| <<<Back 1 day (to 2020/02/18) | Fwd 1 day (to 2020/02/20) >>> | 20200219 |
petercher | I'd like to discuss the bug 688990 - imagemask interpolation doesn't quite match Adobe | 00:45.20 |
| The reference implementation seems to have white pixels before the scan line and black pixels after the scan line. | 00:46.41 |
| gs and my implementation assume white pixels on both sides. Is this acceptable? | 01:16.03 |
| Can somebody comment on the bug 688990, "imagemask interpolation doesn't quite match Adobe" ? | 16:17.12 |
Robin_Watts | Hi petercher | 16:39.40 |
| If you have a mechanism for getting an exact match, then we'd be interested in seeing it. | 16:40.24 |
petercher | The context on the left and right edges used by on Adobe is different. | 16:42.08 |
| Does gs want this bug? | 16:42.22 |
Robin_Watts | petercher: The issue is this... | 16:44.31 |
| It's a device dependent operation. | 16:44.42 |
| If we can exactly match adobe then there is some arguable benefit to us taking a new version. | 16:45.06 |
| If we can't exactly match adobe, then it's a question of us moving from one thing that doesn't quite match to another thing that doesn't quite match - much less of a benefit. | 16:45.45 |
| Whether it's worth our time is a question of complexity etc. | 16:46.26 |
ray_laptop | Robin_Watts: and from your comment #8, does Adobe even match Adobe (presumably we were comparing to CPSI which we no longer have running anywhere) | 16:47.13 |
Robin_Watts | If you have a commit ready to go, then clearly we'll take a look (and we'll honour the bountiable status) as long as it's not stupidly complex. | 16:47.28 |
| I suspect this is one of those bugs that we'd never have bothered with if it came up again today, but given that it has, let's see what you've come up with. | 16:48.03 |
petercher | I want to know what to do with the right edge. | 16:48.22 |
Robin_Watts | petercher: I would not put any more effort into trying to get an exact match to adove. | 16:48.57 |
| adobe. | 16:48.59 |
| as long as it does something that doesn't look awful, we probably don't care. | 16:49.22 |
petercher | OK, I'll submit a patch today. | 16:50.31 |
Robin_Watts | Thanks. | 16:50.42 |
petercher | Please also look at the big 702105 | 16:51.01 |
Robin_Watts | Bear with us a bit though, cos we're in the throes of preparing a release, so we may not get to look at it immediately. | 16:51.03 |
ray_laptop | Robin_Watts: how do we test/verify the scaling -- it is likely to cause LOTS of bitmap diffs, and AFAIK we don't have "real Adobe PS" available (we used to have ScanVec Amiable CPSI, but that dongle no longer works on modern motherboards -- I last tried it > 3 years ago when my MB died) | 16:51.09 |
kens | petercher I'm still awaiting a response on bug #690595 | 16:51.35 |
Robin_Watts | ray_laptop: Presumably petercher has something he's been working against? | 16:52.56 |
| petercher: I'll need to dig into bug 702105 a bit to understand what the patch achieves. | 16:53.33 |
petercher | kens: Oh, that bug affected the spool size rather than the output file. | 16:53.41 |
kens | Well can you tell me how to reprooduce a result please ? | 16:54.15 |
petercher | Yes, the old SAi PhotoPrint. | 16:54.24 |
kens | Could you stick a method in a comment on the bug please ? I'm in the middle of a different problem and I'll only forget.... | 16:54.52 |
ray_laptop | petercher: Yes, that's what we have (we bought a copy years ago) but, I can get it to run anymore. How is yours running? | 16:55.50 |
petercher | Robin_Watts: The old code increased the counter too much and missed end-of-line condition. | 16:56.15 |
ray_laptop | s/can get/can't get/ | 16:56.21 |
Robin_Watts | petercher: Right, but I need to convince myself that's a reasonable fix rather than improving the end-of-line condition. I'm a bit buried at the moment, but I have it on my list now. | 16:57.22 |
petercher | Robin_Watts: I'll attach my full test file to 702105. This file will establish the base state for the old interpolation. | 17:00.03 |
Robin_Watts | ok. | 17:03.40 |
| ok, I've looked at 702105 and it looks plausible to me. I'll get that in. Thanks. | 17:19.36 |
| <<<Back 1 day (to 2020/02/18) | Forward 1 day (to 2020/02/20)>>> | |