| <<<Back 1 day (to 2014/10/08) | 2014/10/09 |
Mulover | Hey there! Any pros here that can me explain a (probably) special one with mudraw? | 09:37.17 |
kens | Don't ask to ask, just ask | 09:38.49 |
Mulover | Got it :-) I created a PDF containing 3 elements: a background and two glyphs. the glyphs are placed widht 50% opacity and overlap a bit. The special thing is, the glyphs are faux bold. If i view the PDF with Adobe reader, it looks like expected. With mudraw / sumatra though you see the faux fragments overlapping. Example: http://ge.tt/1Roz6y02 | 09:42.19 |
kens | So are the 'glyphs' using a real font, or are they drawn as paths ? If they are using a real font, is it embedded ? | 09:43.07 |
Mulover | Embedded, real font. | 09:43.26 |
kens | Looks to me like you have a stroke/fill colour mismatch problem | 09:43.51 |
| Obviously we would need to see the original PDF file | 09:44.12 |
Mulover | Its inside the link: http://ge.tt/1Roz6y02 | 09:44.29 |
kens | Hmm so it is, OK give me a minunte | 09:44.46 |
| As I expected, teh text is drawn with rendering mode 2, so it is both stroked and filled | 09:46.11 |
| Ah, and its stroked and filled with black | 09:47.00 |
| Looks like it might be a bug in transparency, you should probably open a bug report | 09:47.36 |
Mulover | So now? I can somehow understand why this happens, but i dont understand why Adobe Reader gets around this. The color used does not matter. Happens the same for all colors. | 09:47.44 |
kens | As I said, open a bug report | 09:48.17 |
Mulover | OK. Thank you. Firefox PDF-Viewer shows the same bug, Chrome does not. Funny.... | 09:48.50 |
kens | Ghostscritp gets it correct too, like I said it looks liek a transparency bug | 09:49.10 |
Mulover | Thanks. | 09:49.50 |
kens | NP | 09:49.55 |
tor8 | kens: yeah, it's a limitation in that we don't automatically create an isolated knockout group for stroke+fill in mupdf | 10:08.23 |
kens | Ah, I didn't know that..... | 10:12.09 |
| Looks like GS actually gets that one right | 10:12.25 |
| Well he did open a bug report 695582 | 10:13.08 |
| Personally I wouldn't rate it as major, but that's up to you. | 10:13.36 |
Mulover | I dont know if it helps you guys, but older GS (8.62) seem to share the bug too. | 10:14.43 |
kens | Seems like tor8 already knows the problem | 10:15.00 |
Robin_Watts | Mulover: Yes, we fixed it in gs | 10:15.11 |
| well, mvrhel did, by pushing a group as appropriate. | 10:15.25 |
| I am somewhat embarassed that mupdf gets it wrong. | 10:15.34 |
Mulover | So there is hope? :-) | 10:15.45 |
Robin_Watts | Mulover: It's probably a days work, but I'm horribly buried in something else. | 10:16.08 |
kens | It sounds fairly straoight-forward to fix, to me | 10:16.08 |
| But then, I know nothign about MuPDF :-) | 10:18.01 |
| I cna't beleive its harder to fix than Ghostscript though | 10:18.11 |
Mulover | Nice to read that there will be a fix, you guys are great! Looking forward for 1.7 then :-) | 10:19.40 |
tor8 | chrisl: http://ghostscript.com/~tor/ghostscript.com/new/ results of hacked awki script and a new stylesheet and navbar based on your prototype | 11:23.44 |
chrisl | tor8: looks fine | 11:25.36 |
tor8 | chrisl: do you want to take it from there, or should I migrate the files and replace the existing website? | 11:48.56 |
| Robin_Watts: paulgardiner: fix for zeniko's warnings and some mujs bugs on tor/master for when you need a break from sot | 13:24.23 |
nsz | btw i found other issues wrt regexp: inside [] most escaped chars should be accepted as themslefs | 13:30.24 |
| eg /[\#\?]/ should be the same as /[#?]/ | 13:30.47 |
Powl | Hello, is there any Javadoc for MuPdf using in Android? | 13:40.25 |
Robin_Watts | Powl: No. | 13:41.16 |
| Are you wanting to integrate mupdf into your own application? | 13:42.59 |
| Specifically, are you hoping to take our example app and tweak it, or do you want to drive the MuPDF api programmatically? | 13:43.23 |
