| | <<<Back 1 day (to 2017/05/22) | 20170523 |
| ray_laptop | mvrhel_laptop: I need to chat about pdf14 transparency (as it relates to tags). | 00:06.58 |
| | mvrhel_laptop: I'll send an email... | 00:13.27 |
| mvrhel_laptop | Great thanks! | 00:13.36 |
| ray_laptop | mvrhel (for the logs) email sent. cc'ed tech. This looks messy :-( | 00:46.06 |
| | like the "art_composite" function may need to tell us which of the three conditions contributed to the pixel color (source, background or both) as it does all the components | 00:47.26 |
| | mvrhel_laptop: OK? | 17:00.32 |
| mvrhel_laptop | ray_laptop: I don't see another email | 17:00.51 |
| ray_laptop | just sent -- Subject Tags and transparency. | 17:01.25 |
| | mvrhel_laptop: let's go ahead and chat about Blend modes and tags (the email was other problems, mostly). | 17:02.22 |
| | mvrhel_laptop: AIUI, if src_alpha == 0, then we shouldn't change the tag, no matter what BM, right? | 17:02.57 |
| mvrhel_laptop | ray_laptop: I don't know. I will need to review the spec. But I would suspect that to be the case | 17:03.27 |
| | let me open it up | 17:04.05 |
| | yes. if alpha_s is 0 then we just have the backdrop | 17:05.37 |
| ray_laptop | Some Blend modes should ALWAYS "OR": Multiply, Screen, Overlay, HardLight, SoftLight, Difference, and Exclusion | 17:05.49 |
| | but Darken, Lighten, ColorDodge and ColorBurn depend on the component values as to whether the source is used (I think) | 17:07.32 |
| mvrhel_laptop | ok. so the question now is, if alpha_s = 1 what is the backdrop contribution. | 17:08.20 |
| | we just need to go through each blend mode and figure it out | 17:08.40 |
| | do you want to do it or do you want me to do it ray_laptop? | 17:08.54 |
| | The difficulty here, is that you could have the following situation | 17:09.48 |
| ray_laptop | mvrhel_laptop: well, if alpha_b is 0, I suppose the only contribution is from the src, right? | 17:10.31 |
| mvrhel_laptop | well yes | 17:10.41 |
| ray_laptop | ok, so alpha_s == 0 and alpha_b == 0 are the simple cases :-) | 17:11.19 |
| mvrhel_laptop | So looking this over, I am now thinking we always want to or with the non-normal blend mode | 17:12.02 |
| | for example, if I had an image and was blending with Lighten | 17:12.23 |
| | I would have pixels that were image and some that were not within the image | 17:12.38 |
| | that would be weird | 17:12.44 |
| ray_laptop | mvrhel_laptop: so the question is, should the tag be based on whether or not the destination color changes, or just if it involved the source? | 17:12.45 |
| | mvrhel_laptop: so, we always OR in non-Normal (unless the alpha_s or alpha_b == 0) ? | 17:13.55 |
| mvrhel_laptop | I would divide this into Normal and not normal. And then normal would be divided into alpha_s values | 17:13.56 |
| | ray_laptop: oh right | 17:14.45 |
| | always or in non-normal except for the a_s = 0 or a_b = 0. | 17:15.43 |
| | wait isnt that causing you problems now? | 17:15.58 |
| ray_laptop | mvrhel_laptop: probably alpha_b will only be 0 if the pixel was never written, so replace or OR would be the same | 17:16.10 |
| mvrhel_laptop | oh you have the issue when a_s = 1 | 17:16.22 |
| | as you were ignorning the backdrop | 17:16.34 |
| ray_laptop | mvrhel_laptop: did you see the email? | 17:16.37 |
| mvrhel_laptop | which is there when you have the non normal blend | 17:16.50 |
| | ray_laptop: with a_b = 0 is not really going to happen except for unfilled pixels | 17:17.20 |
| | s/with/right | 17:17.28 |
| | ok. so I think we have that resolved right? | 17:17.49 |
| ray_laptop | mvrhel_laptop: yes, I think that covers it. | 17:18.36 |
| mvrhel_laptop | ray_laptop: I got your new email | 17:18.40 |
| | let me open gs | 17:18.44 |
| ray_laptop | mvrhel_laptop: and this discussion allows me to fix/replace the problems identified in the email | 17:19.17 |
| mvrhel_laptop | oh so we are done? | 17:19.38 |
| ray_laptop | the use of dst[num_comp] was REALLY bad | 17:19.47 |
| | mvrhel_laptop: yes, we are done. Thanks. I am going to make those changes/fixes and run regression, then may ask for your review. | 17:20.17 |
| mvrhel_laptop | ok sounds good | 17:20.23 |
| ray_laptop | mvrhel_laptop: so get back to real work ! ;-) | 17:20.44 |
| | mvrhel_laptop: and thanks | 17:21.25 |
| mvrhel_laptop | np. I don't think I did much.... | 17:21.37 |
| ray_laptop | (oh, I _did_ say that) | 17:21.37 |
| | mvrhel_laptop: well convincing me that it's OK to treat all the funky BM's as OR cases helps (your example of the image) | 17:22.24 |
| | Forward 1 day (to 2017/05/24)>>> | |
Log of #ghostscript at irc.freenode.net.