groff with -X for gxditview.  I'll
attach the PNG, but if it doesn't make it through Mailman then see
https://s18.postimg.org/99hxk3ngp/troff-pic-arc-comparison.png

I think the arc should start at the current position, i.e. A.s, be
centred by the "at" at A.n, and be heading down, anti-clockwise.
heirloom matches this.  groff's grops is just wrong. gxditview is close
to heirloom, but with the wrong start position.

Can anyone shed light on why groff deviates from CSTR 54 theory and
historic practice?

-- 
Cheers, Ralph.
https://plus.google.com/+RalphCorderoy

troff-pic-arc-comparison.png
Description: troff-pic-arc-comparison.png