chrisl | tor8: sorry, missed that. Is there anything required in the awki files, or should it all just work? | 13:44.03 |
Powl | i'd like to integrate it and tweak it | 13:44.58 |
| this lib is very fast in rendering, and zoom function works well. I'd like to integrate a custom toolbar, double tap to zoom and a snippet function. | 13:46.38 |
tor8 | chrisl: I just updated the style.css and awkihtml.awk files | 13:47.36 |
| chrisl: feel free to take either the updated awkihtml script or just use the generated html files | 13:47.50 |
chrisl | tor8: I think we'll just remake the html files in situ - obviously, I'll have to update the downloads pages, but I'll do that at some point | 13:49.31 |
henrys | tor8:what is the name of our web git program? | 13:49.37 |
rSilva | Hi guys | 13:51.37 |
| I'm having a some troubles with MuPDF | 13:51.59 |
| documents with images make the app crash (in Android) | 13:52.46 |
tor8 | nsz: right. the spec for regexps in ecmascript is an exercise in unraveling entangled spaghetti... :( | 13:53.22 |
Powl | i also have images in my pdf, but no crash (Android) | 13:53.54 |
kens2 | : rSilva I'm pretty sure thqt's not generally the case. What version are you using, did you compile it yourself, can you postr an example ? | 13:54.07 |
tor8 | but yes, it does look like regexps allow identity escapes (unlike the rest of ecmascript) | 13:54.09 |
rSilva | I've been following the C code and was kind of able to find where the problem occurs | 13:54.43 |
| void fz_fill_image(fz_device *dev, fz_image *image, const fz_matrix *ctm, float alpha) in device.c | 13:55.03 |
| I'm using the latest source code from Git | 13:58.19 |
kens2 | OK so you built it yourself then. | 13:58.38 |
| Now, can you post an example file ? | 13:58.44 |
rSilva | yes | 13:58.44 |
| the pdf that causes the crash? | 13:59.00 |
kens2 | Yes | 13:59.03 |
nsz | i noticed the complicated regexp grammar description.. | 13:59.06 |
| it is also highly inconsistent with posix regexp | 13:59.20 |
| or other variants of regexps | 13:59.28 |
| which does not make things easier to understand | 13:59.47 |
tor8 | nsz: indeed. | 14:00.14 |
nsz | i wanted to eval a part of a big js code (some function definitions) for web automation (these happen not to reference the dom or other browser context) | 14:02.43 |
| but it's hard to cut out the lexically correct part with simple tools because it's hard to figure out where are the comments/strings/regexps begin or end | 14:03.37 |
tor8 | eval as in run through the eval() function? | 14:03.39 |
rSilva | http://we.tl/7vw3bp4357 | 14:03.47 |
nsz | i know which function i want to call with what argument, the function calls other functions but they all do string operations | 14:04.29 |
| and i want to do this without invoking a whole browser thing | 14:04.51 |
| so i need to eval part of a js code | 14:05.18 |
| the solution can be arbitrarily hackish | 14:05.35 |
| and i'm trying to do the js code interpretation part with mujs now | 14:06.10 |
| (more specifically this is for the mundane task of streaming/downloading youtube videos without a browser, i have various scripts to do webautomation but this one broke a while ago..) | 14:08.44 |
kens2 | rSilva : your file works for me. tor8 or Robin_Watts can one of you try this on a proper Android device please ? | 14:08.45 |
rSilva | I can open in other readers | 14:09.07 |
| but using it inside my app will always crash | 14:09.17 |
| I have other documents that the app open just fine | 14:09.32 |
kens2 | Have you tried the standard MuPDF demo ? | 14:09.37 |
rSilva | I will | 14:09.49 |
| my code is based on the demo | 14:10.23 |
chrisl | tor8: I've done the main ghostscript.com site. The download pages will take a while..... | 14:10.33 |
kens2 | 'based on' is not the same as the demo. Its going to be a lot easier if this goes wrong with the standard demo | 14:10.55 |
rSilva | you're right | 14:11.47 |
kens2 | Th image is in an Indexed ICCBased colour space (N=3) which 'might' be the issue,it may be the colour space not the image causing a problem. | 14:13.35 |
