Clipping of subpixel anti-aliased text

[Theo Veenker]

Hi Ben,

You could check if cairo_text_path() followed by cairo_fill() yields a 
different result? Probably not. I would test cairo_in_clip() on all 
pixels for every loop iteration. Perhaps that shows the clip being 
unexpectedly reset? Not sure.

Theo

-- 
Theo Veenker  |  Behavioral Experiment Systems
theo.veenker at beexy.nl  |  www.beexy.nl

On 14/04/2025 17:13, Ben Harris wrote:
> I have a question that essentially boils down to: if I set the clipping 
> region to a pixel-aligned rectangle, should I expect pixels outside that 
> rectangle to be unaffected by drawing operations?
> 
> The specific example that leads me to ask this is demonstrated by the 
> attached program.  It creates a 20 × 20 pixel image surface sets the 
> clipping region to precisely the left half of the surface, and then 
> repeatedly shows a ']' character and then erases it using cairo_paint().
> 
> My hope was that this would leave the whole image black because the last 
> operation is to paint the entire clip region.  But on my test systems 
> (Debian bookworm and trixie, Cairo 1.16.0 and 1.18.4 respectively) I get 
> a column of red pixels just outside the clipping rectangle.
> 
> What is the expected behaviour here?  I can see that for an arbitrary 
> clipping region it's not possible to clip drawing operations in a way 
> that can be perfectly undone.  But in the special case of a pixel- 
> aligned rectangle it should be possible.  Or do I need to use 
> cairo_surface_create_for_rectangle() to get perfect clipping?
> 
> Aside: the context where I found this problem was the GTK version of the 
> PuTTY terminal emulator, which needs to be able to reliably update only 
> part of the terminal screen.
>