| | <<<Back 1 day (to 2015/09/08) | 20150909 |
| fredross-perry | robin_watts: take a look at b137f19 in my repo | 00:20.19 |
| lhx | Question: is it possible to change the jpeg quanitzation away from the default when -sDEVICE=pdfwrite? I really want to change the jpeg quanitzation as opposed to to PPI of the jpegs in some large PDF files. Thoughts? | 02:25.49 |
| | I've tried -dJPEGQ and -dQFactor and it doesn't seem to have any effect | 02:28.36 |
| rayjj | lhx: The QFactors (which apply to JPEG, as far as I know) are described in the notes following: http://www.ghostscript.com/doc/current/Ps2pdf.htm#note_7 | 04:12.12 |
| | Thus, you can set paramters into the ColorImageDict, ColorACSImageDict, GrayImageDict, and/or GrayACSImageDict that should allow you to alter the QFactor | 04:16.27 |
| | lhx: this would be done from the command line with something like the following immediately preceding your input file: -c "<< /ColorImageDict << /QFactor 0.2 /Blend 1 /HSamples [2 1 1 2] /VSamples [2 1 1 2] >> >> setdistillerparams" -f | 04:19.20 |
| | lhx: note that you may find it easier to put the all of the PS that sets the distillerparams into a file and then just put the name of that file just before the input file. In that case, the -c and -f are NOT used, but simply: myparams.ps infile.ps | 04:21.14 |
| | or infile.pdf | 04:21.20 |
| lhx | rayjj: thanks! That was an awesome explanation. Is there a reason why that's such a PITA to do it that way instead of having a nice CLI flag? | 06:21.42 |
| chrisl | lhx: probably because that's how Adobe's Distiller does it. In addition, we already have far too many command line options, if we added ones for everything that can be configured in pdfwrite, we'd end up with a document the size of War and Peace, just for the command line args! | 06:24.57 |
| lhx | chrisl: thanks for the perspective. I try to stay away from acrobat when I can so I didn't have context | 06:26.08 |
| chrisl | lhx: I don't blame you! But unfortunately, Adobe is seen as a benchmark for these things.... | 06:27.28 |
| lhx | chrisl: I do have the ligthroom and PS subscription | 06:27.46 |
| | chrisl: but I refuse to pay $20 a month for PDF software | 06:27.57 |
| chrisl | The thing is, we *have* to support these things through setdistillerparams since that can be called in the incoming Postscript jobs, and since there is a defined way to push Postscript via the command line, it makes sense to utilise that, rather than add to the existing proliferation of (already confusing!) CLI options | 06:30.07 |
| rayjj | lhx: I agree wholeheartedly | 06:30.15 |
| chrisl | Additionally, these are settings that are changed only very rarely..... | 06:30.44 |
| rayjj | chrisl: yeah. I don't want us to turn into the imagemagick CLI ! | 06:31.18 |
| chrisl | Ick! | 06:31.49 |
| rayjj | just 'convert' has 311 lines of emitted by --help | 06:32.01 |
| | although, if we collected the Use.htm and Language.htm options, we may already be close :-/ | 06:33.15 |
| chrisl | Hence preferring not to add even more! | 06:33.51 |
| lhx | heheh... I guess I don't fundamentally understand the technology | 06:34.13 |
| rayjj | chrisl: yes! totally in agreement | 06:34.19 |
| chrisl | And that's before my inherent dislike of ImageMagick! | 06:36.40 |
| rayjj | lhx: the issue is that ghostscript does _so_ much that it has options for this and that and the other. Viewing on the screen, making jpegs, pngs, tiffs, pnms, printer format streams, cups, "high level" PS Level 2, PDF (at anything from 1.2 to 1.7), extracting text, ... | 06:37.15 |
| lhx | rayjj: sure... so why not another flag? ;) | 06:37.34 |
| chrisl | lhx: it's also preferable to avoid having 2 or more different ways of achieving the same goal, where possible...... | 06:38.52 |
| rayjj | lhx: there are basic options in Use.htm, GS specific options in Language.htm and options that only are relevant to a specific device (conversion task) in Devices.htm, then there are the "high level" devices -- ps2write and pdfwrite that add even more options | 06:39.02 |
| lhx | been using pdfwrite | 06:39.40 |
| rayjj | chrisl: so, that would imply that -dQFactor should also control the pdfwrite device ? | 06:39.51 |
