| <<<Back 1 day (to 2014/02/04) | 2014/02/05 |
marcosw_ | I'll be updating the bugzilla server tonight, so if it's not working tomorrow you'll know who to blame :-) | 02:16.29 |
| The update to picas is complete and bugzilla appears to still work, please report anything odd or broken. | 05:50.14 |
| chrisl_away: for the logs: you said the the other day that you changed the repository from ufst-6.2 to ufst-6.x, but the weekly regression tests are using the ufst repository. Should I switch them to ufst-6.x? | 06:23.05 |
chrisl_away | marcosw_: yes, as of a few weeks ago, both our ufst customers are using 6.3, so it would be preferable to test with the ufst-6.x code. | 07:52.28 |
janmoesen | how can i create a two-page spread (left page and right page next to each other [_|_]) from a "normal" PDF? i.e., i want to extract page 2 and 3 into a new PDF so that they are next to each other instead of page 3 below page 2 | 08:03.20 |
| this is what i have now: gs -dQUIET -sDEVICE=pdfwrite -dPDFSETTINGS=/prepress -dFirstPage=2 -dLastPage=3 -o new.pdf old.pdf | 08:03.58 |
kens | The simplest solution is probably to use a page layout tool, or a tool designed to do ths | 08:04.01 |
| While you can do it with Ghostscript you will need to write some PostScript, there is no built-in method | 08:04.22 |
janmoesen | to clarify: it needs to be on the command line/in a script | 08:04.24 |
| right | 08:04.29 |
| any pointers? i checked the faq on ghostscript.com and the linked faq on wikibooks | 08:04.45 |
marcosw | chrisl_away: okay, I'll make the switch. | 08:04.49 |
kens | You need to pre-empt the built-in PDF interpreter so that it soen't execute a 'showpage' for each page in teh document | 08:04.52 |
| You need to set the current transformation matrix for each page, firstly to scale appropriately, and secondly to position the sexcond and subsequent pages in the signature | 08:05.51 |
| WHen the signature is competed you then need to execute the showpage | 08:06.10 |
| Unfortunately, at the moment all of this is hidden away inside the PDF interpreter. | 08:06.25 |
| Essentially you need to call some of the routines built into that (PostScript) code in a way that wasn't originally intended | 08:06.57 |
janmoesen | excuse my reaction, but yikes :-) | 08:07.03 |
kens | If you look in gs/Resource/Init/pdf_main.ps you wil find the code there | 08:07.18 |
| and yes, yikes, is a reasonable reaction | 08:07.27 |
janmoesen | thanks, going to have a look | 08:07.30 |
kens | As I said this isn't something that it was originally written to do, and adding it on afterwards is a difficult retrofit. | 08:08.02 |
| janmoesen : I have some 'similar' code which you are welcome to if you want it. WHat it does is the reverse of what you want. It takes an 'n-up' page and splits it into a number of individual pages | 08:09.13 |
| It is moderately well commented, it may give you some clues as to how to do what you want. | 08:09.58 |
janmoesen | kens: that could be useful, yes; for the splitting I just use `gs` with FirstPage equal to LastPage | 08:10.33 |
kens | janmoesen : yes its a different problem, in this case the code slices up a page into 'n' sub pages, the exact inverse of what you need. | 08:11.05 |
| The code is too large ot put hre, I'll stick it on pastebin, just a minute | 08:11.23 |
janmoesen | cheers | 08:11.27 |
| (git is cloning the repo in the meanwhile; seems large enough, too) | 08:11.42 |
kens | GS is fairly large, especially since it will include all the 3rd party libraries (FreeType, openJPEG, zlib...) | 08:12.12 |
janmoesen | understood | 08:13.16 |
kens | http://pastebin.com/taaLxywf | 08:13.31 |
janmoesen | i was viewing the pdf_main file through gitweb, but i thought syntax highlighting would be helpful, so there | 08:13.42 |
kens | Hmm is PostScript, I'm not sure there is a syntax highlighter for PostScript | 08:14.04 |
janmoesen | vim seems to handle it just fine | 08:15.09 |
| thank you for the script; diving in right now | 08:15.16 |
| (ah, no; vim thinks filetype=grads) | 08:15.40 |
kens | In that pastebin; runpdfbegin starts oiff interpretation of a 'file' object, treating it as a PDF file, pdfpagecount retrieves the number of pages from that file (after initialisation).pdfgetpage renders the page and pdfshowpage finishes drawing it | 08:15.57 |
| I suspect you can't use pdfshowpage, because it will actually draw the page, so you will have to look at the definition of that routine in pdf_main.ps and do all the things it does *except* the showpage | 08:17.10 |
| Then you use 'translate' to move the origin, draw the next page, and then call 'showpage' to actually render the combined pages | 08:17.48 |
janmoesen | i'm with you until now on a general level, but i might show up again with more questions once i have tried to tweak the logic | 08:18.44 |
kens | I'm sure yo uwill, this is not at all obvious. | 08:18.57 |
janmoesen | it's quite a "different" language from what i am used to | 08:18.58 |
kens | Its like FOrth if that helps | 08:19.09 |
janmoesen | (note to self: .ps, not .gs for vim highlighting/ft=postscr, you putz) | 08:19.21 |
kens | Its stack-based so it uses Rverese Polish Notation | 08:19.24 |
| I'll be out between ~10:15 and 1pm (ie leaving in about 2 hours) | 08:20.07 |
janmoesen | only in telling me it has quite a bit of history :-) | 08:20.16 |
kens | But the channel is logged feel free to post messages and I'll read them when I get back | 08:20.25 |
janmoesen | no worries, you've been really helpful so far | 08:20.26 |
kens | thinks I should really bite the bullet and *write* at least a 2-up imposition for GS | 08:21.14 |
janmoesen | kens: thank you for your help, but it seems the PDF i am using, is not OK: http://pastebin.com/DLHi4bvZ | 08:59.55 |
| "File did not complete the page properly and may be damaged." | 09:00.11 |
kens | That's not too encouraging certainly | 09:00.20 |
janmoesen | fwiw, i added the -dFirstPage and -dLastPage to see if those made any difference | 09:00.30 |
