| <<<Back 1 day (to 2016/01/28) | 20160129 |
anes_ | hi all | 04:26.49 |
| right now, i'm looking at epson device driver (gdevepsn.c) | 04:27.56 |
| how to get systemdict value from inside device driver? | 04:28.49 |
| i want to add a control command for the printer, but i have to calculate a value based on -dDEVICEHEIGHTINPOINTS | 04:29.29 |
| sorry, -dDEVICEHEIGHTPOINTS | 04:29.57 |
Robin_Watts | Shiny new bumper sized passport arrived. | 12:00.53 |
| New ESTA obtained, updated info on the upcoming flights. What have I forgotten? :) | 12:01.21 |
chrisl | Oh, good call, I need to renew my ESTA, too..... | 12:02.41 |
kens | got 12 months on mine yet | 12:04.07 |
Robin_Watts | kens: I had 3 years on my last one :( | 12:04.31 |
| But Kenya/Tanzania want 3 double blank pages, and i only had 2. | 12:04.55 |
kens | Not sure I have any doubles.... | 12:05.08 |
Robin_Watts | Hence I have a bumper sized passport, and I've stapled 20 of the pages together :) | 12:05.11 |
| http://www.ebuyer.com/shop/weekly-deals (960GB SSD for 160quid) | 12:07.20 |
kens | getting cheaper every day..... | 12:07.37 |
Robin_Watts | kens: The amazon deal in Jan was cheaper, but 160 is not a bad price. | 12:11.37 |
| I now have a laptop where the only spinning thing is the fan. | 12:11.51 |
tor8 | Robin_Watts: so the AwtDrawDevice looks like it's never been tested... it's got different numbers of arguments between C and java and it segfaults when I fix that... | 12:36.07 |
Robin_Watts | tor8: It has indeed never been tested. | 12:36.18 |
| I sketched it out to convince myself that the framework it sits in was OK. | 12:36.58 |
tor8 | half of the jni c functions take and return 'int' where I believe we ought to be using 'jint' to be on the safe side for 64-bit compiles on windows | 12:37.37 |
Robin_Watts | tor8: I have got all your commits (except the WIP one) and all my commits together onto the robin/harfbuf branch. | 12:37.53 |
| int == jint on windows. | 12:38.03 |
| long != jlong on windows. | 12:38.18 |
tor8 | and I though jbyte was unsigned but it looks like it's signed! | 12:38.34 |
Robin_Watts | but technically, yes, we should return jint. | 12:38.38 |
| tor8: I found a bug in the fallback caching stuff (a 'serif' should have been a 'sans'), and that fix is pickled in. | 12:39.39 |
| Shaping looks better now. | 12:39.44 |
| But the code is not behaving as it should with "arabic digits arabic" | 12:40.05 |
| so I'm looking to fix that now. | 12:40.13 |
tor8 | Robin_Watts: I'm getting dropped words in the ManyLang.epub | 12:44.29 |
Robin_Watts | tor8: With my lastest code? | 12:44.57 |
| lastest. I like that. | 12:45.12 |
tor8 | yes, with your lastest code. the hebrew looks like it's missing some words and has ginormous word spacing. | 12:45.49 |
Robin_Watts | tor8: Leave it with me. | 12:46.11 |
| I reckon all but the last commit on the harfbuzz branch are good to go though. | 12:46.33 |
tor8 | Robin_Watts: the arabic in "English sentence with .... in the middle of it" looks correct at last! | 12:46.38 |
| Robin_Watts: yes, all but the last commit should be good | 12:47.31 |
Robin_Watts | I'll run a cluster test and commit it if it doesn't all fall over. | 12:47.59 |
| Ok, I have it correctly displaying arabic digits arabic. | 14:00.57 |
| but there are still cases that are hard to understand. | 14:01.17 |
| Suppose I have: 12345 abcde 6789 (where abcde is arabic) | 14:01.42 |
| That should be displayed as either: 12345 edcba 6789 or 123456789 edcba | 14:02.19 |