| | chrisl: but the pdfwrite allows for different ImageDict for Color, ColorACS, Gray, GrayACS, ... | 06:40.33 |
| chrisl | rayjj: possibly. But as I said, we *have* to support setdistillerparams for pdfwrite, so..... it's debatable | 06:40.44 |
| rayjj | chrisl: IMHO, having to have "distillerparams" that are ignored by other devices makes sense. | 06:41.42 |
| | and the basic problem is the implicit connection between CLI options and device specific options. For example: -sCompression=lzw is only relevant to TIFF devices, but as a command line option it just "works" without error and is ignored for other devices | 06:43.56 |
| | sort of like lhx's attempt to use -dQFactor=_.__ being ignored by pdfwrite, but affecting the jpeg device | 06:45.03 |
| chrisl | Sure, but changing that stuff now is really out of the question :-( | 06:45.54 |
| rayjj | we might be better off with --device-opts="___" in order to differentiate it from 'generic' or 'parser' options | 06:46.52 |
| | chrisl: well, we /might/ be able to implement a new approach (not necessarily what I had) and deprecate the old. maybe ??? | 06:47.51 |
| chrisl | But we can't drop support for the old, so the confusion would continue to arise | 06:48.17 |
| rayjj | yeah, but if the 'old' options were pre-processed into new options, and those were printed (in a "verbose" mode) the correspondence between the old and the new would become somewhat clearer | 06:49.54 |
| lhx | there was plenty of confusion reading the docs in that I ran 5 trials with different jpeg -dQFactor numbers and wondering what the hell I was doing wrong. The documentation is super unclear in that scenario | 06:49.55 |
| | I eventually found it, after having to read a footnote of a chart... | 06:50.08 |
| chrisl | lhx: the JPEGQ and QFactor options are in the section specifically for the JPEG output device - it's hard to see how that could be clearer | 06:52.07 |
| rayjj | so if someone had -sDEVICE=ppmraw and -dQFactor=0.9 we could (in verbose mode) indicate that -dQFactor=0.9 was used to set a parser variable "QFactor" to 0.9 (and that it did not end up as part of the "device-opts" options | 06:52.09 |
| | or ^^^ with -sDEVICE=pdfwrite (same -- it is ignored, except to set a variable in the PS parser systemdict) | 06:53.10 |
| lhx | chrisl: gs accepted the flag silently, the docs don't say it only works in the JPEG output device, and that's one of the few places JPEG quantization is mentioned. In the http://ghostscript.com/doc/current/Ps2pdf.htm page, you have to read a *footnote* to find this out | 06:53.44 |
| rayjj | lhx: that's sort of what I was trying to get at | 06:54.04 |
| lhx | rayjj: 10-4 | 06:54.39 |
| chrisl | So, in the section on the jpeg output device, it should state that it's only referring to the jpeg output device?? | 06:55.01 |
| rayjj | we don't have a clear disctinction between device options/flags/switches and other (parser or "main" options) | 06:55.16 |
| lhx | chrisl: Granted, I'm a noob to gs, and I'm also not the best programmer, but I do live in tmux, use emacs, code golang, and write in latex all the time, and I was having trouble groking... | 06:55.17 |
| rayjj | lhx: you are not the first. And also tripping over the need to have -d___=___ for some options and -s____=___ for others -- it's documented, but not at all obvious why or which | 06:56.43 |
| lhx | I also try to RTFM when I can :) | 06:57.06 |
| rayjj | of course, to us, it is totally obvious :-) | 06:57.11 |
| lhx | of course :D | 06:57.42 |
| chrisl | That does assume a certain familiarity with Postscript...... | 06:57.48 |
| lhx | which I don't have | 06:58.25 |
| rayjj | but even I sometimes forget and use something like -dBandListStorage=file instead of -sBandListStorage=file | 06:58.31 |
| | but at least that causes a 'typecheck' error | 06:58.50 |
| chrisl | I suppose docs like Device.htm assume a familiarity with Ghostscript's concept of devices - but I don't see how we can avoid that...... | 07:00.19 |
| kens | Write a decent manual :-) | 07:00.33 |
| rayjj | and then there's -diDisplayFormat=___ (for the "display" device") Good thing I sitll have a good memory and learned that I can input values in hex with -dDisplayFormat=16#____ | 07:00.40 |
