[cairo] clipping question

[Bill Spitzak]

Fabien Costantini wrote:
> But using cairo_save/restore api would save the _whole_ context,
> and I only need to deal with the clipping context.
> Furthermore, restoring other undesired context attributes would have side effects on the drawings,
> so I'm afraid it's not usable in my situation.

Actually I think it is ok if pop-clip and pop-transform has undefined 
results on the rest of the graphics state. Otherwise it will be quite 
impossible to use a gsave/grestore api like Cairo for fltk.

The biggest problem we have is the lack of "hairline" mode in Cairo. We 
absolutly need hairlines to draw widgets that don't look like crap. 
Currently we are not using Cairo's transforms at all, doing it all 
ourselves before sending the coordinates. In theory we cannot use 
Cairo's path-construction calls either, as we don't know if they are 
going to fill or stroke until after the path is constructed.

I'm mostly of the opinion that we should use Cairo as much as possible, 
with as little as possible emulation code, but hairline really is 
causing problems.

So again, I really think hairline mode, triggered by 0-width lines (to 
match OpenGL and DirectX and X and every other graphics API I have ever 
seen) is very useful and in high demand by programmers. I think the 
result can be well-defined: horizontal and vertical lines are an integer 
number of pixels thick and positioned such that they hide the 
antialiasing of the same shape being filled, and below/right in cases 
where there is no antialiasing. All other lines at any angle are 
positioned to join the horizontal/vertical ones and are antialiased.

Emulating this in current Cairo is a pain, and the fact that we quickly 
found it necessary is an indication that lots of other users are going 
to find it necessary as well.