| I see this file also uses overprint | 14:14.18 |
| Though not for the image | 14:14.36 |
tor8 | nsz: there's a tentative regexp identity escape fix on tor/master | 14:14.44 |
| if you could try that out and see if it helps I'd appreciate it | 14:14.53 |
rSilva | it opens ok in the PlaySTore demo | 14:15.02 |
tor8 | http://git.ghostscript.com/?p=user/tor/mujs.git;a=summary | 14:15.14 |
kens2 | rSilva : then that suggests to me some problem in the way you have built the library, or the way you are using it | 14:15.25 |
| I'm not really qualified to help with either of those | 14:15.41 |
rSilva | maybe the NDK version is the issue | 14:15.59 |
kens2 | Possibly, what version are you using ? | 14:16.08 |
rSilva | android-ndk-r8e | 14:16.25 |
| because the latest one has issues | 14:16.36 |
kens2 | I thinik that's known to work. | 14:16.40 |
rSilva | that was the one recommended here | 14:17.01 |
kens2 | Like I said, I'm not really qualified to help you, I don't do Android...... | 14:17.55 |
jogux | henrys: are you about? we should revisit our current favourite topic if so :) | 14:18.08 |
rSilva | thanks kens2 | 14:18.40 |
Powl | just one question. In the MuPdf Demo I noticed, that when I highlight text on a page which has two or more columns of text, the whole (in one line) text in every column is highlighted. Is it possible to highlight only the tapped text? | 14:21.36 |
henrys | jogux:ugh | 14:21.43 |
jogux | henrys: sorry :-) | 14:23.10 |
Robin_Watts | Powl: sorry, stepped out. | 14:27.19 |
Powl | ok | 14:27.47 |
Robin_Watts | Powl: It should be possible to do what you describe. | 14:28.08 |
| Highlights in mupdf are done by dragging an area. | 14:28.22 |
| If you only drag the area over the first column, you should be OK. | 14:28.39 |
nsz | tor8: i cannot try it right now, but will do that a few hours later | 14:29.07 |
Powl | So i have to change the code, right? Because highlighting goes over the whole page, no only over the column, where my finger pos is | 14:30.11 |
| at the moment | 14:30.20 |
Robin_Watts | Powl: Ok, that's probably because we are gluing the lines in the two columns together. | 14:30.34 |
| You'd need to look into the hairy structured text code I suspect. | 14:31.07 |
| so yes, tit will require changes at both C and java level. | 14:31.21 |
Powl | hm ok...C also? | 14:31.40 |
Robin_Watts | Powl: Probably, yes. | 14:32.50 |
Powl | Another question. Is it possible to make a snippet funktion. Like drawing a rectangle and cut out the selected area and save it as bitmap? | 14:34.02 |
Robin_Watts | Powl: Sure. But that would probably require C level coding too. | 14:34.52 |
Powl | is there a function to get a image of the current page? | 14:35.32 |
Robin_Watts | Not at the java level, currently. | 14:35.48 |
| The MuPDF API is a C level one. | 14:36.06 |
| We've exposed just enough of it into java to do what we need for the app. | 14:36.29 |
Powl | i need this funtions in a android pdf reader, | 14:36.44 |
Robin_Watts | so to call different bits of the API will probably require more JNI code to be written. | 14:37.17 |
| I have some generic JNI bindings in progress, but they aren't released yet. | 14:37.35 |
Powl | and what about making a screenshot of the layout in which the reader view sits, and extract a bitmap out of the layout? | 14:38.16 |
| ok, I see | 14:38.16 |
francisco_ | hi | 14:41.10 |
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. | 14:41.10 |
aneuronal | anyone knows who make a notes in mupdf ? | 14:42.16 |
nsz | i cannot parse that question | 14:42.46 |
aneuronal | it is possibly create a note in MuPDF ? | 14:43.34 |
| or highlight text ? | 14:43.48 |
Robin_Watts | aneuronal: You can certainly highlight text. | 14:44.08 |
| sticky notes aren't implemented yet. | 14:44.14 |
nsz | last time i checked there were no pdf editing functions | 14:44.15 |
| i guess the point of creating notes is to save it along the pdf | 14:45.45 |
aneuronal | yes | 14:45.57 |