| Currently my code picks the latter. | 14:02.41 |
tor8 | Robin_Watts: looking through the java code, there's a fair bit of java boilerplate to initialization, finalize and destroy | 14:12.44 |
Robin_Watts | tor8: There is. I had more of the finalize/destroy stuff in the JNI, but you nixed that. | 14:13.21 |
tor8 | Robin_Watts: wonder if we could skip on some of that by making a NativeObject superclass that has the boilerplate and native pointer field | 14:13.21 |
Robin_Watts | Possibly. | 14:14.04 |
tor8 | still, it'll save like 5 lines for each class. probably not worth the bother. | 14:14.18 |
Robin_Watts | tor8: might save more than that, but... I'm unoffended by it as it is. We can think about that later if we want. | 14:15.22 |
tor8 | yeah. I need to get things working on desktop java | 14:15.39 |
| there were lots of typos and errors exposed when I added a javah-generated header file | 14:15.52 |
| I'm fixing them up as I go | 14:15.59 |
Robin_Watts | ok. | 14:18.45 |
| Gah. So there is custom allocator stuff in harfbuzz already - but it's still crap. I'll move to using theirs. | 14:19.12 |
tor8 | oh! | 14:19.41 |
| where is that? | 14:19.45 |
| I didn't see any when I was digging around | 14:19.54 |
Robin_Watts | hb-private.hh | 14:19.57 |
tor8 | ohhhh! | 14:20.19 |
| how did you spot that? | 14:20.21 |
Robin_Watts | I missed it too. Hunting for why it doesn't like cluster testing. | 14:20.36 |
| It's in line with what I'd done. it just requires some -Dcrap stuff at compile time. Will fix. | 14:21.02 |
tor8 | Robin_Watts: cool. | 14:22.42 |
| too bad it's compile time :( | 14:22.57 |
suporte_smarket | Hi! I 'm trying to convert a sRGB pdf to CMYK with pure black (0% 0% 0% 100%). I am using to use USWebCoatedSWOP.icc with -dKPreserve=1 but it always returns rich black (72% 67% 67% 88%). Any hints? | 14:30.33 |
kens | Presumably you are lenoardo_ under a new name ? | 14:31.12 |
suporte_smarket | yes! I am using the company name! | 14:31.26 |
kens | Hmm in that case, have you reviewed the AGPL to ensure you are complying with the licence ? | 14:31.57 |
suporte_smarket | no I didn't know about it | 14:32.26 |
| so i can't use a adobe profile in gs? | 14:33.16 |
kens | Ghostscript is open-source software, it is supplied under the Afeero General Public Licence (AGPL). Ths does not prevent its use commercially, provided you comply with the terms of the licence | 14:33.21 |
| You should read teh licence and check that your usage complies with the terms of the licence. THe use of an ICC profile is not relevant. | 14:33.54 |
Robin_Watts | suporte_smarket: Are you distributing Ghostscript to customers? Or are you using it to implement a web service for customers? | 14:33.58 |
suporte_smarket | No I am testing to see if it works | 14:34.18 |
kens | OK but are you *intending* to do any of the above if it does work ? | 14:34.32 |
suporte_smarket | Yes and pay if necessary | 14:34.44 |
kens | OK fair comment | 14:34.55 |
suporte_smarket | and not too expensive | 14:34.57 |
kens | The basic answer to your queston is that you need to get teh colour management sorted out. | 14:35.11 |
| In order to produce a CMYK PDF file from RGB input, we must convert the RGB to CMYK. | 14:35.38 |
| THe fist problem is; "is the input really RGB, or is it an ICCBased space ?" | 14:35.57 |
suporte_smarket | How can I check that? | 14:36.15 |
kens | Well the easiest way is to deompress the PDF file and read it. | 14:36.35 |
suporte_smarket | ok just a moment | 14:36.59 |