| lhx | I appreciate the perspectives and the help. The documentation needs to have some more handholding for the absolute noob I think | 07:01.12 |
| | if you want the absolute noob's impression | 07:01.20 |
| | lol | 07:01.28 |
| rayjj | lhx: yeah. So that's why we are here :-) | 07:01.37 |
| | (job security) ;-) | 07:01.47 |
| lhx | :) | 07:01.52 |
| lhx | secures the jobs with force | 07:02.33 |
| lhx | and uses paper... lots and lots of paper | 07:02.54 |
| rayjj | kens: I'd be inclined to write a sane front end, THEN write documentation for that instead of trying to write a black and yellow book for the current CLI | 07:03.09 |
| kens | rayjj I'd write a sne device API, and a sane fonrt end API first | 07:03.52 |
| lhx | night night everyone | 07:04.05 |
| rayjj | lhx: g'nite | 07:04.16 |
| | for me as well... | 07:04.21 |
| kens | I would say though, that even in my fomrer life which had both, and a manula., people found it hard to get started | 07:04.27 |
| rayjj | kens: getting "started" isn't usually our problem. It's "how do I do this "***""? | 07:06.03 |
| | if they can't get gs installed and display a 'tiger' then, that's one thing, but there are lots of other things that people want to do that is really hard to dig out of the docs (or even the code) | 07:07.29 |
| kens | Given that the code I referred to had no fornt end to speak of, "getting started" means that sort of thing, not firing it up and running tiger | 07:08.27 |
| rayjj | and, of course, all the "lore" incantations that we get with -dWRITESYSTEMDICT -dPARANOIDSAFER -DDWIM ... | 07:08.36 |
| | maybe we should just get rid of all the options except -dDWIM | 07:09.36 |
| | google probably has a module laying around that we could use for that ;-) | 07:10.47 |
| chrisl | We should get rid of everything except "-c" - *everything* set via real Postscript :-) | 07:10.47 |
| kens | You'd still need -f as well though | 07:11.34 |
| chrisl | No, you have "run" | 07:11.50 |
| kens | Hmm yes fair enough | 07:11.59 |
| chrisl | I'm guessing there'd be some resistance to that idea.... | 07:12.31 |
| rayjj | chrisl: yeah, and make PCL users use -J".." with PJL | 07:12.48 |
| kens | boggles | 07:13.01 |
| chrisl | Exactly! | 07:13.06 |
| kens | Someone opened a bug report about using EPS2write to create multi-page PostScript ???? | 07:13.30 |
| rayjj | and gxps would take -X"___" where ___ was some XML gobbledygook | 07:13.54 |
| chrisl | I'd argue XPS could also use PJL | 07:14.36 |
| rayjj | too late for me to re-invent the world. Time for bed. | 07:14.57 |
| kens | It will have to use 'something' for us to get PDF/A out of it | 07:15.05 |
| rayjj | chrisl: XPS has it's own Job level stuff -- we'd just need to have a CLI option to adding/overrnding what's in the XPS) | 07:16.09 |
| | kens: XML can do ANYTHING, hadn't you heard ;-) | 07:16.41 |
| chrisl | rayjj: true, but typing XML would be a bigger PITA than type Postscript.... | 07:16.41 |
| kens | rayjj I try not to listen to evangelists, of any strip | 07:17.02 |
| | stripe | 07:17.06 |
| rayjj | chrisl: well, this discussion wasn't tending toward easier user experience, AFAIK | 07:17.24 |
| chrisl | rayjj: good point, XML it is! | 07:17.45 |
| rayjj | chrisl: now I'm going to have nightmares :-D | 07:18.22 |
| | my kids go to school 1 hr later tomorrow, but even so, it is past time for me to say hello to my pillow. | 07:19.28 |
| | g'nite | 07:19.38 |
| kens | goodnight ray | 07:20.25 |
| chrisl | kens: I wonder if it would be viable for eps2write to warn or error if the output file doesn't have a %d in it? | 08:01.21 |
| kens | But what if you know hte input is a single page file ? | 08:01.43 |
| | Or do you mean only when producing second and subsequent pages ? | 08:02.09 |
| chrisl | Hence possibly a warning would be best | 08:02.14 |
| kens | I think I considered it at the time and decided not to bother, I can't remember why | 08:03.17 |
| | Its a dumb thing to d anyway, and exhibits a complete lack of understanding | 08:03.38 |
| chrisl | Agreed | 08:03.47 |
| | It's unfortunate, and insane, that epswrite ignored the restriction | 08:04.28 |
| kens | Yes, that's just mad | 08:04.38 |