kens | It could be a bug, the current release is 9.10. But I'd have to see teh PDF file to tell | 09:00.35 |
janmoesen | (they didn't) | 09:00.35 |
| i can upload it if you have the time to look at it --- but please don't feel obligated | 09:01.07 |
kens | give me a URL and I'll take a quick look | 09:01.21 |
| Hmm tha'ts using pdf_slice.ps ? Well that could be the problem of course. | 09:02.02 |
| I wrote that program a few years ago for somene else with a specific need that interested me, its possible it no longer works | 09:02.40 |
| By the way, FirstPage and LastPage won't work with that program, it subverts the PDF interpreter | 09:03.53 |
janmoesen | i figured; the output stayed the same | 09:04.06 |
| PDF: http://jan.moesen.nu/iepot.tmp/old.pdf | 09:04.08 |
kens | I get a forbidden page on that URL | 09:04.23 |
janmoesen | go me! | 09:04.27 |
| should be ok now | 09:04.52 |
| sorry | 09:04.53 |
kens | yep hgetting it now | 09:04.58 |
| Well the PDF is OK | 09:06.45 |
| So its the usage of pdf_slice.ps that isn't | 09:07.08 |
janmoesen | right | 09:07.55 |
| thanks for looking | 09:07.58 |
kens | OK I get the error, but the task completes OK | 09:09.02 |
| Probably the error is irrelevant | 09:09.19 |
| Abusing the PDF interpreter is probably just confusing it in some fashion | 09:09.51 |
| If you try running it to some kind of output you'll see that it looks just fine | 09:10.38 |
janmoesen | i get a prompt that says "GS<6>" | 09:14.41 |
kens | You need to set -dBATCH to exit on completion | 09:15.04 |
| THat's just taken you back to the interacvtive prompt | 09:15.16 |
| type quit | 09:15.22 |
janmoesen | that's what i did, but after "quit" i still did not see any new files --- was i supposed to? | 09:15.47 |
kens | Depends what you set the output to | 09:15.58 |
| I left it at the default, which is the display device (naturally, don't set NODISPLAY for this) and saw the output | 09:16.33 |
janmoesen | heh, i had forgotten to re-add -onew.pdf | 09:16.34 |
kens | So you have -sDEV_CEpdfwrite ? | 09:16.50 |
| -o implies -dBATCH by the way | 09:17.02 |
janmoesen | i'm running headless, so no display; i'll set it to the pdfwrite device | 09:17.08 |
kens | -sOutputFile doesn't | 09:17.11 |
janmoesen | ah, i did not know that | 09:17.22 |
kens | Well you can use tiff or pbm or whatever... | 09:17.26 |
janmoesen | that, or pretty much anything else :-) | 09:17.29 |
kens | GS has *waaay* too many switches..... | 09:17.57 |
| Maybe Ray can emeber them all, I certainly can't | 09:18.23 |
janmoesen | to be honest, i usually let imagemagick handle all the GS work for me, but it was too slow on biggish PDFs when you only want to process certain pages | 09:18.50 |
| it executes gs -sDEVICE=pdfwrite and what have you | 09:19.28 |
kens | Yes, I think I've mentioned on Stack Overflow how to use GS instead when ytou only want some pages | 09:19.32 |
janmoesen | that's what got me started :-) | 09:19.46 |
| it's much faster, for sure | 09:19.52 |
kens | Ah, infected by PostScript, there's no escape now....... | 09:20.03 |
janmoesen | hopes are not to be gotten up in this case | 09:20.57 |
kens | :-) | 09:21.11 |
chrisl | Postscript is a great language - as long as you don't put too much value on little things like your sanity! | 09:22.11 |
kens | Or any remaining hair | 09:22.23 |
janmoesen | :-) | 09:22.38 |
chrisl | It has the advantage of being a very powerful language - it has the disadvantage of "enough rope to hang yourself"..... | 09:23.58 |
janmoesen | chrisl: a truly lovely quote, that | 09:25.10 |
| kens: baldness is hereditary, so i guess i can't blame it on a few days of PDF'ing and Postscripting --- but thanks for the heads up | 09:26.10 |
| heads-up for the n-up | 09:26.19 |
kens | Oddly the two staff members with reduced hair don't (much) work on PostScript. Perhaps PDF is even worse | 09:26.46 |
| dir out.pdf | 09:26.57 |
| oops ;-) | 09:27.02 |
| Using pdfwrite got me a decent output PDF file using pdf_slice.ps | 09:27.29 |
| Of course I now have 208 pages..... | 09:27.59 |
janmoesen | yup, that works now | 09:34.03 |
| i think i am going to leave it at that; most pdf viewers support viewing pages side by side, and printers can print two pages on one sheet, two --- if i convert the separate pages to jpeg, i can more easily combine those two in a 2-up way | 09:37.25 |
| (look at me being a quitter!) | 09:37.39 |
kens | :-) | 09:39.45 |
| Maybe I'lltry and do the simple n-up cases as an exercise | 09:40.03 |
janmoesen | i can't justify spending much more of my customer's time on this nice-to-have feature | 09:40.05 |
kens | Keep an eye out and I'll say if I get anywhere, won't be today though I have an overnight delugeto deal with | 09:40.38 |
janmoesen | i'll stick around for a bit to see how that goes --- but again, don't feel obligated | 09:40.44 |
kens | People do ask, so it would be nice to have a demo of sorts | 09:40.58 |
janmoesen | true | 09:41.08 |
| you seem like a right old bunch of Good People --- thanks for all the help, and good luck with the incoming deluge | 09:41.43 |
Robin_Watts | paulgardiner: Morning. | 10:41.52 |
paulgardiner | Robin_Watts: hi | 10:42.02 |
Robin_Watts | I'm having weird things here. | 10:42.03 |
paulgardiner | Oo sounds good. Can I have some too | 10:42.17 |
Robin_Watts | I've removed the /DEBUG from the last link of the test shell, and now the exe doesn't do anything. | 10:42.46 |
| If I copy the exe out of its dir and run it, I get a warning about a missing MSVCR80.dll | 10:43.14 |
paulgardiner | Does /DEBUG control which dlls are linked or do you have to specify those explicitly? | 10:44.07 |
Robin_Watts | I assume /DEBUG controls that. | 10:44.53 |
| BUT... I have that DLL on my machine, I'm sure. | 10:45.07 |
| Let me try a rebuild. | 10:45.13 |
| tor8: I like your idea about doing array construction locally. | 11:52.02 |
| I will look at that when I get back to mupdf. | 11:52.14 |
tor8 | Robin_Watts: great! | 11:52.33 |
