| | <<<Back 1 day (to 2014/10/20) | 20141021 |
| kens | Well that's amusing, my device emits warnings 'implements obselete procedure image_data' and 'end_image' | 09:01.44 |
| chrisl | Are they getting called? | 09:07.29 |
| kens | No idea yet :-) THe warning comes form the graphics library :-) | 09:07.49 |
| | I'm really thinking its time for a good clean out here | 09:08.09 |
| chrisl | Good luck with that - the API for the '80's is with us for ever...... | 09:08.55 |
| kens | I know it would take a lot of time to do, but I think it would be a good idea. Surely nobody can really be using this obselete stuff ? Not when we've changed the name of one method and apparently emit warnings on another 2 | 09:10.03 |
| chrisl | Well, my position is well documented: there is simply no excuse for having 70+ methods in the device API. Frankly, I think the device API has been hampered, if not downright broken, since LL2 came out | 09:11.54 |
| kens | The setpagdevice implementation is likewise hampered by the put/get params problem as well. But lets tackle one thing at a time :-) | 09:12.32 |
| chrisl | And as you have pointed out on more than a few occasions, the current device form prevents us from being able to (easily) create a flexible device chain | 09:14.00 |
| kens | It does indeed, and the fact that the clist uses the same awful hackery (replacing device procs) as Robin's spy module demonstrates that this *has* been a problem in the past, and continues to be one into the future | 09:14.49 |
| chrisl | I also think that the separation of the devices into vector and "printer" devices is an unnecessary complication | 09:16.46 |
| kens | Yes I'm inclined to agree, its making my life more complicated at the moment | 09:17.10 |
| KunMyt | hi every body | 09:31.15 |
| | i have question: how do read pdf file by buffer Stream with mupdf ? | 09:32.08 |
| kens | I don't think I understand the question, so you'll probably have to wait until a MuPDF developer comes online | 09:32.52 |
| KunMyt | i using mupdf for read pdf file in android, i want to read file with bufferStream, can you help me? | 09:33.45 |
| kens | No, I have no idea what a bufferStream is. | 09:33.59 |
| KunMyt | the example of mupdf using URI file path to read it | 09:34.52 |
| kens | Still no idea what you mean | 09:35.17 |
| KunMyt | in mupdfActivities.class, i see it can read pdf file by byte[] in buffer | 09:35.52 |
| kens | As I understand it MuPDF includes curl in order to read from a server, but that's all I know. | 09:35.52 |
| tor8 | KunMyt: you'll need to read the bufferstream into a byte array first | 09:36.06 |
| | kens: the C api is more flexible | 09:36.40 |
| KunMyt | @tor8: I can't send byte array though Intent, why ? | 09:36.49 |
| tor8 | the java code in the android app only takes a filename or a buffer of bytes | 09:36.57 |
| kens | tor8 I was about to start on that when you interjected :-) | 09:36.58 |
| tor8 | KunMyt: but Robin_Watts or paulgardiner are really the ones to ask about android details | 09:37.33 |
| KunMyt | ohh, i can wait for them come back :D | 09:37.59 |
| kens | KunMyt : THe real API for MuPDF is a C API, the Java implementation is (as RObin repeatedly says) only a thin veneer which exposes just enough of the underlying implementation to implement the various apps we build. Its not been designed or built as a real functional API to the library, unlike the C interface. | 09:38.45 |
| tor8 | KunMyt: I guess you could extrend MuPDFActivity to do what you need, but really I'm out of my depth here, I never studied how android intents and activities work | 09:38.53 |
| KunMyt | @kens: no no | 09:39.26 |