| | But then epswrite was, frankly, crap | 08:04.59 |
| | At least the OP is happy with ps2write | 08:05.15 |
| chrisl | Well, that is the correct tool for that job, so.... | 08:05.32 |
| kens | Well I was slightly concerned that there might have been a real problem underneath it | 08:05.57 |
| chrisl | TBH, I first thought the problem might be with the tiff device | 08:06.21 |
| kens | Well, since he didn't supply an example file..... | 08:06.36 |
| | I diagnosed the problem just from the bug title :-) | 08:06.50 |
| chrisl | Yeh, but he mentioned multi-page tiff files - which are (or were) problematic | 08:07.20 |
| kens | That didn't occur to me, but I already knew he was on shaky ground at best because there is *no* way to get eps2write to produce a multi-page EPS :-) | 08:08.03 |
| | As far as I can see, he should have got an EPS with the content from every page in the input written to the same output, ie everything on one page | 08:10.16 |
| chrisl | Which probably explains the description | 08:11.16 |
| kens | has lost the originalbug report.... | 08:11.36 |
| | Well I suppose it depends what he means by 'garbage' TIFF files. To me that rather implied broken, not 'not what I expetced', but yeah. | 08:12.33 |
| | He was probably getting all his marks on a single page. | 08:12.54 |
| | LOL SO the 'intentionally crash' bug was deliberately introduced to Igor ? ROFL...... | 08:15.03 |
| chrisl | I wonder how many people that's going to bounce around before someone just fixes the damned thing | 08:18.26 |
| kens | :-) | 08:18.36 |
| | Are you the memory manager today ? | 08:19.05 |
| chrisl | I suppose on those terms, it will probably be me | 08:19.29 |
| tor8 | kens: chrisl: I did think of reassigning it to Igor. but having a ghost own a bug didn't seem very productive in the end. | 08:19.35 |
| kens | That's what I thought | 08:19.43 |
| chrisl | I just feel it seems like a pretty trivial thing to open a bug for - rather feel that opening the bug took longer than just fixing it would have | 08:20.31 |
| kens | tor8 Yeah I didn't mean to suggest that. But it tickles me immensely to see that Igor deliberately introduced a route to a crash and apparently didn't regard that as a problem. | 08:20.31 |
| | I'd do it, and get you to check it, but I expect its as quick for you to do it yourelf | 08:21.27 |
| chrisl | I'll get to it at some point - but I'm buried in gdevp14.c and ICC profile reference counting..... again :-( | 08:22.37 |
| kens | Oh yeah I saw that | 08:22.48 |
| | I may do it anyway then and pass it by you, I thnk I'm nearly finished making PJL use dynamic memory allocation, and the next thing is to start adding new code for pdfwrite's benefit, I can take a pause there | 08:23.32 |
| tor8 | kens: Indeed! I definitely see your source of mirth. It's so typical of Igor. | 08:33.45 |
| kens | :-D | 08:33.55 |
| | Well it seems I've managed to break the PJL parser more or less completely, I can't say I'm surprised, now to fix it. | 08:35.15 |
| | chrisl ping | 12:53.23 |
| chrisl | kens: pong | 12:53.33 |
| kens | I see a problem in psitop.c in the ghostpdl targte, want me to mail you about it, discuss it here, or esomething else ? | 12:54.01 |
| chrisl | ghostpdl doesn't work, and isn't intended to work just now | 12:54.36 |
| kens | I know, jujst wondered if you wanted to know about ths particular problem | 12:55.00 |
| chrisl | Sure, can you summarise it here? Then we can decide if it's bug worthy | 12:55.26 |
| kens | In ps_impl_allocate_interp_instance() it sets up a fake argv | 12:55.51 |
| | And sets argc to 10, but the argv only contains 4 args | 12:56.05 |
| | THe code in imainarg.c doesn't stop on encountering a 'null' argumen | 12:56.24 |
| chrisl | Oh, I think I noticed that.... but fixing it didn't make it work, so I gave up | 12:56.43 |
| kens | In the current code the uninitialised data happens to work, but when I changed the memory layout, it doesn't | 12:56.51 |
| | Well, when I say 'work' I mean it gets further. | 12:57.20 |
| | I only noticed because I accidentlaly built and debugged 'ghostpdl' when I meant 'ghostpcl' | 12:58.01 |
| chrisl | I guess I could fix it, but as all that will be thrown away, it didn't seem worth the effort. I do wonder how it ever worked..... | 12:58.32 |