| its really fast but cannot add any notes | 14:46.42 |
| Robin how i highlight the text? | 14:47.40 |
Robin_Watts | It's implemented in the android viewer. | 14:48.06 |
aneuronal | I'm use linux version | 14:49.17 |
Robin_Watts | Click the (v) icon, then (+) then select the annotation you want. | 14:49.33 |
| Oh, well, you're out of luck then. | 14:49.42 |
aneuronal | xd | 14:49.47 |
| would be nice to implement these features in the version of Linux | 14:51.12 |
Robin_Watts | yes. | 14:51.25 |
rayjj | kens2: I saw the post on bug 695533 with the huge (20852x13807) image. Is the reason that the whole image is allocated because it's JPX ? | 14:53.47 |
| (I could look, but thought you might know offhand) | 14:54.15 |
kens2 | rayjj yes I believe so, we need a memory area the size of the final image | 14:54.23 |
| I had a quick discussion with SHelly about it, the image has 6 resolution levels, so we *could* decode less of it, but..... | 14:55.01 |
| The canvas is 150x100 inches, so for all but the lowest resolutions we would need the whoel thing anyway | 14:55.08 |
| and the width and height are in the PDF< so we can't easily change it | 14:55.24 |
| raqyjj we know the GS output matches Adobe for bug #695582, we discussed it earlier in the logs | 14:57.22 |
| tor8 says he knows what the problem is | 14:57.35 |
| rayjj ^ | 14:57.53 |
kens2 | decides to leave the customer question to marcos | 14:59.39 |
rayjj | kens2: right (re the mupdf transparency bug) I hadn't read the logs yet, but this way it's captured in the bug report (not just the logs) | 15:01.59 |
| kens2: about the download of the latest version ? | 15:02.25 |
kens2 | yeah that one, leaving it for Marcos to reply | 15:02.42 |
rayjj | kens2: I was leaving that for Marcos as well | 15:03.04 |
kens2 | :-) | 15:03.11 |
henrys | mvrhel_laptop: what are your plans for adobe cs? are you going to sign up for the cloud stuff. I wonder if we should have a business subscription? | 16:08.42 |
mvrhel_laptop | henrys I have been running CS3 for a long time now | 16:09.23 |
| I really dislike the subscription stuff, but I suppose at some point having the latest version is good | 16:10.02 |
henrys | mvrhel_laptop: never really needed an upgrade? | 16:10.03 |
mvrhel_laptop | no not really | 16:10.08 |
| it does all that I need | 16:10.13 |
kens2 | mvrhel_laptop : is there a way to prevent embedding of ICC profiles in output JPEGs from Ghostscript ? | 16:11.08 |
mvrhel_laptop | kens2: well, we would have to add that as a command line option | 16:11.31 |
kens2 | OK so we can't do it now (got a question by email from an old associate) | 16:11.51 |
mvrhel_laptop | who wants it not to be embedded? | 16:11.51 |
| and why? | 16:12.05 |
kens2 | Dwight Kelly, I'm asking him now why :-) | 16:12.06 |
henrys | mvrhel_laptop: I have to evaluate a product that requires indesign, so I may need to subscribe. | 16:12.18 |
jogux | is the cloud subscription shareable? It would probably be handy if we could get access to Illustrator for a bit at some point. | 16:12.49 |
mvrhel_laptop | henrys: find out the pricing for a group license | 16:12.50 |
kens2 | mvrhel_laptop : what ICC profile gets embedded if you set -dUseFastColor ? | 16:12.55 |
henrys | mvrhel_laptop: or it might throw it at you, it'll only take an hour or so | 16:13.04 |
mvrhel_laptop | kens2: that is a good question. I suspect the default device profile. | 16:13.30 |
kens2 | OK that was the one that really surprised me. | 16:13.42 |
mvrhel_laptop | kens2: that would be a good case for it not to be embedded | 16:13.46 |
henrys | jogux: you pay per seat | 16:13.51 |
kens2 | :-) | 16:13.51 |
mvrhel_laptop | in any device | 16:13.54 |
| henrys: that is fine, I can do in design work pretty quick | 16:14.07 |
| kens2: perhaps the png, tiff and jpeg devices should not embed when fastcolor is set | 16:14.48 |
| lets open an enhancement for that | 16:15.03 |
kens2 | mvrhel_laptop : that, at least, seems reasonable to me. I'll let you know what Dwight comes back with as a use case | 16:15.15 |