| kens | One day Robin hopes to expose the C API through the Java interface, but that's a future project, and the developers are too busy to do this right at the moment | 09:39.34 |
| KunMyt | when them come back @tor8 | 09:40.28 |
| paulgardiner | I'm here, but I'm not sure I understand the question | 09:42.29 |
| KunMyt | ok, it is problem for android with mupdf | 09:42.59 |
| Robin_Watts | KunMyt: We've implemented just the 2 ways of working that our viewer needs. | 09:43.21 |
| paulgardiner | MuPDFCore has two constructors: one takes a file name, the other an array of bytes | 09:43.39 |
| Robin_Watts | If you want, you can extend mupdf.c and MuPDFCore.java to cope with another method. | 09:43.40 |
| KunMyt | i know, but i can't call it though Intent | 09:43.47 |
| | with FileName ok! | 09:44.11 |
| | with byte[] fail | 09:44.24 |
| Robin_Watts | Possibly that's a problem with the declarations in the AndroidManifest | 09:45.27 |
| paulgardiner | You could right your byte array to a temporary file | 09:45.32 |
| Robin_Watts | feel free to have a play. If you find a fix, let us know. | 09:45.45 |
| paulgardiner | s/right/write | 09:45.50 |
| KunMyt | i don't understant ... | 09:46.54 |
| kens | OK so if I read the logs correctly, Ray is fine with us creating an 'init' method for devices. I've kind of reached a point where I need one as well, so does nayone think the method needs any arguments other than a pointer to the device ? | 12:56.31 |
| chrisl | kens: if we are adding an init method, should we also take the opportunity to add the opposite (finit)? | 13:21.28 |
| kens | Thre's a finalise routine for devices already if htat's what you mean | 13:21.44 |
| | Or do you mean a separate method, nethier finalise nor close ? | 13:22.07 |
| chrisl | Finalize will be called when we're about to free the memory | 13:22.35 |
| kens | Yeah, and clsoe is called when the device is closed, hence my query. | 13:23.03 |
| chrisl | I guess that's probably sufficient | 13:23.09 |
| kens | OK. | 13:23.20 |
| chrisl | Remote Postscript debugging - oh, the joy...... :-( | 14:01.33 |
| kens | ? | 14:01.39 |
| chrisl | Customer with the problem with Arial CIDFont | 14:01.58 |
| kens | Oh yeah :-( | 14:02.05 |
| | I like the dumb question from Mathew Abin too, he hasn't (of course) set the page size policy, or set FitPages or anything. So what does he expect to happen ? | 14:03.22 |
| Robin_Watts | kens, chrisl: So OpenType fonts can contain a GSUB table that tells fonts to do substitutions on the glyphs before it renders them. | 14:03.27 |
| kens | I htink you have to do somethign to tell the font you want substitutions to take place, I forget what | 14:03.51 |
| | Oh and Mathew Abin hasn't actually attached the test file :-) | 14:04.15 |
| Robin_Watts | The problem I have reported here is that the pdf file supplied by the customer uses a subset font, and therefore doesn't have all the glyphs. | 14:04.17 |
| kens | Robin_Watts : that would seem likely to be a problem certainly | 14:04.33 |
| chrisl | Robin_Watts: in what type of file? | 14:04.54 |
| Robin_Watts | And the substitutions change 'g' for another glyph (presumably another appearance of 'g'), but that one isn't there. | 14:04.59 |
| | This is a PDF file. | 14:05.05 |
| | I'm thinking that possibly we shouldn't be doing glyph substitution in PDF files. | 14:05.21 |
| kens | I'm not sure about PDF files, but don't they support *TrueType* fonts ratehr than OpenType ? | 14:05.24 |
| chrisl | I'm not sure PDF uses the GSUB table..... | 14:05.35 |
| Robin_Watts | Certainly, gs and mupdf don't seem to. | 14:05.42 |
| kens | glyph substitution with a PDF file sounds like a really bad idea | 14:05.46 |