| Robin_Watts: I will soon pester you (or paul) about how to run the javascript stuff in mupdf | 11:53.10 |
| you have a mujstest thing that runs javascript on the command line without the android app right? | 11:53.31 |
Robin_Watts | I do. | 11:54.08 |
| You know how the windows and linux viewers are just a few platform specific classes glommed onto the side of pdfapp? | 11:55.04 |
| mujs test is just another one of them. | 11:55.20 |
tor8 | I fear to look at pdfapp.c these days :) | 11:57.17 |
paulgardiner | Don't say pdfapp.c! Send shivers down my spine. | 12:08.17 |
Robin_Watts | paulgardiner: tgvsimplepdfapp | 12:09.31 |
paulgardiner | Are you trying to push me to insanity?! | 12:10.08 |
| Actually, pdfapp.c really isn't that horrendous | 12:10.36 |
| At least it doesn't have a key binding hard wired to "goto page 3" as unidis-test seems to. :-) | 12:12.34 |
norbertj | chrisl_away you here? | 13:21.16 |
kens | Not at the moment norbertj but he will be bck fairly soon, cna anyone else help ? | 13:21.35 |
norbertj | I have a question on pllfont.c. There the fonts are loaded, but on exit of plmain() are the also unloaded ? In our pdl we enter/leave the plmain while keeping the executable alive (between jobs), but I suspect that there now is a memoryleakage. | 13:23.08 |
kens | Hmm I htink you will have to direct that to either chris or henry. I could go investigate it for you but I suspect it would take me longer to give a definitrive answer than just waiting for one of them. | 13:24.06 |
norbertj | Ok. I'll wait for them | 13:24.42 |
kens | I'll point chris at the logs when he comes in, if you can stick around for a little while. Or you can send an email to support, which will get a reply later | 13:24.44 |
chrisl | I'm back now | 13:24.55 |
kens | :-) | 13:24.59 |
norbertj | Hi chrisl | 13:25.01 |
chrisl | Hi | 13:25.06 |
ghostbot | que tal, chrisl | 13:25.06 |
chrisl | norbertj: I'm not sure if/where the fonts are released, I'd need to dig through the code. | 13:25.54 |
| norbertj: I'll have a look this afternoon, and either ping you here, or drop you an e-mail if you've gone by then | 13:27.36 |
norbertj | chrisl, what I also see is that in pxPassthrough_init -> pcl_do_resets ->pcfont_do_reset ->pcfont_do_reset->pcl_set_current_font_environment-.pl_load_built_in_fonts->pl_load_built_in_mtype_fonts-> etc | 13:28.01 |
| that also a huge malloc is done. If fonts were already loaded I did not expect that a second pl_load_built_in_mtype_fonts would be required. | 13:28.48 |
| This was when using the C706.bin testfile. | 13:29.14 |
chrisl | norbertj: yes, as I understand it, the fonts have to be loaded twice, as the PCL5 environment is separate from the PXL one - we can't risk PXL softfonts contaminating the PCL world, or vice versa | 13:29.45 |
norbertj | But in pcl5 interpreter they were already loaded (this is standard without a job), and in pxl this is done at the first job in pxsession | 13:30.30 |
| and this is during a passthrough from pxl to pcl | 13:30.55 |
chrisl | My understanding, which is, I'll grant, limited: the passthrough from pxl to pcl is totally separate from the "top level" pcl. But that really is henrys's area | 13:32.14 |
Beleg | Hi, I'm feeling like a complete noob here... I've been trying to compile jbig2dec for hours and am still unsuccessful. | 13:36.19 |
norbertj | Ok, I'll wait for henrys fot this. I checked with the debugger and see indeed that the pfontdict passed to pl_load_built_in_fonts() during the pxPassthrough_init...is indeed empty. And I now suspect that it is not released again afer the passthrough. | 13:36.23 |
Beleg | I would be so grateful if anyone could help me compile jbig2.exe for win32 :) | 13:36.42 |
Robin_Watts | Beleg: Is there a jbig2.exe ? :) | 13:36.59 |
Beleg | sorry, jbig2dec.exe :) | 13:37.14 |
Robin_Watts | Beleg: Is there a jbig2dec.exe ? :) | 13:37.23 |
| My (limited) experience of jbig2dec is in using it as a lib in gs and mupdf. | 13:37.47 |
| the standalone exe doesn't get much TLC. | 13:38.00 |
chrisl | norbertj: Hmm, I don't see the fonts being released anywhere in a "normal" start interpreter/stop interpreter cycle, so although I may end up looking at it, we really need henrys's greater experience in the area. | 13:38.12 |
Robin_Watts | and I suspect that when it is built, it's only on unix. | 13:38.24 |
kens | There's an msvc makefile | 13:38.52 |
Beleg | I found an old version, but I need an updated one as the one I'm using doesn't always work... | 13:38.55 |
kens | looks like its supposed to build an executable | 13:39.08 |
Beleg | tried the msvc makefile, but i'm getting compile errors left and right | 13:39.22 |
tor8 | Beleg: jbig2dec has a very simple build system, you ought to be able to just do the equivalent of "cc *.c" | 13:39.28 |
kens | tor8 it really doesn't work | 13:40.26 |
| looks like hte makefile is expecting some libpng stuff | 13:40.38 |
tor8 | kens: the jbig2dec executable (optionally) uses libpng | 13:40.57 |
kens | its trying to open pnglibconf.h | 13:40.58 |
Beleg | i found the libpng stuff and managed to compile a libpng.lib | 13:41.10 |
tor8 | I think if you comment out the LIBPNG bits of msvc.mak it should work | 13:41.12 |
| of course, that's not helpful if you want jbig2dec to write png files | 13:41.37 |
Beleg | but when i try to compile jbig2_image_png.c I get 4 different compile errors | 13:41.50 |
| i'm only writeing jpeg files, so that could work | 13:42.20 |
| i'll try :) | 13:42.22 |
kens | Doens't work for me | 13:42.27 |
| at least not yet | 13:42.32 |
| still trying to include png.h | 13:42.43 |
norbertj | chrisl: I wait for henrys. I see a definite memoryleak for the pcs->built_in_fonts in pcfont_do_reset() being called in the pxpcl_release() and pxPassthrough_init(). | 13:42.43 |
kens | needs zlilb also apparently | 13:43.18 |
| or possibly anyway | 13:43.23 |
tor8 | kens: libpng needs zlib | 13:43.46 |
chrisl | The png parts of jbig2dec have bitrotted - the don't work with recent libpng releases | 13:43.55 |
