| | <<<Back 1 day (to 2016/04/24) | 20160425 |
| nemo | So. I wanted to experiment w/ reducing size possibly simplifying rendering of a pdf by stripping embedded fonts | 15:02.36 |
| | ghostscript seemed the natural way to do that | 15:02.41 |
| | but, well, every time I try to do something w/ gs I end up spending half an hour on the man page | 15:02.57 |
| | didn't run into a simple commandline that did *only* that in a quick search, so thought I'd ask here | 15:03.19 |
| kens | Beware, Ghostscript and the pdfwrite device do not simply 'manipulate' a PDF file, they create a brand new PDF file where only the visual appearance is likely to be consistent. | 15:03.27 |
| | There is no method in GS+pdfwrite to remove a font | 15:03.46 |
| | You might be able to do it with the NeverEmbed array | 15:04.32 |
| | I haven't specifically tried that | 15:04.39 |
| nemo | mm | 15:04.42 |
| | kens: I knew a new pdf was created, but I thought it could be kinda like w/, oh, movie encoding where you can chose to copy audio or video but just shift container | 15:05.09 |
| kens | Which version of GS are you using ? | 15:05.10 |
| | nemo, no, that's not how it works at all | 15:05.24 |
| nemo | that is, I'd generate a new pdf but it wouldn't change any of the internal image compression, vectors, whatever | 15:05.26 |
| | just new pdf - fonts | 15:05.31 |
| kens | If you are using the latest version, 9.19, then read the 'Overview' in VectorDevices.htm | 15:05.39 |
| nemo | heh. 9.10 is installed, but can do a new build I guess | 15:06.05 |
| | have had to do that before.. | 15:06.13 |
| kens | 9.10 is a little old now | 15:06.18 |
| nemo | so, can gs do roughly what I'm hoping for, before I burn too much time on this? | 15:06.49 |
| kens | THere is a NeverEmbed distiller parameter which will take a list of fonts which should not be embedded in the output. You could probably use that. No guarantees on what the final result will look like though | 15:06.56 |
| Robin_Watts_ | nemo: MuPDF allows you do manipulations on pdf files where the internal compression, vectors etc, are unchanged, but even that won't let you strip fonts. | 15:06.57 |
| nemo | kens: the goal was mostly to reduce load time / render time of a simple pdf in PDF.js | 15:07.03 |
| | hm... I wonder if this is amenable to just manual editing in a hex editor or something ⺠| 15:07.23 |
| kens | -c "<</NeverEmbed [/NameOfFont]>> setdistillerparams -f | 15:07.27 |
| nemo | and yeah, might look like crap, just wanted to see | 15:07.45 |
| kens | If you delete data in a PDF file then the xref table offsets will be incorrect, makign it an invalid file | 15:07.52 |
| nemo | kens: PDF.js already has a fallback to helvetica for the embedded gothic apparently | 15:07.58 |
| kens | Well, see the command above | 15:08.10 |
| nemo | tries | 15:08.14 |
| kens | The codumentation notes, in particular (note 13) that you cannot use NeverEmbed with the ps2pdf script | 15:09.16 |
| nemo | in this particular case going from powerpoint â PDF | 15:09.41 |
| | durn. /NeverEmbed doesn't take wildcards. was hoping ⺠| 15:10.31 |
| | I see someone else tried that too from a search | 15:10.36 |
| kens | No, you must specify the font name(s) precisely | 15:10.46 |
| | Oh, and if the font is a subset you will need to include the subset prefix | 15:12.32 |
| tor8 | kens: nemo: with mutool run you could save http://pastebin.com/raw/FbHGi5nq and run that script to delete embedded font files | 15:25.00 |
| | it will go *very* badly if you have identity-h encoded fonts though. | 15:25.09 |
| kens | That kills all fonts ? | 15:25.34 |
| tor8 | it loops through all objects in the file, if it has /Type/Font then it nukes the /FontFile entries in the font descriptor | 15:26.10 |
| kens | Right | 15:26.32 |
| tor8 | you'll probably want to run it through pdfclean -g afterwards, since the script doesn't actually remove the objects it just unlinks them | 15:27.58 |
| nemo | gm | 15:28.06 |
| | er | 15:28.07 |
| | hm | 15:28.08 |
| | I've never run mutool before. how do I pass it your script? | 15:28.16 |
| | can't hurt to try this | 15:28.18 |
| tor8 | nemo: save it as killfonts.js and then run "mutool run killfonts.js input.pdf output.pdf" | 15:28.40 |
| | mutool run is very new, you'll need the latest mupdf release. preferably from git. | 15:28.59 |
| nemo | ah | 15:29.13 |
| | how about this 1.9 release timestamped the 19th? | 15:29.56 |
| kens | you want 1.9a | 15:30.06 |
| tor8 | nemo: 1.9 should also work, I still haven't uploaded a new windows build | 15:30.21 |
| | for the 1.9a fix | 15:30.31 |
| kens | chides tor8 | 15:30.52 |
| nemo | just gonna be a source build | 15:30.56 |
| tor8 | nemo: you might need to change the 'new PDFDocument' line to 'new mupdf.PDFDocument' | 15:31.52 |
| tor8 | dinners. | 15:32.32 |
| nat54 | I was having problems building mupdf for android with ndk r10e and r11. Most errors are from ../thirdparty/mujs/jsmath.c with bunch of undefined reference errors. Can some one help me out please?? | 18:30.40 |
| nemo | nat54: folks who were helping me disconnected 2h ago | 18:41.27 |
| | nat54: mupdf website does have android prebuilts, but if you're missing deps I guess you're just gonna have to track down and install 'em | 18:41.56 |
| nat54 | memo:can you help me? | 18:42.32 |
| | Ya I know that, but just trying to build my self like removing some cjk fonts. | 18:44.26 |
| nemo | shrugs | 18:46.50 |
| | not really no | 18:46.55 |
| | not too familiar w/ building for android | 18:47.05 |
| | if I was maybe our Hedgewars android release wouldn't be 4 years out of date | 18:47.14 |
| Robin_Watts_ | nat54: (For the logs) probably that you are building with the 64bit version of the ndk. | 19:01.52 |
| rayjj | mvrhel: (for the logs) I am trying to enable the UseFastColor "nocm" mode when the ICC input profile data_cs is one of gsRGB, gsGRAY or gsCMYK. It is working, EXCEPT I am tripping over a memory issue | 20:59.52 |
| nat54 | nemo: now it is working with ndk r10e. Thanks. | 21:01.45 |
| rayjj | mvrhel: specifically, the link_handle is opaque to the GC (as it generally should be for a "real" cms), so I changed its allocator over to mem->non_gc_memory. That helps, but the nocm_link_t has a "*pis" element in it, and that is in GC memory which can be fiddled with by restore, or reloc | 21:02.45 |
| | It seems like the code is actually copying those into a "fake" imager_state, so I think all I will need is to make those also be in non_gc_memory. Does that sound right ? | 21:08.24 |
| | mvrhel: (for the logs) Thanks for the phone call. I am trying changes to use non_gc_memory for the nocm_link_ structure and its components | 22:13.41 |
| | Forward 1 day (to 2016/04/26)>>> | |
IRC Logs