| Robin_Watts | chrisl: That is the conclusion I have come to. | 14:05.51 |
| | OK, so if you two are in agreement, I will look at disabling the use of the gsub stuff in PDF. | 14:06.17 |
| kens | I think that's absolutely the right thing to do | 14:06.28 |
| | Robin_Watts : the PDF reference seems to imply that PDF files are only expected to support version 1.0 of the TT spec, so that's *TrueType* fonts, not OpenType. | 14:07.33 |
| | Since OTF are backwardly compatible, its safe to include an OTF and use it as a TT, but we shold not use OTF features. | 14:07.57 |
| | OK I have to go see the vampies, back later | 14:09.53 |
| | vampires* | 14:10.00 |
| chrisl | Actually, according to the ISO PDF spec, PDF *does* support OpenType, but shouldn't rely on any optional tables in OpenType, and GSUB is an optional table, so...... | 14:10.35 |
| | Robin_Watts: either way, you should disable the GSUB table for PDF...... | 14:11.31 |
| henrys | meeting time | 14:31.38 |
| marcosw | morning | 14:32.07 |
| henrys | marcosw: saw a few bugs for xpswrite, when can we actually test with it? | 14:32.42 |
| | marcosw: we wanted to get mvrhel_laptop's changes tested | 14:33.03 |
| mvrhel_laptop | oh crap. I completely forgot to get chrisl the gsview installers | 14:33.54 |
| chrisl | Figured you were busy...... | 14:34.33 |
| mvrhel_laptop | I will carve out some time from sot to get it done | 14:35.04 |
| henrys | tor8: miles wants to do a mupdf android release but I've put it off until the London meeting... | 14:36.06 |
| | tor8: and how is epub coming? | 14:36.33 |
| Robin_Watts | henrys: We did a mupdf android release less than a month ago. | 14:37.09 |
| | but I assume you mean a chargeable one. | 14:37.23 |
| henrys | Robin_Watts: right | 14:37.27 |
| marcosw | sorry, afk. | 14:37.56 |
| henrys | chrisl: so it looks like kens is going to the device init stuff... | 14:38.33 |
| chrisl | henrys: er, yes? | 14:39.08 |
| marcosw | henrys: ghostxps locks up reading many of the xps produced by xpswrite. So I'm not sure what testing can be done at this point. | 14:39.34 |
| henrys | marcosw: are just doing the pdf subset? | 14:40.06 |
| | s/are/are we/ | 14:40.20 |
| marcosw | you the mean the pdf fts subset? I wasn't, but am not sure why it would be different. But I'll try it today. | 14:41.02 |
| henrys | marcosw: maybe it is one issue can you report a bug and I'll look at it. | 14:41.33 |
| Robin_Watts | marcosw: A list of the files that lock enough might be enough of a result for the first round of tests :) | 14:41.34 |
| chrisl | They probably don't lock up, just take a long, long time to run...... | 14:42.02 |
| henrys | right as long as they all lock up like the previous version it's tested ;-) | 14:42.24 |
| rayjj | sorry I'm a bit late. School field trip for my youngest threw the schedule off this AM | 14:42.24 |
| marcosw | already done: http://bugs.ghostscript.com/show_bug.cgi?id=695618 | 14:42.30 |
| | chrisl: probably. I was using 1 hour === forever | 14:43.04 |
| henrys | marcosw: I do think we want to stick with very simple fts files though. | 14:44.09 |
| marcosw | Robin_Watts: I thought about generating a list and adding it to the bug, but figured that would just people sad. | 14:44.27 |
| rayjj | kens: on the device 'init' call, I would include *both* the static device (srcdev) and the device being initialized (destdev) | 14:44.29 |
| mvrhel_laptop | yes just the fts pdf files for now | 14:44.38 |
| marcosw | henrys: I could limit the test to files that run in a reasonable amount of time (<1 minute ?) and just not test the rest. | 14:44.55 |
| mvrhel_laptop | the time outs are one of the issues that I do want to get fixed | 14:45.19 |