mvrhel_laptop | kens2: ok good idea. perhaps there are cases where you want color management to occur but the profile not to be embedded | 16:15.54 |
jogux | henrys: it looks like if we get 'creative cloud for team' we can transfer licenses between team members as/when we want (but not sure how much more expensive that is :-) ) | 16:16.16 |
kens2 | COUld be, I expect Dwight has a reason (he usually does) I'm just not sure what it is. | 16:16.17 |
mvrhel_laptop | can't think of any... | 16:16.21 |
| but would be interested to hear | 16:16.27 |
kens2 | mvrhel_laptop : Wantr me to open an enhancement request ? | 16:16.31 |
mvrhel_laptop | kens2: yes please | 16:16.37 |
kens2 | OK will do it now. | 16:16.41 |
mvrhel_laptop | thanks | 16:16.45 |
henrys | jogux: yeah I saw that when you have to contact sales you know it's going to be high prices | 16:17.33 |
jogux | :-) | 16:17.47 |
| henrys: one thing I saw (two years old) said 'normal' was $49/month, 'team' was $70/month. | 16:18.21 |
| henrys: http://www.adobe.com/creativecloud/buy/business.html?PID=2159997 says 49.99/month? | 16:19.04 |
henrys | jogux: any chance we can get what you want done with inkscape, we have a relationship with those guys and they'd help if needed | 16:20.31 |
jogux | henrys: I've not had any success with such in the past. :( | 16:21.23 |
kens2 | RIght, I'm off. G'night all | 16:22.56 |
henrys | jogux: I guess we could get a licence or two at that price. One of my todo items for the meeting was hammer down what exactly we need to do graphics wise. I'll move to skype | 16:25.55 |
rayjj | I have CS3 as well. I generally don't like to upgrade because they "improve" the UI every time and move stuff around (you'd think they made money selling training or something). | 16:51.08 |
mvrhel_laptop | yes. I agree rayjj | 16:51.37 |
rayjj | the only nice thing about subscription licensing is that you don't have to jump through hoops when getting a new laptop (contacting them to get the re-activation) | 16:54.32 |
Robin_Watts | rayjj: Bah. They should just have done it through steam. | 16:56.00 |
rsc | Does somebody have a hint what could be wrong if a PDF file generated using ghostscript has text that can be marked in a PDF reader but copy & paste leads to strange symbols only? | 20:05.51 |
gokce | Hello, I have just installed MuPDF android application and I couldn't really figure out the purpose of the left-most button (that looks like a chain link), anyone? | 20:13.38 |
Mulover | @gokce: The most simple case i can think of: the application / system where you paste is not capable of viewing the selected symbols? | 20:30.37 |
gokce | Mulover: Can you rephrase that? The most simple case for the button? I didn't really get it.. | 20:34.28 |
Mulover | You could debug the content by using mudraw -t pdffile.pdf > debug.txt and view the textfile to check the text and its codepoints. Maybe some weird font with invalid codepoints was uses. | 20:35.00 |
gokce | Mulover: Are you talking about the android application? | 20:36.02 |
rsc | Oh, TrueType vs. Type1 seems to be the reason for my garbled copy & paste results. Does that wake somebody up? :) | 20:36.35 |
Mulover | Oh damn, sorry my eyes :-( I was talking about the ruined copy & paste results. | 20:39.00 |
gokce | Mulover: Ahh ok, that makes sense now. | 20:39.54 |
rsc | Ah, so you meant me. | 20:39.56 |
Mulover | rsc: I dont think the font type does matter. Could you provide the PDF? | 20:40.09 |
rsc | Mulover: not publically, but yes | 20:41.28 |
Mulover | Then try to "debug the content by using mudraw -t pdffile.pdf > debug.txt and view the textfile to check the text and its codepoints. Maybe some weird font with invalid codepoints was uses.". I got no E-Mail at the moment. | 20:44.14 |
| gokce: i had to LOL reading my answer to your question :-) | 20:50.17 |
gokce | Mulover: I was already feeling stupid for not figuring out what a button does and when you said that stuff I got even more suspicious of myself lol. | 20:53.19 |
