Re: Gnuplot win terminal regression

[Bastian M. <bma...@we...>]

I found out what's happening:

In case of "lt palette z" and "lc rgb variable" there's a
polygon with the same start and endpoint. Windows optimizes this away 
and the drawing position is not updated. The next call to W_vect
therefore uses the wrong starting point.

Possible fixes:
a) a MoveTo() at every W_move call
b) a MoveTo() after all Polyline calls
c) filter out such polygons and replace by calls to MoveTo()


I'd prefer option b) because it is straightforward and does not 
introduce many additional GDI calls:

	if ((lastop==W_vect) && (curptr->op!=W_vect)) {
	    if (polyi >= 2) {
		Polyline(hdc, ppt, polyi);
		MoveTo(hdc, ppt[polyi-1].x, ppt[polyi-1].y);  /* make sure we move the 
drawing position in case of empty polygons */
	    } else if (polyi == 1)
		LineTo(hdc, ppt[0].x, ppt[0].y);
	    polyi = 0;
	}



   Bastian


Ethan A Merritt schrieb:
> Bastian:
> 
> Back in June I applied a patch from you that had the following ChangeLog
> entry:
> 
> 2008-06-24  Bastian Maerkisch  <bma...@we...>
>         * src/win/wgraph.c (drawgraph): Don't forget to draw a polyline even
>         if it has only a single segment.
> 
> There is now a report on the newsgroup that for data files containing 
> multiple polyline definitions (e.g. blank lines in the file), each supposedly
> independent polyline is connected to the previous one.  From inspection of
> the code, I conclude this is due to your change above.  That is, the initial
> move command sets polyi=1, but your change causes this to be treated as a
> length 1 polyline rather than a move.   Strangely, this happens only for
> plots with either "... lc rgb variable" or "... lt palette z".
> I don't fully understand why the linetype makes a difference.
> 
> Could you have another look at this issue please?  Perhaps the previous
> patch should be reverted, or perhaps an additional fix is needed somewhere
> in the W_pm3d_setcolor code block in wgraph.c.
> 
> I'd like to release an incremental version 4.2.4 sometime soon, and this
> is the only regression I know of that would hold it up.
> 
> 	Ethan
>