| | they are killing any reasonable printing for gsview | 14:45.33 |
| | but just working with the fts files makes a reasonable subset for now | 14:46.03 |
| henrys | marcosw: that's reasonable if there are a good number passing... What is it like 50% or 1%? | 14:46.14 |
| marcosw | I'll run the PDF_1.7_FTS files, making a list of any that don't complete in a reasonable time and running the comparison to direct to bitmap for the test. | 14:46.17 |
| mvrhel_laptop | that would be great | 14:46.26 |
| henrys | marcosw: oh and are you using -dNOCACHE? | 14:46.51 |
| marcosw | henrys: I'm thinking 20-30%, I didn't test many files after I noticed that realized that most were not finishing. | 14:47.17 |
| | henrys: no, I'm assuming from you question I should be. | 14:47.32 |
| henrys | marcosw: definitely yes | 14:47.44 |
| marcosw | perhaps I'm confused, but isn't a cache typically used to speed up a program (I'm old, so forgive me if you kids have changed the meaning)? | 14:49.04 |
| henrys | rayjj: is there anything you want miles to pester company C or M about today? I have a meeting with him after this one. | 14:49.07 |
| rayjj | xps is a pig even with mupdf -- that's why I didn't include XPS in any of the company M timings. Memory use and CPU load are *much* worse than for the original PDF (xps created by MS Document Writer) | 14:49.27 |
| chrisl | marcosw: disabling the cache means xpswrite gets glyph outlines rather than bitmaps | 14:50.05 |
| henrys | marcosw: with cache the fonts are rendered as rectangles from rasterized images instead of outlines. or what chrisl said ;-) | 14:50.28 |
| rayjj | henrys: I'd sort of given up (for the meantime) on company C, and I'm waiting for a response from company M to the timings I sent. Please ask Miles if I should dive back into company C work | 14:51.08 |
| marcosw | chrisl: <sarcasm>when I think about it I can fully understand why that option would be called NOCACHE</sarcasm> | 14:51.31 |
| henrys | rayjj: that's the preferred platform so yes he does want you to work on it. I assumed you couldn't without help. Can you do it without help? | 14:52.34 |
| chrisl | marcosw: we discussed the crapulence of options called "NO...." a while back - but we're stuck with it :-( | 14:52.45 |
| Robin_Watts | GLYPHSASOUTLINES would have been too clear. | 14:53.12 |
| chrisl | I think, DISABLEGLYPHCACHE would have been my choice | 14:53.41 |
| henrys | NOGLYPHSASOUTLINES=false | 14:54.02 |
| | anyway meeting over, I thought we wouldn't have anything today but somehow we managed to fill the 1/2 hour. | 14:55.09 |
| rayjj | henrys: the lack of source and documentation for the networking module make it quite difficult, but I can work on just getting gs/pcl6 built and timed from a kloodge debugmoniitor method | 14:55.13 |
| henrys | anything else for the meeting? | 14:55.55 |
| marcosw | chrisl: just don't get me started on the ATS, there is a VM called ldap but it doesn't provide LDAP, another VM, homeserver, which is not the home directory server (no idea what it does, but the system falls over with a loud thud if it's not running), the server intranet2, is not the server for the intranet home page, etc. | 14:57.12 |
| chrisl | marcosw: clearly a system created by systems admins! :-( | 14:58.14 |
| rayjj | henrys: I'll put together what I think I need most. Note also that either I need to get a license for the ARM-DS toolchain. I never got the quote I requested from ARM. | 14:59.38 |
| | henrys: but since then I've found that newark sells a linux node locked license for $3,300 (which is less than the $7,500 I've seen mentioned from ARM) | 15:00.27 |