| Mulover: Do you happen to know the answer by the way? | 20:53.41 |
Mulover | I have got no fu.... clue ;-) Maybe just check the codepoints? ;-) | 20:57.46 |
rsc | Mulover: "warning: not building glyph bbox table for font 'AWSBZT+DejaVuLGCSans' with 4141 glyphs" and debug.txt only contains lots of "?" characters | 20:58.31 |
Mulover | The first one should not be a problem. I see those all the time and never saw any gotchas. This might be caching thing or so. If you turn your text editor into hex mode, do you see the same bytes for all characters? | 21:02.05 |
rsc | All characters (except the spaces and line breaks) are 3f | 21:03.20 |
| Sorry, there are no spaces. Everything except line breaks (0a) is "?" (3f) | 21:04.01 |
Mulover | Without seeing the pdf if have got no more ideas. | 21:04.22 |
rsc | Okay. Did you receive my impolite /msg? | 21:04.33 |
Mulover | Actually not. I suppose the web client here is not capable. | 21:05.05 |
| rsc: ah there is message! i will take a look (de, ja? Guten Abend :-)) | 21:10.58 |
rsc | Mulover: hehe, it's German, yes. | 21:12.41 |
henrys | wow I really missed this when it came along: http://proceedings.spiedigitallibrary.org/proceeding.aspx?articleid=1348353 | 21:21.00 |
Mulover | I had a look at rsc's secret pdf. He is right and i can only make assumptions. I have to admit, i am no expert at all. While viewing his PDF is was able to copy & paste its textual contents perfectly with google chromes and firefox internal pdf reader. Mudraw rendered a nice png from it. But mudraw -x and mudraw -t only produced garbage. Adobe Reader and Sumatra showed the text, but copy & paste broke the text content. | 21:30.40 |
| I suppose the font is broken?! | 21:31.38 |
rsc | The font or the usage of the font? | 21:31.54 |
| DejaVuLGCSans should not be that special | 21:32.30 |
Mulover | I saw that.... Better ask someone who can analyse your case better than me. Sorry. | 21:33.22 |
rsc | Some hints (which might be wrong) at Google brought me to TrueType rather Type1 or Identity-H encoding as possible cause. However I don't have any clue here. | 21:33.31 |
Mulover | How do you generate that pdf? Not within gs itself i suppose. | 21:35.53 |
| You use gs to convert something to pdf? | 21:36.07 |
rsc | Mulover: A binary (compiled from C code) generates PostScript (%!PS-Adobe-3.0) and that one gets converted to PDF. | 21:39.38 |
Mulover | Only idea i have: the font gets mangled within the process referencing weird codepoints. But thats just a guess, sorry. | 21:49.20 |
rsc | Aha, if I replace "Identity-H" in the postscript by "Identity-UTF16-H" then I get a broken looking PDF but copy & paste works. | 21:58.20 |
| (well with spaces after each character) | 21:58.29 |
Robin_Watts | gokce: It enables/disables links in the document. | 21:58.45 |
Mulover | Well: then supply two versions! One for copy & paste and a good looking one! LOL. Do you have any chance bringing your process to a full lutf-8 (or latin1) workflow? Maybe using iconv? | 22:03.58 |
rsc | Mulover: it should be full UTF-8 already. | 22:04.36 |
Mulover | Like is said before, only (probably bad) guesses from me: If you look at the postscript file(editor), it looks good? Tried another gs version? | 22:07.56 |
| Robin_Watts: Did you fix my transparency bug? The day is over! Just kidding ;-) | 22:08.30 |
gokce | Robin_Watts: Ahh, that was my first guess but it doesn't really work for me. I guess I don't have an appropriate example pdf to test it. Anyway thanks.. | 22:13.38 |
rsc | Mulover: independent of version as it seems. | 22:15.09 |
Mulover | It "smells" like an encoding issue to me. Browsers are kind of clever in detecting encodings. That would explain why both chrome and firefox copy & paste the content correctly. | 22:16.54 |
| If you get some of the pros here for your issue, they can tell you the reason within minutes.... | 22:17.52 |
| Good night everybody, at least for the CETs and around. | 23:03.11 |
| Forward 1 day (to 2014/10/10)>>> | |