kens | If you can't do that (not many people can) then you will ahve to make an exmaple avialable so that we cna tell you | 14:37.01 |
suporte_smarket | ok I will make an example | 14:37.39 |
kens | As simple as possible please | 14:37.50 |
suporte_smarket | ok just a moment | 14:38.02 |
| hey | 14:56.08 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 14:56.08 |
suporte_smarket | here is the link | 14:56.13 |
| http://www.filedropper.com/zwucuzjb100531454079705 | 14:56.14 |
| of sRGB PDF | 14:56.25 |
| kens | 14:56.33 |
kens | That's a 2Mb file, not exactly simple ? | 14:56.57 |
| storming great image in it for starters I thought we were talking about text ? | 14:57.14 |
suporte_smarket | I must generate about 20 pages like this one | 14:57.21 |
kens | Its too complex, I'm not going to attempt to look at that | 14:57.37 |
suporte_smarket | if I remove images is better? | 14:57.56 |
kens | a single libe of text is better | 14:58.06 |
| In one fot | 14:58.12 |
suporte_smarket | oh | 14:58.13 |
kens | font* | 14:58.15 |
suporte_smarket | ok | 14:58.18 |
| http://www.filedropper.com/zwucuzjb100531454080194 | 15:02.51 |
| sorry but i cannot remove the marks on corners | 15:03.11 |
| hope its simple enough | 15:03.19 |
kens | Pretty crappy PDF file, the colour space isn't referenced from the Resources dictioanry | 15:06.13 |
| Oh actually it is | 15:06.26 |
| Its not sRGB its plain RGB | 15:06.36 |
| So basically you want to control an RGB to CMYK conversion | 15:07.07 |
| Ths is done via the colour management system, in our case we use LittleCMS, specifically LCMS2 | 15:07.25 |
| As I indicated last night, you will really need to talk to our colour expert, Michael Vhrel, but he's not online yet. | 15:08.16 |
| I thnk you will need to modify the default RGB and CMYK ICC profiles that we use for the RGB and CMYK spaces, and replace them with a pair where equal RGB values end up as K in the CMYK space. | 15:09.01 |
suporte_smarket | use -sDefaultRGBProfile and -sDefaultCMYKProfile ? | 15:11.06 |
kens | I don't know, I don't usually deal with teh CMS | 15:11.24 |
| More likely just DefaultRGBProfile, and set an OutputICCProfile for the CMYK output | 15:11.53 |
suporte_smarket | I got it, I will try that. Anyway, what time Michael's usually online? | 15:12.23 |
kens | CIOuld be any time now, he is on the West Coast of the US, so its just afetr 7 am there | 15:12.56 |
Robin_Watts | mvrhel_laptop: ping | 15:13.52 |
kens | LOL didn't notice he was here..... | 15:14.07 |
Robin_Watts | Well, his computer is :) | 15:14.19 |
suporte_smarket | Ok kens thank you for yout attention. Here it's lunch time if I cannot make this, I come back and talk to Michael! | 15:15.10 |
kens | I have a feeling Michael has dealt with thsi before somewhere | 15:15.43 |
suporte_smarket | Ok so I will try to google some more. Thank you! | 15:16.44 |
mvrhel_laptop | Robin_Watts: pong | 15:43.36 |
| going out to breakfast in a minute... | 15:44.00 |
Robin_Watts | mvrhel_laptop: Hi. i was just pinging you cos Kens/suporte_smarket were talking about you without using your nick. | 15:44.13 |
| I think suporte_smarket has gone to lunch though. | 15:44.33 |
mvrhel_laptop | ok. I will be back in about 30 | 15:44.47 |
Robin_Watts | mvrhel_laptop: Have a milkshake for me. | 15:44.52 |
mvrhel_laptop | a little early for that | 15:45.05 |
| eggs and bacon | 15:45.19 |
Robin_Watts | It's never too early for a milkshake. | 15:45.48 |
mvrhel_laptop | :) | 15:45.52 |
