| <<<Back 1 day (to 2014/07/27) | 2014/07/28 |
kens | Hmm, another complaint about broken GS downloads.... | 07:02.58 |
chrisl | I guess I might need to brave godaddy support :-( | 07:48.16 |
kens | For the logs; anyone got access to an HP printer which accepts PostScript as an input ? | 12:34.51 |
chrisl | kens: I think tkamppeter has at least occasional access to one | 12:36.04 |
kens | I was hoping someone at Artifex would have one and could get back channel communication out of it... | 12:36.37 |
chrisl | None of the HPs I've dealt with have a backchannel - at least, not accessible to anyone but an HP service rep | 12:37.21 |
kens | :-( | 12:37.35 |
chrisl | If none of us do, I can ask around the Hants Linux user group | 12:38.03 |
jogux | I have a hp inkjet from their business line, I presume it doesn't accept postscript but I really have no idea :) | 12:38.15 |
kens | Most Hp printers don't sady, or only as an optional add-on | 12:38.49 |
jogux | http://www.amazon.co.uk/HP-Officejet-37ppm-colour-4800dpi/dp/B000BRXTSW claims it does. not sure I believe that. | 12:39.26 |
chrisl | The Deskjets and their derivatives almost all use a customised (bastardised) PCL3 | 12:40.28 |
kens | It says PCL, I don't see PostScript on there, maybe I'm looking in the wrong place | 12:40.32 |
pedro_mac | kens> sorry, I have a(very) old hp3500 but only PCL (via JetDirect) | 12:41.04 |
kens | Yeah that's normal | 12:42.09 |
| http://www.hp.com refuses to admit they ever made a K550 :-) | 12:42.28 |
chrisl | According to HP, Officejet Pro K550 is "HP PCL 3 enhanced" | 12:42.28 |
| http://h10025.www1.hp.com/ewfrf/wc/document?docname=c00488708&tmp_task=prodinfoCategory&cc=us&dlc=en&lc=en&product=501369 | 12:42.41 |
kens | PCL3 >? <boggles> | 12:42.41 |
chrisl | kens: like I said, most of the Deskjet, BusinessJet etc use various hacks on PCL3 | 12:43.23 |
kens | OK found hte data sheet, its says PCL3 there | 12:43.26 |
chrisl | kens: and don't forget, different HP printers use different "versions" of their PS, so whatever you need to debug, no guarantee another printer model will exhibit the problem | 12:45.30 |
kens | Yeah I know. I've suggested to Todd he goes and dumps the problem in HP's lap, I can't make head or tail of it | 12:46.05 |
chrisl | Is this the filter one? | 12:46.42 |
kens | yeah SubFileDecode, he replied to my mail after I sent another one prompting to see if he'd received the first (which he had, and forgotten about) | 12:47.28 |
| The error clearly occurs when creating a font, but the same routine is used to execute the page description, and (if I remove teh font) that works correctly. | 12:48.37 |
chrisl | I take it you've tried eliminating the subfiledecode filter? | 12:48.48 |
kens | I'm not sure I can | 12:49.06 |
chrisl | Can't you just pull out the font from the ps2write output? | 12:49.33 |
kens | Fonts aren't a single thing in ps2write output, but I can extract (or remove) the FontFile object which defines the font, certainly | 12:50.11 |
chrisl | Is it a TTF? | 12:50.24 |
kens | Removing the FonFile object makes the problem go away | 12:50.28 |
| No, its a type 1 font | 12:51.03 |
chrisl | Okay, so the type 1 stream should be a complete definition | 12:51.32 |
kens | More specifically its Nimbus Sans L but called Helvetica-Bold | 12:51.34 |
| chrisl yes, the stream its executing is a compete font | 12:51.49 |
tkamppeter | chrisl, I have some HP PS printers, and they do backchannel. CUIPS is able to ask the printer for the default option settings following the appropriate entries in the manufacturer-supplied PPD files. | 12:52.32 |
chrisl | kens: So, you can pull that out, and send it to the printer with and without the subfiledecode | 12:52.50 |
| tkamppeter: I thought when we tried before you couldn't get the backchannel - maybe that was a specific model.... | 12:53.18 |
kens | chrisl yes.... But that's not quite the same as how ps2wsrite executes the font (I think) | 12:53.24 |
| I think I'll put this aside for a few days, I'm in the middle of something else anyway | 12:54.54 |
chrisl | kens: and, IIRC, ps2write defines the font, then does something with the encoding? That might cause a problem | 12:55.11 |