| Robin_Watts | rayjj: Will one of the free codesourcery toolchains not do what you want? | 15:01.33 |
| rayjj | henrys: if I don't get the toolchain, I'll have to spend an unknown amount of time trying to get things working with another compiler (codesourcery or the linux gcc arm toolchain) | 15:01.47 |
| | Robin_Watts: all of their build stuff seems locked to the ARM-DS tools, and I don't know how hard it will be to switch over. | 15:02.42 |
| | Robin_Watts: for one, will the binary only libraries work with the "free" tools ? | 15:03.19 |
| Robin_Watts | rayjj: That is the question. But I wouldn't imagine it would take a huge amount of time to try calling libtool on them or something and seeing if it barfs. | 15:04.05 |
| rayjj | Robin_Watts: 90% of the stuff is *not* source as it was with company M (linux based) code | 15:04.09 |
| henrys | rayjj: don't you live right next to them, did you try pounding on the door ;-) | 15:04.53 |
| jogux | rayjj: does it need to be ARM-DS? artifex/emobix both have licenses for older ARM toolsets. | 15:07.49 |
| | I can't keep up with their renamings though | 15:08.07 |
| rayjj | henrys: They moved out of So Cal (the original SOC came out of Boston). I think all of their developers have gone or been "outsourced" to .in or somewhere. | 15:08.07 |
| | jogux: hmm... I didn't realize that Artifex may have gotten licenses in the acquisition. How do I access those ? | 15:09.40 |
| jogux | rayjj: ah. hm. good question. | 15:10.20 |
| Robin_Watts | rayjj: They are on points in /mnt/something | 15:10.51 |
| rayjj | jogux: there are tests in the build process to verify compiler/toolchain version. In order to build the examples, I had to edit the list in the test proc to use the newer toolchain (free trial has expired) | 15:12.05 |
| | Robin_Watts: thanks -- I'll go fishing | 15:12.30 |
| jogux | rayjj: ah. sounds painful :( | 15:13.07 |
| | rayjj: I think we have RVDS 4.1. Don't imagine Picsel had anything newer than that. | 15:15.32 |
| | we == emobix in this case :) | 15:16.02 |
| Robin_Watts | Picsel has RVDS 4.1 on points, yes. | 15:16.21 |
| | but I wouldn't like to bet where the license details are. | 15:26.14 |
| rayjj | Robin_Watts: I found one in /dataserver/central/transfer/alexms | 15:26.29 |
| | Robin_Watts: arm-rvds3.1-nl-PIC6368.lic | 15:26.42 |
| Robin_Watts | <StatingTheObvious>That's a key for 3.1 not 4.1</StatingTheObvious> | 15:27.09 |
| rayjj | Robin_Watts: still looking. That was just the first one I found | 15:27.32 |
| | hmm... don't see any others | 15:27.49 |
| | I see a bunch of .lic files for other things (GoodShareServer.lic) | 15:28.44 |
| | and some that appear to be for Intel VTune, but in directories named "expired", so probably not of much use :-) | 15:29.50 |
| | one for VTune 7.0 that may not be | 15:30.21 |
| | in dataserver/resources/dev_tools/Intel VTune 7.0 | 15:30.38 |
| | and another in dataserver/resources/dev_tools/WinCE/Intel XScale C++ compiler | 15:31.29 |
| | I guess I'll play with another toolchain first until the brick wall gets to me | 15:32.30 |
| jogux | the keys will likely be node locked anyway, so the tricky will be finding the login details for the arm website. | 15:33.22 |
| | or at least the email address and doing password recovery. | 15:33.32 |
| Robin_Watts | rayjj: Some of the ARM things require a license manager (flex lm, IIRC), to issue floating licenses. That would be another layer of complexity. | 15:34.00 |
| kens | is back, drained..... | 15:47.00 |
| | Oh great, so his software doesn't have a problem on a different Windows machine. Possibly the one he's running on just doesn't have the fonts installed ? | 15:51.04 |
