Re: tltk backend printing

[Ben Abbott]

On Jan 19, 2010, at 2:39 AM, Michael D Godfrey wrote:

> Using the current hg source I noticed 2 items about printing:
> 
> 1. Rendering and print() are not synchronized.  If you type print()
>     before rendering of the screen plot is complete, it fails.  I am
>     not sure if this happened with gnuplot, but it is more obvious
>     now.  It would make sense to put a refresh in the print command
>     so that it is not necessary to precede each print() with a refresh.
>     This obviously usually happens in functions or scripts.  In addition,
>     something odd seems to happen if you plot a number of plots in a
>     script following each plot with a refresh() and print().  At about plot 4 
>     or 5 the refresh stops working in the sense that the plot window 
>     stays blank (typically black), but the print() produces the correct plot.
> 
> 2. When printing a fairly large number of plots using just
>     print("file.ps") --no color or other options -- occasionally the plot is 
> in color.
>     So far, it has always been that the plotted curve is in blue.
>     Most of  the plots are B&W as documented.  This must be some uninitialized
>     variable.
> 
> I will see what I can do about both of these, unless some one knows
> exactly how to fix them.
> 
> Michael

Regarding (1), there is already a "drawnow()" early in the print process, so 
I'm confused why there is a problem.

Ignoring that, I'd guess that the current approach of modifying some of the 
figure's properties during the process of printing is interfering with 
rendering the figure (and vice versa). An easy solution would be to make an 
invisible copy of the figure being printed and print that instead.

Ben