| kens | If you intend to discard it, then ignore me. | 12:58.50 |
| chrisl | Well, if you notice, it sets a bunch of options non-optionally, which seems broken to me. | 12:59.24 |
| kens | I can't believe it ever worked reliably, since it depends what's in the uninitialised memory. Mostly (for me) it seems to be poitning to real memory, but I ended up with one set at 0xa which seg faulted of course | 12:59.36 |
| chrisl | I think that either needs all replaced, or *majorly* rewritten | 13:00.42 |
| kens | ANyway, as long as its defunct that's fine | 13:00.48 |
| chrisl | I wonder what the "proper" way is to define a device with a device finalize method might be....... | 13:01.30 |
| kens | As I recall, there is no macro for that. | 13:01.44 |
| | I discovered when doing the device subclassing htat there is no simple way to do such a thing | 13:02.01 |
| chrisl | Wonderbar :-( | 13:02.05 |
| kens | ps_impl_allocate_interp_instanceIt certainly reduces the utility of the finalize method..... | 13:02.43 |
| chrisl | Probably explains why no device actually uses one | 13:03.13 |
| kens | OK so the PJL parser no longer seg faults, now it just doesn't work, progress! | 13:03.18 |
| | and yeah, its xetraordinarily difficutl to write a device with a finalise, which is undoubtedly why nobody does | 13:04.06 |
| chrisl | Does that imply it's effectively deprecated, and to get that functionality, it should be done with a memory manager finalize method, I wonder? | 13:05.03 |
| kens | I admit that hadn't occured to me | 13:05.28 |
| chrisl | In this case, either approach will work - so I'll use the easier one | 13:06.29 |
| kens | :-) | 13:06.36 |
| chrisl | But I think I will need to get rayjj to check this out before handing it to the customer...... | 13:07.52 |
| kens | I think it should be possible to write a macro for a device that includes a finalise routine | 13:11.44 |
| bananapie | Hello. I am writing an entry for cidfmap to use a custom ttf file for a specific font. I have two fonts to subsitute, one is called 'Arial', and with the entry I added ( with help from bug 691513 ), the arial font works perfectly. My problem is I also need "Times New Roman", cidfmap doesn't take spacing and the font gs is looking for is 'Times New Roman' and not 'TimesNewRoman' or 'Times New Roman'. How do I add spaces | 13:11.56 |
| | in a font name in cidfmap ? ( gs 9.16 from sources / Ubuntu 14.04 ). Thx | 13:11.56 |
| chrisl | kens: yeh, but I don't want to do that if it's effectively deprected | 13:12.51 |
| | bananapie: try doing "(Times New Roman) cvn" | 13:13.18 |
| kens | cidfmap is PostScript, so writ the fontname as a PostScript string then use the PostScrip cvn operator to convert it into a name | 13:13.21 |
| bananapie | Thanks kens and chrisl, I'll try what you have suggested. | 13:14.58 |
| | given, that I'm a noob with gs and postscript, it'll take me a few minutes. I'll post back to let you know if it worked. | 13:15.46 |
| henrys | kens, chrisl :I'm removing psitop and all of PSI_INCLUDED as well, I thought that implicit in the project I'm doing. | 13:17.08 |
| chrisl | henrys: I wasn't aware you were doing so, but I felt it would be required | 13:17.54 |
| henrys | chrisl: we did agree to not let anyone build that sometime ago, didn't you cause the build to fail in some way? | 13:19.01 |
| chrisl | henrys: not in the new scheme - I was just going to remove the target for the release. I must admit, I had hoped to have made more progress on it by now :-( | 13:19.59 |
| kens | henrys, chrisl, just ignore me then, I'm not up to speed on what the 2 of you are doing, and its only because I built the wrong target by accident | 13:26.28 |
| chrisl | I'd really wanted the gpdl project hidden (for now) in the VS solution, but Robin_Watts had other ideas....... | 13:27.50 |
| kens | Its just one letter different and in my befuddled state I got confused.... | 13:28.22 |
| henrys | chrisl: is the cluster building that? | 13:29.22 |
| chrisl | henrys: I don't believe so | 13:29.35 |
| bananapie | ok. I don't know what I am doing. I tried '/(Times New Roman) << /FileType /TrueType /Path (/usr/share/fonts/truetype/msttcorefonts/TimesNewRoman.ttf) /SubfontID 0 /CSI [(Identity) 0] >> ;' and it complains "While reading gs_cidfm.ps: | 13:33.00 |