suporte_smarket | Just a question. How much does it cost the commercial licence for gs? | 16:11.53 |
Robin_Watts | Depends on your usage. And to stop the next question, however much info you give us, we will not name a dollar amount. | 16:13.12 |
| We will however point you at a man that can :) | 16:13.22 |
| We can put you in touch with Scott Sackett, our VP of sales. You talk to him, and answer his questions, and he'll give you a quote based on the info you give him. | 16:14.28 |
suporte_smarket | Ok I will have to ask my boss to contact him and she is not here yet. | 16:17.29 |
| We still have only one client using the pdf stuff so we still don't have much precise usage measure. But I believe we can reach an agreement. | 16:21.14 |
| Michael is back? | 16:22.57 |
Robin_Watts | He arrived, then left for breakfast. Should be back in less than 30 mins. | 16:23.31 |
suporte_smarket | ok! | 16:23.41 |
tor8 | Robin_Watts: hah! I finally got something drawing :) | 16:33.19 |
Robin_Watts | fab. | 16:33.30 |
tor8 | I have however horribly butchered the jni code | 16:33.37 |
HenrySti` | tor8: did you see Ron's email? | 16:33.40 |
Robin_Watts | I'm edging closer to having this stuff working. | 16:33.49 |
tor8 | HenrySti`: the one about the whatismybrowser.com thing? I went to that page and submitted the info to ron. | 16:34.32 |
HenrySti` | tor8: great thanks we're paid for support might as well get our money's worth | 16:35.10 |
| I'm in a coffee shop waiting on a car repair and can't readily reach #artifex if anyone wants something... | 16:36.26 |
jogux | HenrySti`: I need to chat about TC but that can wait :) | 16:39.08 |
HenrySti` | jogux:okay, emails fine too. | 16:42.15 |
| this is going longer than expected. I may head home in a bit. | 16:42.54 |
mvrhel_laptop | ok I am back | 16:50.15 |
| suporte_smarket: Let me read the logs to see what is going on before I ask you any questions | 16:50.47 |
Robin_Watts | mvrhel_laptop: suporte_smarket was called leonardo yesterday. | 16:51.18 |
| hence you might need to look at that conversation too. | 16:51.29 |
mvrhel_laptop | so the -dKPreserve option is to ensure that gray source colors are mapped to K only | 16:52.09 |
| it will not work with RGB source colors | 16:52.34 |
| so you do have DeviceRGB and you want to go out to CMYK though | 16:53.56 |
| hold on let me look at something | 16:54.55 |
suporte_smarket | hey im back | 16:55.45 |
| The only thing people is complaining is about the small text color | 16:57.12 |
| its becoming rich black | 16:57.27 |
| and they are saying that it goes bad at the printer | 16:57.49 |
| if i could replace the color ou just make small black text k only | 16:58.07 |
| i would be awesome | 16:58.13 |
| it | 16:58.17 |
jogux | finally twigs which window the ctrl-w that 'did nothing' went to. | 16:59.44 |
Robin_Watts | tor8: Aha. I'm there I think. | 17:07.44 |
| Updated version on robin/harfbuzz that looks good. | 17:07.54 |
| Except that all my right to left text that is supposed to be left aligned is coming out right aligned. | 17:08.13 |
| That's because of a commit of yours. Was there a reason for that? | 17:08.22 |
tor8 | Robin_Watts: the WIP commit is supposed to fix that | 17:09.22 |
| Robin_Watts: the default text-align is 'left' but if the document has dir=rtl then text-adjust: left=right and vice versa | 17:09.46 |
Robin_Watts | Is it finished enough for me to grab? | 17:09.59 |
tor8 | and if dir=auto then it's auto detected | 17:10.01 |
| Robin_Watts: yeah, should be | 17:10.08 |