tor8 | I don't think jbig2dec depends directly on zlib | 13:43.55 |
henrys | I heard passthrough, my ears started burning and I woke up | 13:43.57 |
kens | yeah, but in this case its the inclusion of png.h | 13:44.01 |
henrys | did I mention I hate passthrough | 13:44.13 |
norbertj | hi henrys: Look at pl_dict.c/pl_dict_init() it just resets the pdict->entries and pdict->entry_count to 0. (called from pcfont_do_reset(). | 13:45.06 |
chrisl | henrys: is there a reason we can't reuse the pcl5 interpreter instance created at startup for the passthrough function? | 13:45.46 |
kens | tor8 well I can get it to compile, but not link | 13:45.49 |
chrisl | kens: did you follow the instructions in the makefile? | 13:46.37 |
henrys | chrisl: yes | 13:46.39 |
kens | chrisl no, because I don;t have a suitable libpng,so I commetned out the libpng stuff | 13:46.57 |
chrisl | henrys: okay.... erm, thanks | 13:47.09 |
kens | but OBJS still includes jbig2_image_png so either I get compile errors or link errors | 13:47.43 |
henrys | chrisl: there are different flavors of passthrough:snippet and full page - also you just select a font which then gets used by XL, if I started up a a complete instance it would reset everything - I could parameterize the instance of each type of passthrough but that seemed like a lot of work at the time. | 13:49.20 |
norbertj | henrys: isn't it safer in pcfont_do_reset() to call pl_dict_free on a dict that is not empty? Before calling pl_dict_init() on same dictionary? i.e. pcs->built_in_fonts | 13:50.32 |
henrys | norbertj: it's like a regression, but I need to look, the fonts should be freed | 13:51.21 |
| s/like/likely | 13:51.31 |
chrisl | Oh, I bet this my fault :-( | 13:51.54 |
kens | chrisl tor8 jobig2_halftone.c isn't ioncluded in the makefile, but functions used by it are referenced in the main code. I suspect the makefile has rotted wrt the code | 13:52.32 |
chrisl | kens: that's not surprising, since it didn't do haftone areas when that makefile was last touched :-( | 13:53.43 |
henrys | norbertj: best to create a bug and make it p1, I'm in the midst of something else (another p1) and I'll tend to it next. | 13:53.45 |
kens | chrisl yes, figured that, I'm trynig to fix it | 13:54.13 |
henrys | what a horrible thing to wake up to, passthrough ! geez ;-) | 13:54.25 |
chrisl | henrys: Actually it doesn't look like it's my fault after all, but if you want me to look at it, let me know | 13:54.28 |
| henrys: just go back to bed ;-) | 13:54.46 |
| kens: I'll leave it your hands, then..... | 13:55.09 |
kens | chrisl not getting very far.... | 13:55.30 |
henrys | chrisl: yes I can't imagine it would be you, I had this idea that I could just keep the fonts around - a single allocation and forgot about the dreaded passthrough and its shenanigans | 13:56.03 |
kens | chrisl Beleg I now have an executable | 13:57.04 |
chrisl | henrys: no, I moved a check about whether the path to the fonts was NULL or not, I was worried I'd inadvertently moved something else at the same time, but I didn't | 13:57.35 |
Beleg | kens: i'm very interested :) | 13:59.31 |
kens | Yeah give me aminut to tidy up | 13:59.41 |
| Of course, I have no idea if it actually works | 13:59.53 |
Beleg | :) | 13:59.59 |
kens | Beleg try this: | 14:01.13 |
| http://pastebin.com/KeJ8WbKC | 14:01.13 |
| Note that this does not use libpng, but the instructions are in the makefile if you want to try that too | 14:01.42 |
norbertj | henrys: bug 695017 submitted | 14:02.38 |
kens | chrisl I'm too dumb to do the Unix makefile, want me to open a bug report ? | 14:02.50 |
chrisl | Hmm, I care so little :-( No, I'll do it now | 14:03.37 |
kens | OK you want my stupid msvc changes ? | 14:03.49 |
| Its pretty trivial.... | 14:03.58 |
chrisl | Just commit them | 14:04.05 |
kens | (otherwise I wouldn't be able to do it ;-) | 14:04.09 |
Beleg | Thanks kens, I now get an executable | 14:05.33 |
chrisl | kens: the Unix one works.... | 14:05.42 |
kens | But does it work ? :-) | 14:05.44 |
Beleg | But i'm afarid the error I was hoping to eliminate is still there | 14:05.50 |
| does anyone want the pdf I'm trying to convert? | 14:06.26 |
kens | You can't convert a PDF file with jbig2dec | 14:06.44 |
| You ouwld have to extract the JBIG2 data first, possibly also decompressing it | 14:07.00 |
| Drat, chrisl can I send you the makefile ? the pre-commit hook won't let me commit it withg a tab | 14:07.45 |
chrisl | are you using the command line? | 14:08.01 |
kens | No, git gui | 14:08.07 |
| I only use tghe command line under protest | 14:08.16 |
chrisl | Ugh :-( Yes, send me the change, then | 14:08.22 |
kens | Thanks | 14:08.30 |
Beleg | i'm using a pdf viewer which in turn uses jbig2dec as a plugin to convert pdfs to jpegs | 14:09.08 |
kens | Beleg, we aren't going to look at someone else's code, we have enough to do ;-) | 14:09.43 |
chrisl | Beleg: you could try running the PDF through Ghostscript - it can output jpegs | 14:10.35 |
henrys | chrisl:it is -24 here maybe going back to bed was the right call. | 14:15.31 |
kens | is that centigrade or funny scales ? | 14:15.47 |
henrys | cent | 14:15.54 |
kens | Wow | 14:15.58 |
| Remind me not to visit Henry in the winter | 14:16.08 |
chrisl | henrys: stone me! I thought the wind and rain was bad here...... | 14:16.23 |
henrys | kens:very unusual here there's a theory it's to do with the arctic melting and the jet stream slowing leaving these cold fronts "stuck" but I don't know. | 14:17.55 |
tor8 | Robin_Watts: henrys: hm, Javascript is a trademark owned by Oracle... you need to buy a license to actually use the name. maybe we should go with something based on the "real" name of the language: ecmascript. | 14:32.42 |
chrisl | "occamscript" | 14:33.46 |
| tor8: ^^ | 14:33.50 |
tor8 | chrisl: :) | 14:33.59 |
henrys | tor8:I don't see how any of the chosen names would be the same as the trademark | 14:34.09 |
| legally | 14:34.38 |
Robin_Watts | mujs still wins as the name for me. | 14:55.39 |