| | Error: /rangecheck in --.dicttomark-- | 13:33.00 |
| | " and I tried using Fontmap.GS "/(Times New Roman) /TimesNewRoman ;" and have TimesNewRoman in cidfmap, and it simply complains "Can't find CID font "Times New Roman"" | 13:33.00 |
| | I know the solution is probably really simple and straightforward, but I can't figure it out. | 13:33.17 |
| kens | '/' introduces a name, if you are converting a string to a name you don't put a '/' there | 13:33.24 |
| chrisl | replace "/(Times New Roman)" with "(Times New Roman) cvn" sans quotes | 13:33.49 |
| kens | Oh and you din't put 'cvn' to convert the string to a name | 13:33.56 |
| bananapie | "(Times New Roman) cvn << /FileType /TrueType /Path (/usr/share/fonts/truetype/msttcorefonts/TimesNewRoman.ttf) /SubfontID 0 /CSI [(Identity) 0] >> ;" | 13:35.17 |
| | without the quotes ;) | 13:35.25 |
| chrisl | Indeed | 13:35.31 |
| bananapie | thanks. | 13:47.51 |
| chrisl | You got it working? | 13:48.04 |
| kens | Did it work now ? | 13:48.06 |
| bananapie | Yes. | 13:51.44 |
| kens | Great! | 13:52.00 |
| bananapie | I'll post the final line in a few moments | 13:52.13 |
| | I made a stupid mistake that I am fixing. | 13:52.21 |
| kens | Story of my life.... | 13:52.43 |
| bananapie | :D | 13:53.28 |
| | Thanks for your help guys. | 13:53.32 |
| | "(Times New Roman) cvn << /FileType /TrueType /Path (/usr/share/fonts/truetype/msttcorefonts/TimesNewRoman.ttf) /SubfontID 0 /CSI [(Identity) 0] >> ;" was wrong because the file was actually Times_New_Roman.ttf | 13:53.46 |
| kens | oh well, at least its easy to fix :-) | 13:54.16 |
| bananapie | The PDF rendering is still slightly off, but it's probably because I didn't list bold and italics in the cidfmap | 13:54.21 |
| | thanks a lot kens and chrisl | 13:54.24 |
| kens | You're welcome | 13:54.31 |
| chrisl | No problem | 13:54.36 |
| | bananapie: Remember, you are (probably) using a substitute font, so you may never get the rendering *exactly* right. | 13:55.38 |
| bananapie | oops, wrong button | 13:56.05 |
| chrisl | bananapie: You only missed me saying: Remember, you are (probably) using a substitute font, so you may never get the rendering *exactly* right. | 13:56.41 |
| bananapie | ok thanks | 13:56.51 |
| | I am comparing the rendering to evince which is, as far as I know, using the same fonts on the machine. | 13:57.33 |
| kens | Well, it might be.... | 13:57.51 |
| chrisl | Well, it should close then. | 13:58.01 |
| bananapie | There is still some funky spacing in the letters in the title. The rest of the text is close enough that most people wouldn't notice | 13:58.34 |
| chrisl | Well, my stabbings at the pdf14 compositor don't seem to have introduced problems - so that might be a good sign | 13:59.26 |
| kens | Its certainly better than the opposite.... | 13:59.44 |
| | bananapie : is the file publicly available somewhere ? | 13:59.58 |
| bananapie | The pdf says 'version 1.5', btw. It uses the h-identity-16 thingy. | 14:00.09 |
| | with utf16 | 14:00.12 |
| | This makes it work almost perfectly : | 14:00.23 |
| | (Times New Roman) /TimesNewRoman ; | 14:00.24 |
| | /TimesNewRoman << /FileType /TrueType /Path (/usr/share/fonts/truetype/msttcorefonts/Times_New_Roman.ttf) /SubfontID 0 /CSI [(Identity) 0] >> ; | 14:00.33 |
| | But (Times New Roman) << /FileType /TrueType /Path (/usr/share/fonts/truetype/msttcorefonts/Times_New_Roman.ttf) /SubfontID 0 /CSI [(Identity) 0] >> ; , causes funky spacing issues. | 14:00.56 |
| kens | You're missing a 'cvn' in that last one, I hope that's a paste error :-) | 14:01.37 |
| | Or are you really substituting like that ? | 14:02.10 |
| | TO be honest, its kind of hard to comment without seein a concrete example | 14:02.42 |
| bananapie | I can send you the file over private chat if you don't show it to anyone... | 14:03.00 |
| kens | If you mail it to ken.sharp at artifex.com it'll reach me hopefully without any translation going on | 14:03.41 |
| | I will delete it afterwards | 14:03.51 |
| | Of coruse, my version of MS Times New Roman may not match yours precisely either :-) | 14:04.23 |