Robin_Watts | I'm still getting hebrew coming out right aligned. | 17:11.05 |
| Maybe calibre is just getting it wrong, and it's SUPPOSED to be right aligned. | 17:11.34 |
tor8 | calibre and firefox are bolloxed and do the wrong thing. you might need to set the default dir to be LTR not auto | 17:12.36 |
| in order to match calibre and firefox | 17:12.41 |
Robin_Watts | tries "EPUBReader" extension for firefox | 17:14.19 |
tor8 | Robin_Watts: bah, this nativeinfo, nativeresource stuff is pretty nightmarish | 17:14.25 |
| Robin_Watts: just unzip the epub and open the xhtml file with firefox | 17:14.36 |
Robin_Watts | That shows it on the left too. | 17:15.22 |
tor8 | so the only reason for nativeInfo, nativeResource and the lock/unlock is to call AndroidBitmap_lockPixels/unlockPixels for the fz_pixmap samples? | 17:15.27 |
Robin_Watts | Yes. | 17:15.45 |
tor8 | Robin_Watts: near fz_xml_att(node, "dir"), change the else to if (!strcmp(dir, "auto")) and add a new else base_dir = DIR_LEFT_TO_RIGHT | 17:16.21 |
mvrhel_laptop | suporte_smarket: So I see two possible approaches. I am wondering if doing -dUseFastColor and setting the UCR/BG to ensure that equal amounts of RGB get mapped to pure K. The other approach would be to use a specific output CMYK profile for text objects which always maps to pure K for color close to the neutral. That profile would be specified using -sTextICCProfile=filename | 17:16.22 |
tor8 | that should amke it match calibre | 17:16.25 |
mvrhel_laptop | suporte_smarket: the latter approach would provide you with more control | 17:16.49 |
| it does require you to create an CMYK profile to perform the mapping | 17:17.12 |
| kens would have to chime in, but I assume pdfwrite respects the object based color conversions | 17:17.43 |
suporte_smarket | So I like the last option. but I have no idea how to do it. | 17:17.45 |
tor8 | Robin_Watts: how much performance would we lose by drawing to a mupdf-owned pixmap and have the android code memcpy the data from the fz_pixmap to the android bitmap at the end? | 17:17.49 |
kens | mvrhel_laptop : I would not guarantee that | 17:17.59 |
Robin_Watts | tor8: Massive amount. | 17:18.10 |
kens | Only if the PostScript colous space hgas a relevant ICC profile | 17:18.22 |
tor8 | under the hood, I expect the androidbitmap to just copy the data to an opengl texture | 17:18.32 |
mvrhel_laptop | kens: how is he forcing everything to CMYK in the output doc? | 17:18.48 |
Robin_Watts | tor8: What is the problem with doing the same for Awt stuff? | 17:19.00 |
kens | -sColorConversionStrategy=CMYK | 17:19.07 |
mvrhel_laptop | and so when that option is selected where does it call the cmm to do the conversion | 17:19.25 |
kens | whenever it needs to :-) | 17:19.40 |
tor8 | Robin_Watts: there is no equivalent to androidbitmap in awt | 17:19.46 |
suporte_smarket | my current command | 17:19.55 |
mvrhel_laptop | not when, but where | 17:20.02 |
tor8 | the closest thing is a BufferedImage and then magic happens to convert the stuff to XImage or XPixmap in awt drawing code | 17:20.08 |
Robin_Watts | tor8: Well, the equivalent is an int array which we fill with RGBA. | 17:20.14 |
kens | mvrhel_laptop : I don't know what you mean. Itas called form the pdfwrite code whenever it needs to convert a colour space | 17:20.28 |
tor8 | RGBA is not an available pixmap format. only ARGB. | 17:20.33 |