janmoesen | call it jhostscript and be done with it | 14:56.20 |
Robin_Watts | chrisl: Did you ever have the joy of programming in occam? | 14:56.38 |
chrisl | Robin_Watts: no, never appealed, really | 14:57.10 |
Robin_Watts | It wasn't that it appealed to me :) | 14:57.28 |
chrisl | When suggested that above, I'd forgotten occam was already a language | 14:58.07 |
Robin_Watts | Anyone here have any experience with windows manifest files? | 15:01.28 |
kens | Not much | 15:01.35 |
| I looked briefly at it as part of the 'signed driver' that came up a while back | 15:01.56 |
Robin_Watts | The final link that the FBS does produces da-test.exe and da-test.exe.manifest. | 15:02.27 |
kens | IIRC the manifest files are part of the whole 'signing' thing | 15:02.59 |
| Oh yes, they are also (yet another) way of application-specific storage (.ini files, registry etc) | 15:04.21 |
Robin_Watts | and it's the da-test.exe.manifest that tells it what DLL to link with - attempting to rename the exe or move the exe away from the manifest file results in an exe that won't start as it can't find the MSVCR80.DLL. | 15:05.21 |
| I'm struggling to see why the FBS ends up with a manifest where, for instance, ghostscript does not. | 15:05.41 |
kens | its a C# thing | 15:06.11 |
Robin_Watts | No C# involved in the FBS. | 15:06.24 |
| or C++. | 15:06.31 |
kens | .com,.net, etc | 15:06.34 |
Robin_Watts | It's all just vanilla C. | 15:06.36 |
chrisl | Does it use a makefile or a project? | 15:06.52 |
kens | decscribes classes, required resources etc | 15:06.53 |
| the4 manifest is XML | 15:07.03 |
Robin_Watts | chrisl: The FBS uses neither makefiles or projects. | 15:07.05 |
| kens: Indeed. The Manifest produced basically points to the C runtime DLL used. | 15:07.25 |
| and we only use the standard C runtime one (MSVCR80.DLL) | 15:07.39 |
| same as gs. | 15:07.43 |
kens | you doin;'t have to have one | 15:08.01 |
| but if you do, it has to be correct | 15:08.21 |
Robin_Watts | How do I tell it not to make me one? | 15:08.28 |
kens | edelete it ? | 15:08.39 |
Robin_Watts | Then the exe won't run. | 15:08.53 |
kens | IIRC its a compiler switch to make it | 15:09.14 |
| possibly you need to remove it | 15:09.39 |
| "/MANIFEST~maybe | 15:10.17 |
Robin_Watts | kens: I can't see any flags asking for a manifest. | 15:10.43 |
| and I've asked for /MANIFEST:NO and ended up with something that wouldn't run :( | 15:11.06 |
chrisl | I think it's the linker that creates the manifest | 15:11.12 |
kens | Its all a bit new | 15:11.34 |
Robin_Watts | chrisl: It is. | 15:11.48 |
| kens: This is VS2005, so not that new :) | 15:11.58 |
| It's the linker line that I've been looking at. | 15:12.22 |
kens | yeah I meant all this manifest stuff is (compratively) new. its all part of the whole 'managed code' stuff | 15:15.26 |
Robin_Watts | kens: I don't see how that's the case | 15:15.47 |
kens | Well, that's my impression | 15:16.03 |
Robin_Watts | Manifests have been there since VS2005 at least, and there is no managed code involved in this case. | 15:16.28 |
kens | Well I think MS have been moving towards it for a while, piecemeal | 15:17.03 |
chrisl | Robin_Watts: how about creating a dummy project, and using the GUI to select whether to create a manifest or not, and see what difference it makes to the linker command line? | 15:17.45 |
ray_laptop | Note that Ghostscript does have .manifest files, e.g. psi/dwsetup_x64.manifest | 15:18.11 |
Robin_Watts | ray_laptop: Oh,wow. | 15:18.41 |
| only for the setup and uninstaller though? | 15:19.17 |
ray_laptop | there are 3 for 'setup' (install) and two for uninstall | 15:19.21 |
Robin_Watts | I only see 2 each. | 15:19.34 |
ray_laptop | sorry, typo | 15:19.46 |
| OBO | 15:19.48 |
| I have NO idea what they are for | 15:20.02 |
| I'm sure they were created by Russell | 15:20.18 |
Robin_Watts | It looks like gs/bin/gsdll32.dll.manifest is created by the build. | 15:20.29 |
| OK, I have a command line I can use to embed the manifest. Now I just need to make the FBS call it :( | 15:24.58 |
chrisl | henrys: I got the cycle computer that's matched to my turbo trainer a couple of weeks ago - I think it's going to be a useful training aid | 16:00.21 |
henrys | chrisl: ah cool, which trainer do you have? | 16:19.14 |
chrisl | henrys: kurt kinetic | 16:19.25 |
henrys | looks just like mine | 16:20.19 |
chrisl | henrys: what's yours then? | 16:20.43 |
henrys | cycleops I don't use it much the spinning classes here have all gone high tech with a board up front to display watts so you can compete with your comrades | 16:21.24 |
chrisl | I prefer to suffer in private :-) | 16:21.50 |
henrys | well I'm stuck at home for now, I might use it, my car doors are frozen shut ;-) | 16:23.06 |
kens | would think car frozen to floor at that temperature | 16:23.31 |
chrisl | I had that last winter.... | 16:23.33 |
| henrys: so, according the computer, for the 30-60 minute sessions I do, I was holding 60-70rpm and ~170 watts output, and then one minute in ten, I sprint at 90-100rpm and ~350 watts. | 16:25.41 |
henrys | I tried them briefly while I was shoveling, couldn't open any of the doors but I didn't really try hard. I could take a blow dryer to them I suppose. | 16:25.42 |
chrisl | henrys: I've now upped that to 70-80rpm for the session, which is ~210 watts output, still with the sprints - and it's freakin' hard work! | 16:30.13 |
henrys | chrisl: I looked at age group athletes top 3 for ironman and they do something like your sprint numbers for the entire race - 112 miles. | 16:30.46 |
chrisl | henrys: yeh, the one hour, fixed wheel distance record, the guy sustained ~120 rpm cadence for the 60 minutes - amazing stamina | 16:32.34 |
henrys | chrisl: be careful with bodyweight it's watts/per pound that count ⦠you can lose some weight and thing your getting worse watts wise. | 16:33.23 |
| s/thing/think | 16:33.37 |
chrisl | henrys: That's part of why I wanted the computer - I felt I was in a comfort zone, so I wanted solid numbers to work to. As I said before, for my purposes, I'm more interested in cadence and heart rate than watts | 16:35.07 |