| chrisl | Then 9.10 would fail, too, just later on, and in a non-recoverable way..... | 15:57.14 |
| kens | I guess that's true. | 15:57.25 |
| chrisl | Can you check in your \windows\fonts directory for a font file called arialuni.ttf, please? | 15:58.03 |
| kens | Not there. | 15:58.42 |
| | I have 4 variants of arial, but not that one | 15:58.55 |
| jogux | chris> I have Arial Unicode.ttf (I'm on a mac) | 15:59.33 |
| chrisl | jogux: this is very much a windows problem :-( | 15:59.54 |
| kens | actualy might be 5 variants, I'm not sure what ariblk.ttf is | 16:00.25 |
| Robin_Watts | I have arialuni.ttf on windows. | 16:00.28 |
| | windows 7. | 16:00.32 |
| kens | I'm also on Windows 7 | 16:00.44 |
| Robin_Watts | I have 10 arial variants. | 16:00.48 |
| chrisl | Huh, that's weird - it's not on any of my Windows7 machines..... | 16:00.59 |
| kens | OK I have 5 that last one is Arial black | 16:01.02 |
| chrisl | kens: The output from the PS I sent them shows a mapping for arialuni.ttf (and many other font files) but none of the "normal" arial family mappings are there - which suggests gs couldn't stat the "normal" arial TTFs during initialisation | 16:01.54 |
| kens | That's really strange. | 16:02.10 |
| chrisl | Robin_Watts: how big is arialuni.ttf? | 16:02.29 |
| kens | My laptop (also WIndows 7) also only has 5 arial variants, and not arialuni.ttf | 16:02.58 |
| chrisl | How can I find what variant of Win7 I have? | 16:03.42 |
| kens | control panel-> System and security->system | 16:04.18 |
| jogux | wonders if there's a correlation with which of you have installed office / office365? | 16:04.18 |
| kens | Jogux, that could well be | 16:04.32 |
| | I *don't* have a copy of that | 16:04.42 |
| | chrisl both my copies are windows 7 professional | 16:05.01 |
| chrisl | I have Enterprise | 16:05.14 |
| kens | chrisl: | 16:05.20 |
| | http://www.microsoft.com/typography/fonts/font.aspx?FMID=1081 | 16:05.20 |
| | that lists the products supplying the font | 16:05.28 |
| | WIndows is not one of them | 16:05.41 |
| | Various flavours of office however do | 16:05.52 |
| jogux | weird that office 2013 doesn't. | 16:06.05 |
| kens | Well the page may be older than that.... | 16:06.16 |
| chrisl | Okay, well, that clears it up, but doesn't really help much :-( | 16:06.18 |
| Robin_Watts | 23275812 | 16:06.33 |
| | I have office 365 installed here. | 16:06.45 |
| kens | yeah not surprised, it lists pretty comprehensive coverage of Unicode values | 16:06.52 |
| chrisl | Robin_Watts: Hmm, I was going to ask if you could stick a copy up on casper...... | 16:07.34 |
| Robin_Watts | chrisl: That'd be illegal. | 16:07.47 |
| chrisl | Er, um, so it would..... | 16:08.05 |
| kens | Bad Chrisl.... | 16:08.32 |
| chrisl | Robin_Watts has set me straight - I shall prey to the great M$ for enlightenment | 16:09.08 |
| | So, Windows sort of supports symlinks now, they can't be created in Explorer, but they can at the command prompt. And it supports Windows ShortCuts, which can only be created from Explorer, and not from the command prompt - erm, doh!? | 16:13.37 |
| kens | What, you wanted consistency ? You should load WIndows 8..... | 16:14.01 |
| chrisl | And then Explorer doesn't let you see actual font files in the \windows\fonts directory...... <sigh> | 16:14.57 |
| rayjj | chrisl_away: Explorer _does_ let you see the font file -- you have to right-click and look at the properties, or select the "Details" view and then right-click on the header pane and click (check) "Font file names" | 18:46.03 |
| mvrhel_laptop | off to run an errand. | 20:29.43 |
| | bbiaw | 20:29.45 |
| | Forward 1 day (to 2014/10/22)>>> | |
IRC Logs