| bananapie | I sent you the file with my notes. | 14:05.57 |
| kens | OK let me go look | 14:06.03 |
| | GOt it, give me a minute | 14:06.16 |
| bananapie | it doesn't load with Ubuntu 14.04's included ghostscript, I had to compile because of Identity-H-utf16. | 14:06.20 |
| kens | WSell I'll be using the HEAD of Git and I'm on WIndows anyway :-) | 14:07.00 |
| | So there are 4 missing CIDFOnts, Arial, Arial,Bold, Times New Roman and Times New Roman-Bold | 14:08.05 |
| | OK so the default font substitution **really** messes it up :-D | 14:09.00 |
| chrisl | Identity-UTF16-H is a custom CMap. unique to Ghostscript, And because Ubuntu (and other distros) share CMaps between GS and all the other PDF consumers, they only have the standard Adobe CMap set | 14:09.43 |
| bananapie | The first 3 words should be "Resume de votre" in the title. Both es in the word 'Resume' have accents pointing up | 14:10.13 |
| kens | Its OK I cna view it in Acrobat | 14:10.41 |
| bananapie | with msftt-core-fonts on Ubuntu, it renders perfectly in evince with no patches. | 14:11.08 |
| | at least it seems perfect. But Ghostscript screws the pooch on this one. | 14:11.33 |
| | :d | 14:11.37 |
| chrisl | There could be many, many reasons for that, the missing CMap being only one | 14:12.19 |
| kens | well there is no Times_New_Roman.ttf in mY windows fonts, btu there is a times.ttf | 14:12.52 |
| chrisl | bananapie: my guess is that, at best, you end up with the default CIDFont fallback of DroidSansFallback | 14:14.02 |
| bananapie | yes. It tried to fallback to 'Adobe-Identity' and landed on 'DroidSansFallback' | 14:14.27 |
| | front 'pdffonts' | 14:15.52 |
| | name type encoding emb sub uni object ID | 14:15.52 |
| | ------------------------------------ ----------------- ---------------- --- --- --- --------- | 14:15.52 |
| | Times New Roman CID TrueType Identity-H no no yes 3 0 | 14:15.52 |
| | Times New Roman,Bold CID TrueType Identity-H no no yes 4 0 | 14:15.52 |
| | Arial,Bold CID TrueType Identity-H no no yes 5 0 | 14:15.54 |
| | Arial CID TrueType Identity-H no no yes 6 0 | 14:15.56 |
| chrisl | bananapie: so, here's the rant: Embed the font! To PDF really portable, *always* embed the font(s) - that way you know that the correct font is always available to the interpreter | 14:16.16 |
| kens | Hmm I can't seem to get GS to use my modicfied cidfmap :-( | 14:16.23 |
| chrisl | kens: with -I ? | 14:16.49 |
| kens | yeah | 14:16.54 |
| | -I/ghostpdl/Resource/Init | 14:17.01 |
| bananapie | it's fine, I found the issue | 14:17.04 |
| chrisl | kens: try a relative path? | 14:17.27 |
| kens | Umm, I could try that I guess | 14:17.42 |
| bananapie | http://pastebin.com/KUbA0hj0 | 14:17.45 |
| | The spacing issue was that the bold fonts were not defined in the cidfmap | 14:18.01 |
| | I added the two bolded fonts as suggested by kens, and it worked. | 14:18.15 |
| kens | OK well that would help anyway of course. | 14:18.35 |
| bananapie | I wasn't getting any warnings about Arial,Bold and Times New Roman,Bold because it was substituting Times New Roman,Bold with Times New Roman | 14:18.35 |
| | that's why the two different syntaxes earlier made a difference. | 14:19.02 |
| kens | chrisl sorry, when I say it doesn't work I mean it throws me a PostScript errror | 14:19.20 |
| chrisl | Oh, well, probably a typo then..... | 14:19.39 |
| kens | Yes it was sorry I wasn't clearrer | 14:20.23 |
| | Anyway, since the problem is fixed, I'll delete the file. | 14:20.58 |
| chrisl | NP, it's just I messed with the file permissions a while back, so...... got a little paranoid | 14:21.14 |
| kens | It was my own fault, I was just slow to spot it | 14:21.31 |
| bananapie | Thanks kens and chrisl. Your help is very much appreciated. | 14:21.47 |
| kens | NP | 14:21.57 |
| bananapie | I added my solution to the web for the sake of https://xkcd.com/979/ | 14:22.05 |
| kens | :-D | 14:22.28 |
| bananapie | is there a handy command that I can run on a PDF file that will list the fonts that will be sustituted if converted in gs ? | 14:24.13 |
| kens | No | 14:24.20 |
| bananapie | :( | 14:24.24 |
| kens | Any font which is not embedded in the PDF file will be substituted by another | 14:24.36 |