henrys | chrisl: watts/pound really does work though as a metric ⦠it does predict race placement. rpm's obviously won't work. | 16:36.26 |
ray_laptop | henrys: if you pull hard enough you can usually open the doors (IME). Of course, also IME, you usually leave part of the gasket stuck to the door jamb ;-) | 16:36.57 |
henrys | ray_laptop:that was my concern thought it best to let the sun come up a little more and see if that does any good | 16:37.49 |
chrisl | henrys: true, but as I'm primarily using it as a training aid for another sport, as long as I have measurements that I can monitor my own progress (or otherwise) that's all I need. If I start actually cycling more seriously, I'd obviously be more interested in watts/pound | 16:39.01 |
ray_laptop | if your car has remote start, you can let the heater do the trick (presuming you left it on). That way the door jambs warm up. Of course if you don't have remote start... | 16:39.07 |
henrys | chrisl: I really do most races and training by heart rate only, just the spinning class has gotten to be a social competitive type thing | 16:39.34 |
ray_laptop | henrys: I also have experience gluing door gaskets back together with contact cement, if you need it ;-) | 16:39.47 |
henrys | chrisl: when I was training indoors I liked the sufferfest videos, kept me motivated. | 16:41.46 |
chrisl | I have music on, takes my mind off the pain...... | 16:42.16 |
ray_laptop | getting further with psdcmykog. I now get the first 3 pages out looking reasonable. Pages 4-9 all give Photoshop 'unexpected EOF' even though the file sizes are the same as the pages that work. | 16:44.21 |
Robin_Watts | paulgardiner: OK, I have the FBS hacked about enough that it now embeds manifests into the exe's. | 16:57.10 |
| And I have a build underway which embeds the windows fonts in instead of the picsel fonts. | 16:57.41 |
| so I should be able to run stuff side by side and look for differences. | 16:58.05 |
kens | goodnight all | 17:20.38 |
Robin_Watts | mvrhel_laptop: Morning. | 17:22.21 |
| Did you get to look at the lcms threading stuff? (Not hassling, just interest) | 17:22.49 |
mvrhel_laptop | Robin_Watts: oh yes. sorry I did look it over I need to run a couple tests though. I will finish up with that today | 17:25.55 |
| been to busy working on gsview | 17:26.05 |
| s/to/too/ | 17:26.23 |
Robin_Watts | no worries. | 17:26.36 |
ray_laptop | OK. psdcmykog is now producing "reasonable" output. Colors look strange, but at least the pages aren't garbled and no errors produced. Basically I had to have it ignore the SeparationOrder parameter, locking it at 6 components. | 19:08.53 |
Robin_Watts | ray_laptop: Great that it's working, but it's a bit of shame. | 19:09.29 |
ray_laptop | output is different to the psdcmyk device, but that one produces most pages blank which also doesn't seem quite right. | 19:09.30 |
| Robin_Watts: why is that a shame ? | 19:09.46 |
Robin_Watts | It was intended at least partly as an example device to show how such devices should be written. | 19:09.58 |
ray_laptop | well, the problem is that on a device with max_separations==0 setting arbitrary SeparationOrder names doesn't really make sense | 19:11.02 |
Robin_Watts | I guess. | 19:11.36 |
ray_laptop | It could be that the SeparationOrder handling in the devn code is flawed, but the use of SeparationOrder to produce subsets of the available separations is rarely used, and never from PDF. The problem file(s) are PS CET nonsense | 19:12.51 |
Robin_Watts | That would be my fear. | 19:13.24 |
| Or at least, my fear is that in my fiddling, I have broken the SeparationOrder stuff. | 19:13.44 |
ray_laptop | Robin_Watts: I think the tiffsep device used to use the SeparationOrder stuff, but what it did/does with a set of names that didn't exist was never clear | 19:16.03 |
| I'm going to open a non-customer bug to gripe that psdcmyk and psdcmykog differ because I'm not at all sure what "right" is. | 19:18.25 |
Robin_Watts | OK, I've finally got some results out of ghostdocs with the windows fonts. | 19:18.36 |
ray_laptop | that way I can assign it to mvrhel_laptop as a 'color' bug :-) | 19:18.41 |
Robin_Watts | and the fonts *do* make a different. | 19:18.43 |
| difference. | 19:18.46 |
ray_laptop | Robin_Watts: I guess I'm not surprised at that. | 19:19.03 |
Robin_Watts | So, the question is, do we own rights to a set of "windows equivalent" fonts? | 19:19.24 |
| ray_laptop: Well, it's a shame that the epage 'clone' fonts aren't. | 19:19.45 |
ray_laptop | Robin_Watts: well, what we really need to do is to use the glyph metrics from the fonts, and we can actually render the glyphs with whatever fonts we want to use. It'll look better if they are close to the metrics we use | 19:20.51 |
albanyeye | hello, is this where I ask a question about pdf2ps also? | 19:21.39 |
Robin_Watts | ray_laptop: Well, the epage font engine has the ability to show a font with a different set of font metrics. | 19:21.58 |
ray_laptop | albanyeye: you can ask, | 19:22.06 |
| Robin_Watts: that sounds good. | 19:22.28 |
Robin_Watts | In particular it uses that ability to load the PDF base 14 fonts font metrics so that PDFs work. | 19:22.32 |
ray_laptop | Robin_Watts: sort of allowing for something like 'multi-master' font substituton | 19:23.02 |
Robin_Watts | It doesn't do proper multi-master. | 19:23.24 |
albanyeye | I have a pdf that has radio buttons that have either been selected or not. The PDF gets sent to hylafax to be faxed. The pdf recieved has " x" marked where the radio buttons are supposed to be. Regardless of it being selected or not | 19:24.14 |
| I have looked on google search for a couple days without success | 19:24.37 |
Robin_Watts | albanyeye: Sounds like a PDF with a form in it. | 19:25.35 |
| And it further sounds like the form is being improperly rendered. | 19:25.50 |