tkamppeter | chrisl, probably it was really a special mode, I do not remember. I have a Color LaserJet CM3530 and polling the option defaults from that printer is no problem. | 12:55.40 |
kens | chrisl I don't believe it gets that far, it seems to be the execution of the fontfile which causes the problem (of course, no back-channel, so its just guesswork) | 12:55.56 |
chrisl | kens: well, as I say, my next step would be pulling the font stream out of the ps2write output, and trying it with and without the subfiledecode | 12:57.42 |
kens | chrisl I guess, I may try it later, for now I'm lost in charproc capture. | 12:58.09 |
| Given it took Todd over a week to reply, and only because I pinged him, I guess ti can't be urgent | 12:58.41 |
chrisl | kens: that's true. One thing that does spring to mind: eexec may be implemented as a filter, or implemented using a filter, and it HP's was one of the implementations we had trouble with multiple filter combinations | 13:00.06 |
kens | I tried it with the font uncompressed, but yeah you could be correct, that would show up with a SubFileDecode I guess | 13:01.58 |
| But I don't think there's any way we can address that, we rely on SubFileDecode | 13:02.14 |
| I suppose I could decrypt the font :-) | 13:02.30 |
chrisl | I was about to say that! More accurately, we not not encrypt the font.... | 13:02.57 |
| But these are just speculations to test and eliminate - no sense trying to come up with solutions until we actually know the problem. | 13:03.41 |
| Even if *we* can't fix it, the more we can narrow the problem, the better report we can send to HP | 13:04.28 |
kens | I meant I could decrypt the font for testing purposes, if it turns out to be the SubFIleDecode | 13:11.58 |
Ilzar | Hello! Could you tell, please, can I somehow reduce the size of the library mupdf for Android? It weighs about 5 MB in a zip-compressed. | 13:52.59 |
kens | You cna remove CMap files, ifyou don;t require CJKV support | 13:54.23 |
| What CPU are you building for ? | 13:54.49 |
tor8 | Ilzar: build with -DNOCJK | 13:54.49 |
| Ilzar: that will drop all the Cmaps and the CJK substitute font from the build, but then you won't be able to open PDF files with chinese, japanese or korean text | 13:55.43 |
Ilzar | I`m using arm-v7. | 13:56.47 |
| I'll try with -DNOCJK, thank you | 13:57.02 |
chrisl | tor8: can you build with the CMaps but without the substitute CJK font? | 13:57.41 |
tor8 | chrisl: yes, with -DNOCJKFONT | 13:57.57 |
| and there is a smaller CJK font (saves about 1M, but misses out on some more uncommon characters) with -DNOCJKFULL | 13:58.28 |
chrisl | I'd probably start with -DNOCJKFONT - there's no *requirement* for a substitute CIDFont | 13:58.59 |
tor8 | I am considering adding the source han sans font as an option if you really don't care about size (it adds 10M more) but gets you language specific glyphs and even better coverage | 13:59.22 |
| chrisl: yeah, that's probably a better recommendation. a proper CJK pdf will have embedded the font, but may not have embedded the CMaps | 14:00.16 |
chrisl | tor8: I think we should encourage people to embed (CID)Fonts, not give them more and more reasons not to | 14:00.27 |
tor8 | chrisl: in an ideal world... | 14:01.38 |
| in reality, we have to live with crap like the S22PDF creator produces | 14:01.56 |
| CJK text masquerading as WinANSI | 14:02.14 |
chrisl | tor8: the trouble is, there will *always* be cases when a random substitute won't work, and the more effort we make to handle them, the harder it becomes to explain the ones that don't | 14:04.05 |
tor8 | yes, there is that. | 14:06.39 |
Ilzar | -DNOCJK is really helpful. It is 1.5Mb now. Thank you! Maybe there is any kind of flags, which can reduce size? | 14:08.57 |
kens | chrisl I sent a couple of test files to todd, My eexec decrupting program doesn't run under WIndws 7 (curse you Microsoft!) So I'll have to rewrite that if it looks like I need it. | 14:13.20 |
chrisl | kens: I think you might be able to use GS - can't remember for sure | 14:14.20 |
kens | I thought I probably could, but I figured I'd send off the test I had ready and see what comes back. A simple SubFileDecode might not fail of course :-( | 14:15.03 |
