DSP dances

[Ineiev]

Diagonal squared pads approaching to square (that is those whose width is much more than the distance between their endpoints, as in 1206.fp) are shown in a delightful way. this is because round-off errors are multiplied when drawing the pad as a line. this can be seen when changing zoom in gtk gui (I think in lesstif, too). I dare not bother you with my explanations why this effect is not considerable for pads with rounded ends.

Gerber output is not affected, because such lines are drawn as polygons and the coordinates are rounded after computing the polygon vertices. PS output is not affected, because the coordinates are not rounded at all.

In the same piece of code (draw.c) there is another gorgeous thing. it is how the diagonal pads are drawn in thinline mode. square pads are shown as not diagonal; and it is almost inexpressible how rounded pads are drawn.

The patch applied eradicates those things. squared diagonal pads are drawn as polygons; thinline code for rouded pads is adjusted, too.

I pray dear developers don't fix this. PCB will not be PCB without it. your humble servant trusts your wisdom.

[Ineiev (ineiev)]

Alas! somebody fixed thinline draw. fortunately, "solid line" squared diagonal pads dance as they did.

dspdances.bis.patch fixes the rest. I also changed thinline mode drawing for squared pads because the patches' get_pad_polygon() should be faster (single sqrt() vs atan2()+sin()+cos()).

[Ineiev (ineiev)]

I'm sorry. of course, there is no need to check if the pad is squared in the last set_line_cap() call.

dspdances.ter.patch attached.

[Ineiev (ineiev)]

updated to current CVS state;
the speed improved in some cases

[Ineiev (ineiev)]

updated; fixed mask drawing

[Traumflug (mah-jump-ing)]

Hmpf. Launchpad swallowed my text again.

Short version:

- Tried the provided layout and couldn't find anything unusual.

- Tried to apply the patch, but it's (no surprise) outdated beyond recognition.

[Traumflug (mah-jump-ing)]

Got a private mail:

> Awesome! people still recall me.
>
> I attach a rebased patch.
>
> Both lesstif (--with-gui=lesstif) and plain gtk (--with-gui=gtk --disable-gl)
> are still affected. gtk-gl is not affected because it feeds its backend
> with floats rather than ints (not an option for lesstif and gtk).

[Traumflug (mah-jump-ing)]

It took me a while to see what's going on. When one zooms in closely, the square pad suddenly starts to rotate. Is this the bug we're talking about?

[Traumflug (mah-jump-ing)]

The new patch applied flawlessly and this unexpected rotation is gone. Excellent! Pushed directly to master as the fix is very apparent. If there's something missing, please tell us.

[Traumflug (mah-jump-ing)]

No support for admins actively blasting developers away.

[Traumflug (mah-jump-ing)]

Please don't assign other people than yourself to a bug.