ray_laptop | albanyeye: we'd have to look at the PDF. Ghostscript doesn't look at the 'value' of a field, but uses the AP (default appearance) | 19:26.02 |
| albanyeye: just open a bug and post and example (small preferable) file | 19:26.55 |
albanyeye | yes it is a PDF form | 19:27.37 |
troubles | hi, any mupdf folks around, i have a problem hopefully they know what to do | 19:27.54 |
albanyeye | oh great | 19:27.55 |
| I can do that. Where to post the file? | 19:28.06 |
| its a 1 page PDF... very small | 19:28.25 |
ray_laptop | troubles: Robin_Watts is your man | 19:28.31 |
troubles | okay, i see he is logged in, thanks. perhaps he is busy | 19:29.05 |
Robin_Watts | troubles: Ask your question, don't ask to ask. :) | 19:29.06 |
ray_laptop | albanyeye: bugs.ghostscript.com -- open a New bug against Ghostscript and attach the file | 19:29.06 |
albanyeye | Robin..you might be right about form being imporperly renderd | 19:29.16 |
| it is being renededred by html2pdf.fr via php | 19:29.37 |
Robin_Watts | albanyeye: That sounds the pdf creation stage. | 19:30.02 |
albanyeye | will do ray. | 19:30.21 |
troubles | hey thanks. robin, i am trying to make an ereader to replace kobo erader (does not do pdf well). my problem is that i have two different jpeg librs needed. it seems mupdf needs version 9 (which i built static, works fine), but qtwebkit uses version 6.2, and i cannot get both of these in one app, as far as i can tell. | 19:30.27 |
Robin_Watts | The pdf 'rendering' stage (i.e. the stage where we go from pdf -> bitmap) in order to be faxed is the one that is going wrong. | 19:30.46 |
ray_laptop | albanyeye: "improper" is a matter of interpretation. Ghostscript is a non-interactive processor of PDF's, so it ignores any javascript and, as I said uses the AP of fields | 19:30.48 |
Robin_Watts | probably. | 19:30.50 |
troubles | robin: so maybe i could build mupdf with old version of jpeg, but i gather with api change that may not work. cant think of other idea except to try to rebuild qt entirely also! | 19:31.06 |
Robin_Watts | troubles: Just a mo. | 19:31.20 |
troubles | sure, thanks. | 19:31.28 |
albanyeye | thanks Robin / Ray | 19:32.18 |
Robin_Watts | troubles: At what point are you hitting the problem? | 19:32.32 |
albanyeye | will post file in bugs | 19:32.32 |
| by | 19:32.34 |
| ebye | 19:32.37 |
Robin_Watts | Is it when linking at the end? | 19:32.52 |
troubles | well, when i try to build my app using qmake, no, it links okay | 19:33.01 |
| robin: sometimes i can get it to link with the static jpeg, then the mupdf renderer in my app works fine | 19:33.33 |
| robin: but then the qtwebkit crases. if i change link orders around, then the qtwebkit works, but the mupdf does not | 19:33.59 |
| the mupdf works on SOME jpegs, but on others, i get console message that it is the wrong version. | 19:34.16 |
Robin_Watts | I have only limited experience with qmake, but the experience I have had suggests that I am not touching it with a gargepole :) | 19:34.29 |
| OK, there may be a trick you can use. | 19:34.38 |
troubles | so mupdf is much better behaved: it fails cleanly when the jpeg library cannot do the job | 19:34.41 |
| i dont blame you about qmake, but using qt have little choice! | 19:34.56 |
Robin_Watts | Presumably you have the mupdf version in thirdparty/jpeg? | 19:35.05 |
troubles | right, i just built the stock mupdf, it uses the "Thirdparty" jpeg etc. | 19:35.29 |
Robin_Watts | If you look at jpeglib.h in there, you'll note there is a #ifdef NEED_SHORT_EXTERNAL_NAMES | 19:35.36 |
troubles | yes... but | 19:36.14 |
Robin_Watts | If that's defined, it renames all the jpeg entrypoints to something different. | 19:36.35 |
troubles | ah, okay, i see. so then i could call the library something else, and load both librares? | 19:36.56 |
| so everyone else links shared with version 62, but mupdf uses the new version under a different name? | 19:37.21 |
Robin_Watts | So if you put a #define NEED_SHORT_EXTERNAL_NAMES at the top of that file, then everything that file (which should be all the mupdf code and the library code itself) will use different names, | 19:37.27 |
| Thus the two copies should no longer conflict. | 19:37.36 |
| troubles: yes. | 19:37.47 |
| You need to ensure that the mupdf and jpeg-9 builds see that header and define, but that nothing else does. | 19:38.35 |
troubles | wow, what a kludge! i dont like using two libraries mostly the same, but if it works fine, and the jpeg lib is not too big i have had enough for today/yesterday, so will try later tonight. can you tell me why the version 62 seems to be the version most stil use (fedora 20 ships with it for example) | 19:39.00 |
Robin_Watts | Cos they are slow to update with fixes etc. | 19:39.18 |
troubles | just jpeg-9? what about the other things in thirdparty? dont some of them need the jpeg? no, i guess not. So, re fedora 20, imagemagick, everyone else, you guys are ahead on this? | 19:39.56 |
Robin_Watts | yes. we do our best to track current releases. | 19:40.17 |
troubles | would you confirm i cannot build mupdf with the earlier jpeg? if this rename does not work? | 19:40.28 |
| okay, thanks i lot, i will work on this for many hours if necessary, can i report back the results. i think this could be useful for others who might also have this problem | 19:41.39 |
Robin_Watts | The alternative way to do this would be to build mupdf and to statically link it with the jpeglib, to give a mupdflib. Then you can hide the jpeg entrypoints in mupdflib. | 19:43.06 |
troubles | ah, i like that idea, okay, i will try both of these. will take me a while i bet but i will report back, thanks | 19:46.25 |
ray_laptop | Robin_Watts: does that had the names of the jpeg entrypoints ? or does this work IFF you have the jpeglib before the mupdflib ? | 19:46.29 |
| s/had the/hide the/ | 19:46.44 |
Robin_Watts | ray_laptop: The trick is to link 2 libs together to give a third. | 19:47.27 |
| Then you rename all the jpeglib entrypoints within that new lib. | 19:47.43 |
