#1474 "with dots" broken with 3D plots

[Karl Ratzsch]

The plot style "with dots" does not create any output with splot (2D plot works). This seems not terminal-dependent, e.g. the example "figure_dots" is empty in both html (png) and pdf output, also nothing appears in the wxt or windows terminal.

plot sin(x) with dots # works

plot sin(x*y) with dots # empty coordinate system

[Ethan Merritt]

Works fine here. You can see the pdf output for the dots figure in the bundled gnuplot.pdf.

[Karl Ratzsch]

Can´t be. The "figure_dots.png" image created in docs/windows/ is just an empty box, also the pdf. Same for the picture in http://www.gnuplot.info/gnuplot_cvs.pdf, or gnuplot.pdf from the zipped rc2 source tree downloadable from sf.net

The pictures created from my 4.6 source tree look OK.

[Ethan Merritt]

Weird.
The copy of gnuplot.pdf that is included in the distribution packages
for 5.0.rc1 and 5.0.rc2 is created in each case by the code in that
same package. The dots figure in -rc1 is OK, but you are correct
that the one in -rc2 is blank. I wonder what went wrong?

... Aha. The 2D code has a separate routine for plot_dots(), while
the 3D code essentially maps this onto "splot ... with points
pointtype 0" . It was broken when the generic draw_3dpoints routine
was modified to distinguish positive/negative point types in order to
support special cases like "pointtype 'A'". This is easy to fix.

There's another strange thing about 3D dots. If you do "set hidden3d;
splot ... with dots", the dots do not show up immediately. But as
soon as you rotate the view they suddenly appear. This oddity has
been around since at least 4.6.1.

[Christoph Bersch]

I can also confirm the bug using CVS version. Another thing I came across with this plotting style is, that the dots don't scale with the linewidth, contrary to what the documentation says:

plot sin(x) with dots linewidth 10

[Ethan Merritt]

The only thing I see in the docs is the statement

"For some terminals (post, pdf) the size of the dot can
be controlled by changing the linewidth."

which is true. Is there another mention?

[Christoph Bersch]

Oh, you are right. I must have skipped half of the sentence... Sorry.