Ilzar | Is there any flag to switch off js-executing? | 14:37.59 |
kens | chrisl just the font with SubFileDecode works :-( | 15:07.52 |
chrisl | kens: Oh :-( | 15:08.31 |
kens | That is' still with eexec encryption. So its somtehign weird with teh way the HP printer is dealing with the font, I'll have to start mimic'ing more of the way ps2write handles fonts, which is going to be painful | 15:08.53 |
chrisl | kens: you mentioned changing the font name - did you try that? | 15:09.30 |
kens | Yes, that failed too | 15:09.42 |
| with an invalidaccess on SubFileDecode | 15:09.53 |
chrisl | kens: well, the only other option is to pepper the font dict with showpages, and have him count how many pages get emitted | 15:10.51 |
kens | chrisl I'm not certain its the font that's the problem, the error occurs when executing 'filter' on the FontFile object. | 15:11.27 |
| That's before it actually executes the FontFile stream | 15:12.16 |
chrisl | It's a strange error from filter | 15:13.20 |
kens | Yeah I know :-( | 15:13.29 |
| considering the only arguments are currentfile an integer and an empty striong | 15:14.09 |
chrisl | local v global maybe? | 15:14.50 |
kens | Well only currentfile and the string can have a VM type and there was something in filter about being local if all the objects were local and global if any one wasn't, I'll just go look at the PLRM again | 15:15.43 |
| Yeah it says filter disregards the current VM mode (!) | 15:16.34 |
chrisl | Hmm, but that doesn't mean that the HP does that | 15:16.48 |
kens | Well true. | 15:16.55 |
chrisl | In particular, I could see currentfile being global | 15:17.13 |
kens | But still, what I'm doing in the tets file is basically the same as ps2write | 15:17.19 |
| If it is, and its a problem, then my test file woudl fail, it also uses currentfile | 15:17.38 |
| THis would all be a lot easier with a relevant printer and a back channel :-( | 15:18.13 |
chrisl | Doesn't ps2write fiddle with the VM mode? | 15:18.26 |
kens | I didn't think it did, but let me check | 15:18.36 |
| It does, but it puts it back, I'd have to dig further to see if fonts are afected | 15:19.39 |
| Looks like it probably is affected | 15:20.45 |
| I suppose I can try 2 files, one with a locally allocated string and one in global VM | 15:21.25 |
| Obviously I can't affect the allocation mode of currentfile | 15:21.41 |
chrisl | You could have it print the currentvm, currentfile's VM etc before it calls filter | 15:22.35 |
kens | Yes, kind of tedious though | 15:22.48 |
| And ExecuteStream gets called for other reasons in the course of the file :-( | 15:23.11 |
chrisl | I meant specifically for the problem stream | 15:23.26 |
kens | I can't do it for just the problem stream, its all or nothing | 15:23.53 |
| I suppose I coudl set a flag, bit icky though | 15:24.09 |
chrisl | That's what I've had to do in the past | 15:24.19 |
kens | I have had a thought. I think it executes the stream while its already in a stream context, I should check that. | 15:24.51 |
| Because that would mean its executing a SubFileDecode while its inside a SubFileDecode, possibly the HP doesn't like that | 15:25.22 |
chrisl | Could be. Aren't command line options to prevent that? | 15:25.56 |
kens | I'd be doubtful. I'm not even completely sure that's true, I need to check | 15:26.23 |
| THe ps2write output is terribly confusing | 15:26.43 |
chrisl | You're not wrong | 15:27.06 |
kens | Oh, someone else complaining about PCL performance | 15:27.25 |
kens | suspects the correct answer will be 'use the current version' | 15:27.48 |
| Yeah, lets send a 17Mb uncompressed file by email..... | 15:29.38 |
| 1000 pages of identical dense dumb text. | 15:30.56 |
chrisl | Leave it to Marcos, it's almost certainly much better in the current code | 15:32.12 |
kens | Its reasonably quick converting to PDF | 15:32.24 |
| Certainly not 10 minutes | 15:32.33 |
chrisl | IIRC, they may be using UFST - which would make it even worse | 15:32.55 |
kens | 29 seconds to convert to PDF on my PC | 15:33.20 |
chrisl | Have to go - bye all | 15:37.45 |
kens | bb | 15:37.52 |
| Goodnight all | 16:18.39 |
| Forward 1 day (to 2014/07/29)>>> | |