troubles | ray_laptop: have you ever had an issue like this yourself also? | 19:48.30 |
Robin_Watts | The two halves of the lib can still call one another as they agree on the renamed (or hidden) names. | 19:48.34 |
ray_laptop | Robin_Watts: do you want to comment on my latest patch, please ? | 19:48.41 |
Robin_Watts | now, I have to confess it has been many years since I've had to do this myself. and when I did it I had a funky tool to do the renaming. | 19:49.02 |
troubles | Robin_Watts: well, i am optimistic that it should work if I keep messing with it. as long as the only dependency is mupdf on jpeg | 19:50.23 |
Robin_Watts | ray_laptop: It's a shame that you have to save/set to zero/restore the sep names. | 19:51.27 |
| If we're hooking put params, can't we throw a rangecheck if they are non null ? | 19:51.54 |
ray_laptop | Robin_Watts: I also opened bug 695018 for the discrepancy between psdcmykog and psdcmyk. I really am not sure what we should be doing. | 19:51.54 |
Robin_Watts | If the device only operates when SeparationOrderNames is empty, then surely it's better to enforce that with an error? | 19:52.42 |
troubles | Robin_Watts, thanks very much, I will go and try these things. | 19:53.00 |
ray_laptop | Robin_Watts: I can rty that, but I'm not sure that's really better | 19:53.35 |
Robin_Watts | oh, it's a file that sets SeparationOrder ? | 19:53.46 |
| (Just reading the bug) | 19:53.52 |
| What is the correct behaviour here? | 20:01.55 |
| If a bit of postscript attempts to set an illegal setting for a device, is a rangecheck not the correct response? | 20:02.34 |
| What happens for other devices? Like the display device for instance? | 20:03.19 |
| ok, display device ignores it, afaict. | 20:04.31 |
| Well, what you've got seems reasonable then. | 20:05.03 |
troubles | Robin_Watts: Thanks a lot, it works! (Now to see when I try building for other platforms... (But it is nice that mupdf lives in its own world with its own jpeg not visible to anyone else (Oh, I used the "short name" technique, more kludge, but, since you set it up already... works well it seems, thanks] | 20:35.46 |
ray_laptop | hmm.. bad news. The -dDisplayFormat=16#a0800 is broken since 9.10 | 20:55.36 |
| I'm bisecting to see when it broke... | 21:09.43 |
mvrhel_laptop | wow | 21:43.24 |
| that took me forever to find | 21:43.36 |
| I had a crash with the xps printing in the new gsview | 21:43.47 |
| I could not find this damn thing | 21:43.53 |
| it turns out that if I have the "Debug native code" checked in the project settings that it crashes for some unknown reason. If that is not checked during the printing then all is fine | 21:44.25 |
ray_laptop | your forever is a lot shorter than mine :-) | 21:44.32 |
mvrhel_laptop | well something as crazy as that is very irritating. | 21:44.49 |
| The native debugging worked fine coming from csharp and going into gs or into mupdf. You would think that having that checked when going into microsofts native code things would be OK | 21:45.28 |
| The crash is about in a stack of 50 different calls from the XPS DLL into some core library when it explodes. | 21:46.08 |
ray_laptop | git bisect has henrys' patch of 82fc3bd as the culprit | 21:49.42 |
| now to find out why | 21:49.54 |
daniel1 | Hi/part | 21:50.20 |
mvrhel_laptop | ok. and with that check box off, I am now printing | 21:56.43 |
| henrys: so I have gs creating xps content with xpswrite and then that is getting pushed through a print queue in the windows print system. | 21:57.26 |
| need to clean up a few things but it seems to be working fine | 21:57.39 |
henrys | mvrhel_laptop: great my xps doesn't blow up windows - shocking | 21:57.48 |
| ? | 21:57.50 |
mvrhel_laptop | at least tiger worked :) | 21:57.59 |
| I wish that the gs api had some callback mechanism that I could use to show distilling progress. | 21:58.28 |
henrys | mvrhel_laptop: I wonder if check_interrupts could be hijacked for something like that, ray_laptop would know | 22:00.03 |
mvrhel_laptop | ray_laptop: | 22:00.22 |
| ? | 22:00.23 |
| a page by page call back would be nice | 22:01.06 |
| oh he is not here | 22:01.55 |
| anyway I will put that on my christmas wish list | 22:02.11 |
henrys | mvrhel_laptop: why don't you use the run_string stuff | 22:02.15 |
| mvrhel_laptop: example 3 in the api.htm | 22:03.00 |
| ? | 22:03.01 |
mvrhel_laptop | hold on let me look at it | 22:03.08 |
henrys | mvrhel_laptop: of course telling how many bytes are left does not say how long it will take | 22:03.59 |
mvrhel_laptop | no, but at least it gives a percent through byte wise | 22:04.24 |
| henrys: this could work. I will go back and see about this after I clean up my printing stuff | 22:05.00 |
| I have everything ripped apart trying to track down the cause of my crash | 22:05.25 |
| which was MS falut | 22:05.31 |
| fault | 22:05.33 |
henrys | mvrhel_laptop: I thought the strdup punishment didn't fit the crime... | 22:06.05 |
| geez | 22:06.11 |
mvrhel_laptop | well that was an easy one | 22:06.21 |
| this last thing was the type that has you pulling your hair out | 22:06.49 |
| hopefully I wont have anymore of that nonsense. most of the hard pieces are in place | 22:07.15 |
| it is a matter of adding minor features | 22:07.21 |
| I do have some mupdf related stuff to add, text search, links | 22:07.36 |
| but those should be pretty simple since I already did them in win 8 | 22:07.51 |
| bbiaw | 22:28.04 |
daniel1 | Hi, I want to print multiple pages of a pdf into one. My printer hase this option but it offers me not enough controll over the process. Are there some tools with which I can rescale pdf pages manually without loosing the vectors ? I tied it allready with the freepdf printer + editing via inkscape but freepdf rasters the outcome. | 22:58.20 |
Robin_Watts | henrys: You here? | 23:59.24 |
henrys | yup | 23:59.33 |
Robin_Watts | So, I've done some testing. | 23:59.40 |
| The epage fonts do NOT match the windows ones in terms of metrics. | 23:59.53 |
| Forward 1 day (to 2014/02/06)>>> | |