Fix the clippath renderering so that it uses no-clip unsigned chars#1846
Conversation
|
Note to self: before merging this, it should have a test. |
|
Wow, this is low level. Nice spot.
Agreed. Whilst I agree it's a bug fix, I also wonder if we are better off sticking this in v1.3 to avoid any potential negative impacts this change could have? |
|
I think I'm leaning toward putting this in 1.2.x -- while it may change the appearance of some plots, it should do what the user means and give lines that are the requested thickness. Since none of the tests failed or required tweaking, I'm even more on the side of calling this a simple bugfix. Any tie-breaking points of view? 😉 |
|
1.2 sounds good to me. Before you push the green button to release the magical GitHub ponies, is there still going to be a test? |
|
Yes -- a test shall appear shortly. |
|
@mdboom Thanks :) |
|
Travis failure is false. |
Fix the clippath renderering so that it uses no-clip unsigned chars
This was discovered during the discussion in the e-mail thread "removing paths inside polygon".
When paths are affected by a clip path, the scanline type being used for rendering was wrong. It was
grey8, but should beno_clip_grey8. This resulted in horizontal lines looking thicker than they should.This also fixes a related performance bug where the buffer for the alpha mask was over-allocated by a factor of 4.