| <<<Back 1 day (to 2016/12/29) | 20161230 |
wich | Hi guys we noticed that the ghostcript 9.20 tarball has changed since it's original release | 13:15.25 |
| It also seems to be different from what is in the 9.20 git tag | 13:15.40 |
| can you confirm this? and if so, why was this change made? | 13:15.55 |
Robin_Watts | wich: What files are different from the git tag ? | 14:10.54 |
wich | Robin_Watts: e.g. all of expat is removed in the downloaded tarball | 14:34.08 |
| but more is changed | 14:34.16 |
| see https://sobukus.de/smgl/ghostscript-9.20.diff.bz2 | 14:34.38 |
Robin_Watts | wich: expat is only used by gxps. | 14:40.43 |
| Yeah, bz2 not so useful here. | 14:41.58 |
wich | also all kinds of fonts are changed | 14:59.33 |
| and some pcl/pxl stuff | 14:59.38 |
| don't know what the changes are though | 15:00.03 |
| the diff only shows that the binary data is different | 15:00.12 |
| Robin_Watts: uh, you can't bunzip2? | 15:00.29 |
Robin_Watts | I'm on a windows box at the moment. | 15:04.40 |
| The person who will know the answer to your question is chrisl. He's off until the 2nd or 3rd of jan. | 15:20.09 |
wich | hmmn, that's a bummer, no one else knows? | 15:20.35 |
sebras | tor8: I've set up sebras/mupdf-android-old.git now | 16:22.43 |
| tor8: do you mind pulling and checking that there's a sebras/wip? | 16:22.56 |
| and that you can see it. | 16:23.03 |
Robin_Watts | http://git.ghostscript.com/?p=user/sebras/mupdf-android-old.git;a=shortlog;h=refs/heads/wip | 16:30.13 |
| Urgh. I've spent all day digging into a scan converter problem, and there are 2 issues. | 16:31.02 |
sebras | I think I configured the thing correctly using the git-daemon-export-ok thing, but I'm not sure. | 16:31.13 |
| Robin_Watts: are you sure you're working on the scan converter? it soulds like you might dabbling in jpeg2000-related code... ;) | 16:31.50 |
Robin_Watts | Firstly, it's a huge screen area, so it's triggering banding even despite a large max bitmap, and that's showing up a problem with the banding code. | 16:32.05 |
| Secondly, the problem only occurs if I print a letter first. | 16:32.18 |
| Printing a letter somehow causes the subsequent fills to be done with a different fill adjust. | 16:32.35 |
| So, something is screwed with the clist and fill adjusts. | 16:32.48 |
sebras | is one with the cardboard. | 16:39.49 |
brettk80 | Hi all. Wondering if anyone can tell me why I'm seeing different quality of grayscaling output when using the same device settings in windows vs. linux. | 17:07.13 |
sebras | tor8: ok, so there are a few patches on mupdf/sebras/master and mupdf-android-old/sebras/master | 17:09.37 |
Robin_Watts | brettk80: Same version of gs ? | 17:09.55 |
sebras | tor8: using those I'm able to build an apk I can run on the device. | 17:10.39 |
brettk80 | Robin_Watts: Pretty sure. I'll double check. | 17:10.46 |
sebras | using the latest SDK/NDK combination. | 17:10.49 |
brettk80 | Robin_Watts: Actually no. Win is 8.64 and generates better output using the same settings than linux on 9.10 | 17:14.07 |
Robin_Watts | brettk80: Well, there is your first answer. | 17:14.28 |
| The obvious followup question: "Why is a later version worse?" is more interesting though. | 17:14.48 |
| Can we see the source file ? | 17:14.59 |
| v9 introduced color management, so it could be that we're getting a 'truer' response now, but it's one you don't like. | 17:15.47 |
brettk80 | Robin_Watts: Right.. Is the sDevice a config file where I can compare the settings? It hundres of source files that vary on color usage, image quality, etc. The issue is how the dithering is compared between the versions. The older is more smooth, the newer is more chunky. | 17:15.59 |
sebras | tor8: the same apk works in the emualtor too. | 17:16.17 |
Robin_Watts | brettk80: No, the -sDEVICE settings choose which device to use (devices are C compiled things). Not just "configuration files". | 17:16.43 |
tor8 | sebras: nice. | 17:16.46 |
Robin_Watts | What device are you using? | 17:16.57 |
brettk80 | Tiffg4 | 17:17.19 |
Robin_Watts | At what resolution? | 17:17.34 |
tor8 | sebras: mupdf:sebras/master LGTM. | 17:17.39 |
| sebras: the stuff on mupdf-android-old.git also LGTM | 17:17.50 |
brettk80 | Robin_Watts: Let me check the developers code. I'm just a net admin. I know it varies for fax between two settings. | 17:18.20 |
Robin_Watts | OK, fax is generally 200dpi. | 17:18.32 |
| So instead, try: gs -dDownScaleFactor=3 -sDEVICE=tiffscaled -sCompression=g4 -o out.tif inputfile | 17:21.54 |
| So instead, try: gs -r600 -dDownScaleFactor=3 -sDEVICE=tiffscaled -sCompression=g4 -o out.tif inputfile | 17:22.12 |
| (sorry). | 17:22.16 |
brettk80 | Robin_Watts: 204X98 or 204X192. gs -q -sDEVICE=tiffg4 -r$gs_r -sPAPERSIZE=$ps -sColorConversionStrategy=Gray -dProcessColorModel=/DeviceGray -dNOPAUSE -sOutputFile=$outfile $infile -c quit | 17:22.22 |
Robin_Watts | tiffscaled will give you better quality. It renders in 8 bits, then error diffuses down. | 17:23.02 |
brettk80 | Robin_Watts: Ok I can give that a try.. Thank you. | 17:23.10 |
Robin_Watts | hence 600dpi, with a DownScaleFactor of 3 gives you 200dpi. | 17:23.30 |
brettk80 | Robin_Watts: But fax has two settings.. Standard and Fine based on the resolution. | 17:23.45 |
Robin_Watts | So, you can try -r612x296 or -r612x584 | 17:24.35 |
brettk80 | sweet | 17:24.44 |
Robin_Watts | I tend to recommend that people use 600dpi, because PCL stuff can sometimes only render correctly at 600dpi. | 17:25.15 |
| but if your input is ps or PDF then you will probably be OK. | 17:25.30 |
brettk80 | it is all pdf | 17:25.42 |
| Robin_Watts: That looks super.. Thank you so much. | 17:27.55 |
Robin_Watts | Fab. | 17:28.03 |
sebras | tor8: pushed. | 17:52.29 |
| Forward 1 day (to 2016/12/31)>>> | |