suporte_smarket | gs -dNOPAUSE -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -sOutputICCProfile=/home/user/CMYK/USWebCoatedSWOP.icc -dOverrideICC=true -sColorConversionStrategy=CMYK -dKPreserve=2 -dUseFastColor=true -sOutputFile=/home/user/teste97.pdf /home/user/zWUCUZJB_10049.pdf | 17:20.35 |
kens | SO 'where' would be 'lots of places' | 17:20.37 |
Robin_Watts | tor8: ok, ARGB :) | 17:20.44 |
tor8 | which might be equivalent to BGRA depending on endian-ness | 17:20.45 |
| on desktop java I'm not going to worry about an extra memcpy | 17:20.57 |
mvrhel_laptop | kens: ok fair enough, but what procedure do you call to do the mapping | 17:21.01 |
suporte_smarket | tried many different combinations | 17:21.13 |
Robin_Watts | Even with that fix, I still get my hebrew lines right aligned :( | 17:21.14 |
kens | I don't recall offhand, I'll look | 17:21.19 |
mvrhel_laptop | I just want to see where it is hooked in the the cmm kens | 17:21.19 |
Robin_Watts | Maybe it's something dumb I'm doing | 17:21.22 |
tor8 | Robin_Watts: I'll need to hack up the nativeinfo stuff; it currently assumes all native devices are draw devices | 17:21.53 |
Robin_Watts | It does? | 17:22.30 |
| It assumes that all native devices are CDevices. | 17:22.48 |
tor8 | struct NativeDeviceInfo has a fz_pixmap and pageX0, pageY0 fields | 17:23.01 |
| sorry, CDeviceNativeInfo | 17:23.20 |
Robin_Watts | tor8: Right, but not every device needs to use all those fields. | 17:23.25 |
tor8 | okay. | 17:23.35 |
Robin_Watts | But yes, currently in the implementation, it's quite possible that it assumes that all devices will use all those fields. | 17:24.04 |
| Again purely because the only 2 devices I sketched out were the Android and Awt draw ones. | 17:24.21 |
| so yes, some hacking of that stuff into a different form would not offend me. | 17:24.35 |
mvrhel_laptop | kens; maybe in write_color_as_process? | 17:25.16 |
suporte_smarket | how do I set UCR/BG in gs? | 17:25.21 |
kens | mvrhel_laptop : I call get_link and map_color | 17:25.42 |
mvrhel_laptop | kens: ok thanks. I will take if from there | 17:25.55 |
| kens: where do you call get_link? | 17:27.28 |
| I see a big chunk of commented out code in gdevpdfg.c | 17:27.42 |
kens | mvrhel_laptop : no idea, can't find it now | 17:28.13 |
| I can trace through the code tomorrow, but its kind of late now | 17:28.37 |
mvrhel_laptop | kens: ok no problem. have a good night. thanks | 17:28.48 |
| suporte_smarket: I am going to have to look at this a bit later. Sorry. I have too many things on my plate this second | 17:42.14 |
| suporte_smarket: pdfwrite is a little different in its conversions and it will take me a little time to see how it is handling things | 17:43.18 |
suporte_smarket | I have until tomorrow to solve this | 17:48.26 |
| I am trying to follow your tips right now | 17:48.58 |
Robin_Watts | suporte_smarket: While I have sympathy, your deadline is not our problem. | 17:49.32 |
| We have paying customers to deal with and they have to take priority. | 17:49.52 |
suporte_smarket | I know! I didn't mean to pressure or being incovenient... sorry | 17:52.16 |
| just saying I will probably be online for the next 48 hours on this... :) | 17:53.22 |
Robin_Watts | suporte_smarket: Sure. :) | 17:55.46 |
HenryStiles | Robin_Watts: looks like your Lenovo has gone mad ;-) | 18:12.33 |
Robin_Watts | I think network issues, plus firefox updates :) | 18:24.15 |
suporte_smarket | Hey guys! Just wanna tell you I fixed my thing another way. Thank you for your attention and see you some day! | 21:55.15 |
| Forward 1 day (to 2016/01/30)>>> | |