| bananapie | ok | 14:24.40 |
| | thanks | 14:24.42 |
| kens | THe substution process for simple fonts is complicated and impossible to describe | 14:24.52 |
| | For CIDFOnts either you have a substitute defined in cidfmap, or you get DroidSans | 14:25.09 |
| rayjj | well, it _is_ /ghost/script | 14:27.45 |
| | oops, my window was scrolled *way* up | 14:28.12 |
| kens | wonders what rayyjj was replying to..... | 14:28.39 |
| rayjj | nm | 14:28.48 |
| chrisl | rayjj: do you know if the device "finalize" method is something we still use, or should we use the finalize method in the struct descriptor? | 14:29.38 |
| rayjj | chrisl: hmm... I'd have to research that | 14:30.12 |
| kens | device finalize methods are *definitely* executed when the device is freed | 14:30.36 |
| chrisl | rayjj: well, none of the device init macros allow the device finalize to be set, which is why I ask | 14:30.59 |
| rayjj | chrisl: some of the devices have specific finalize functions, but looking at them, they are all hooked into the structure descriptors, *not* the dev->procs.finalize procedure | 14:35.43 |
| chrisl | rayjj: so I guess that's the preferred method these days - I must have missed those, as I didn't see them | 14:36.24 |
| rayjj | the only usage I see is int gx_device_finalize that calls it if it is non-null | 14:37.03 |
| kens | Yes, that's the one that calls it on a device being freed | 14:37.20 |
| chrisl | rayjj: so, this fixes the mutex leak on error/interrupt: http://git.ghostscript.com/?p=user/chrisl/ghostpdl.git;a=commitdiff;h=0dd8e23 | 14:37.47 |
| rayjj | chrisl: LGTM | 14:38.58 |
| | thanks for finding that one | 14:39.12 |
| chrisl | rayjj: thanks. It was "good" it still manifested on the master code - made it easier to test! | 14:39.45 |
| mvrhel_laptop | tor8: so I have a commit on my mupdf repos for the issue with the forms values. This fixes the issues seen with that one file anyway | 15:49.51 |
| tor8 | mvrhel_laptop: I started looking a bit myself as well, and I think something similar to what you've done is the way forward for now | 15:51.04 |
| | however, pdf_to_utf8 allocates a new string, whereas pdf_to_str_buf returns a pointer to the internal storage | 15:51.35 |
| | so your patch will leak memory | 15:51.41 |
| mvrhel_laptop | ah | 15:51.46 |
| tor8 | mvrhel_laptop: there's a function, pdf_get_string_or_stream which does exactly what pdf_to_utf8 does (except re-encode as utf-8) which is called internally by pdf_field_value | 15:52.25 |
| mvrhel_laptop | ok let me have a look at that | 15:52.49 |
| tor8 | there are a few other spots in the pdf-forms.c that call pdf_to_str_buf which should be replaced with pdf_to_utf8 | 15:53.10 |
| | but changing all of them does imply having to free the pointers somewhere as well | 15:53.26 |
| | which is more than I wanted to deal with today | 15:53.53 |
| mvrhel_laptop | tor8: do you want me to just leave this to you and I will push on working with files that have decent text encoding... | 15:54.55 |
| tor8 | mvrhel_laptop: you can do that. I expect to have to deal with this within a week or two anyway. | 15:55.36 |
| | I've spent the day preparing the opengl based viewer for doing separate page and annotation rendering | 15:56.02 |
| | all in preparation to do form filling | 15:56.18 |
| | mvrhel_laptop: not sure, but the functions I just added in the latest commit on tor/master might be of use to you | 15:57.24 |
| mvrhel_laptop | ok thanks tor8 | 16:01.41 |
| kens | night all | 16:02.18 |
| rayjj | interesting. The new Dean of the Engineering College at Mizzou is a woman with a biomedical engineering background: Loboa was associate chair of and professor in the Joint Department of Biomedical Engineering at the University of North Carolina-Chapel Hill and North Carolina State University. | 19:08.26 |
| | I guess emphasis in the Engineering college will be toward that :-) The Bioengineering dept. were the "stars" when I was there because we brought in the most grant money. I'm glad that "it's back". | 19:10.31 |
| | There was some grant mismanagement (to put it in a P.C fashion) and the department had some rough times for a bit | 19:11.18 |
| | Forward 1 day (to 2015/09/10)>>> | |
IRC Logs