Hi, Have prepared a page of tables and out of habit used -Tps piped into ps2pdf(1), I realised I should switch to gropdf(1) but instantly noticed the output was a lot more ugly; enough to force me back.
$ cat bug #! /bin/sh cat >line.tbl <<\E .TS l. _ foo = .TE E tbl <line.tbl >line.tr groff -dpaper=a4 -Tps line.tr >line-grops.ps ps2pdf line-grops.ps groff -dpaper=a4 -Tpdf line.tr >line-gropdf.pdf qpdf -qdf line-grops.pdf line-grops.qdf qpdf -qdf line-gropdf.pdf line-gropdf.qdf $ $ ./bug $ The horizontal lines in the table are much thicker with gropdf(1), so much so they dominate instead of the text they demarcate. I think tbl(1) draws them with the `\D' escape, that doesn't give a thickness. $ grep '\\D' line.tr \v'.25m'\s[\n[3lps]]\h'|\n[3cd0]u'\D'l |\n[3cd1]u 0'\s0 \v'.25m'\s[\n[3lps]]\v'-2p'\D'l |\n[TW]u 0'\v'2p'\h'|0'\D'l |\n[TW]u 0'\s0 $ line-grops.ps has .4 LW % setlinewidth 85.33 4.5 72 4.5 DL % Draw line. /F0 10/Times-Roman@0 SF % Set font. (foo)72 14 Q % Show text. 85.33 18.5 72 18.5 DL 85.33 20.5 72 20.5 DL 0 Cg % setgray EP % End page. that ps2pdf turns into, as shown by qpdf(1), with my guesswork comments. q 0.1 0 0 0.1 0 0 cm % Matrix, 10th scale? /R7 gs 4 w % Line width, equivalent to 0.4. 1 J 1 j 0 G 720.25 8374.25 m % Move. 853.25 8374.25 l % Line. S % Stroke. 0 g q 10 0 0 10 0 0 cm % Matrix, scale back up to 1. BT % Begin text. /R8 10 Tf % Set font. 1 0 0 1 72 827.89 Tm % Matrix, move. (foo)Tj % Show text. ET % End text. Q 720.25 8234.25 m % Move. 853.25 8234.25 l % Line. S % Stroke. 720.25 8214.25 m % Move. 853.25 8214.25 l % Line. S % Stroke. Q gropdf's qdf next, I'll comment significant differences. q 1 0 0 1 0 0 cm % No 10th scale. 1 J % No setting line width, default 1? 1 j 0 g q BT % Begin text, two moves, but no text. 1 0 0 1 72 840 Tm 0 Tc 1 0 0 1 72 837.5 Tm 0 Tc 0 g ET Q 0 G 72 837.5 m % Coordinates differ, see below. 85.33 837.5 l s % Stroke, but `s' v. `S'. q BT 1 0 0 1 85.33 837.5 Tm % Redundant move? 0 Tc /F5 10 Tf 1 0 0 1 72 828 Tm 0 Tc 0 Tw (foo) Tj 1 0 0 1 72 823.5 Tm 0 Tc ET Q 72 823.5 m 85.33 823.5 l s q BT % Empty `Begin text'? 1 0 0 1 72 823.5 Tm /F5 10 Tf 0 Tc 1 0 0 1 72 821.5 Tm 0 Tc ET Q 72 821.5 m 85.33 821.5 l s q BT % Empty `Begin text'? 1 0 0 1 85.33 821.5 Tm /F5 10 Tf 0 Tc 1 0 0 1 85.33 0.11099999999999 Tm 0 Tc ET Q Q Here's the grops coordinates of the three lines, multiplied up by ten. The y-coordinates don't change for horizontal lines. 72.025 837.425 85.325 +0 '' 823.425 '' +0 '' 821.425 '' +0 And gropdf's. 72 837.5 85.33 +0 '' 823.5 '' +0 '' 821.5 '' +0 The difference, pdf-ps, -0.025 0.075 0.005 I think grops should copy grops's choice of default line width. And the differences in coordinates seem odd. Even if they're correctly compensating for line-cap differences, should those differences exist? -- Cheers, Ralph. https://plus.google.com/+RalphCorderoy