| | <<<Back 1 day (to 2019/03/11) | 20190312 |
| mrd0ll4r | aha! | 09:08.56 |
| | I added a reproducer pdf with just ten lines to the git | 09:11.29 |
| | what happens is: when the curve intersects itself, it doesn't paint over itself again, i.e. there's not two strokes of half-transparent gray, but just the one | 09:12.00 |
| | but when it intersects the other curves, it does overpaint them | 09:12.07 |
| kens | mrd0ll4r: I should warn you; I tried two other independent PDF rendering engines yesterday, and all of tehm rendered the lines you pointed out. | 09:12.08 |
| | I believe what you see is correct, but it'll haev to wait until this afternoon to be sure. | 09:12.29 |
| mrd0ll4r | alright alright :D I'm not bug hunting | 09:12.38 |
| kens | Each operation is atomic in PDF, if an object self-intersects it is drawn as opaque throughtout. The result is then blended with the underlygin canvas | 09:13.00 |
| mrd0ll4r | as an idea to fix this: I guess I could split the curves somewhere in the bottom so at the intersection it's different curves, never the same ones with themselves? | 09:13.12 |
| | hmm i see | 09:13.20 |
| kens | Objects can be 'grouped' and then each object in the group is drawn opaque, and teh result of the entire group blended with the underlyign canvas | 09:13.49 |
| | Groups can be nested etc. | 09:13.57 |
| | But I don'tbelieve that the graphcis model blends a self-intersecting object with itself | 09:14.27 |
| mrd0ll4r | I see I see | 09:14.36 |
| kens | Yes I see what you mean from the PDF file | 09:16.26 |
| mrd0ll4r | alright yeah, I just split the curves into their segments, and now it works according to what I expected | 09:16.36 |
| | thanks! | 09:16.38 |
| kens | I'll check with the US engineers this afternoon, this isn't my area really | 09:16.49 |
| | Nice job on r4eproducing it BTW | 09:17.29 |
| mrd0ll4r | :) | 09:29.21 |
| | Forward 1 day (to 2019/03/13)>>> | |
Log of #ghostscript at irc.freenode.net.