| <<<Back 1 day (to 2014/09/04) | 2014/09/05 |
rayjj | gotta love http. I started a download 2+ hours ago, shut down my laptop, opened it back up and it picked up and finished :-) | 01:23.32 |
| I haven't had anywhere the same experience with being able to re-connect SSH sessions (which is _supposed_ to be possible) | 01:24.35 |
snocean | what's this | 04:21.14 |
leena | I saw in the documentation a section for "Rendering stack operators". http://ghostscript.com/doc/current/Language.htm. The layers its referring there, is it same as layers in Adobe Illustrator? If not, whats is this layers and groups about? | 06:33.37 |
kens | leena the purpose of those operators is described in the documentation you refer to: | 07:06.01 |
| "The interpreter state is extended to include a (per-context) rendering stack for handling transparency groups and masks (generically, "layers")." | 07:06.01 |
| Its not the same thing as Illustrator layers | 07:06.14 |
| While PDF has some concepts which are similar to the use of layers in Adobe applications they are not precisely the same. | 07:06.38 |
| If you want to be able to open a PDF or EPS file in Illustrator and have it maintain the layers that were in the original document you are going to have to use Illustrator to create the PDF or EPS file. | 07:07.38 |
richud | hi | 08:50.00 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 08:50.00 |
chrisl | richud: I had a look at a the gstoraster source, and it just calls the "gs" executable, so if you keep your existing gstoraster exe, you should be able to use any gs version | 08:51.20 |
richud | ah cool | 08:51.50 |
| i just have compiled in gdb to my little system | 08:52.09 |
chrisl | But, fwiw, here's a 9.07 I built here: http://www.ghostscript.com/~chrisl/gs-i686-linux-gcc | 08:52.17 |
richud | i am breaking at main | 08:52.19 |
| 139 i have gs_malloc_release(mem); | 08:53.12 |
| 135 exit_status = 255 | 08:53.22 |
| (about 9 steps after main) | 08:53.43 |
chrisl | What's the value of "code"? | 08:54.47 |
richud | hang on i will rerun it | 08:55.18 |
chrisl | richud: are you actually running an i386? | 08:57.33 |
richud | code is 0 at gs_main_init_with_args(minst, argc ,argv); | 08:57.34 |
| then if (code >= 0) it is -15 | 08:57.51 |
chrisl | that's a rangecheck error | 08:58.08 |
richud | its my embedded i386 system (uclibc) on an i5 test pc | 08:58.20 |
chrisl | So which is it: i386 or i5? | 08:59.26 |
richud | sorry I am not sure what you are asking exactly, the OS is all compiled as i386 | 09:00.07 |
| linux kernel/busybox/ <stuff> gdb, ghostscript | 09:00.32 |
chrisl | And is the intention to use an i386 cpu? | 09:00.42 |
richud | its to cover various things from say a p3 to a core i7 | 09:01.09 |
| so i stuck with i386 as target | 09:01.19 |
| ive been using it for years (its for pc deployment) | 09:01.37 |
chrisl | Well, you can try the exe I linked above - although it says i686 I'm not sure that's accurate now..... I suspect it's more basic x86 | 09:02.35 |
richud | ok | 09:02.59 |
| (i just tried running gs with --help and after printing the help text its error code -110) | 09:03.20 |
chrisl | Hrm, I don't see how it would get -110...... | 09:05.47 |
| Oh, actually, that seems to be correct, according to ierrors.h | 09:06.08 |
kens | Yes, -110 is e_Info, ie you selected --help | 09:07.03 |
| Its a special case error for windowed executables (well WIndows really) | 09:07.25 |
chrisl | Yeh, I misread the final switch statement in main() - I initially thought all those values would get overwritten by that switch | 09:08.04 |
richud | chris: that file you linked works | 09:09.04 |
chrisl | Okay, here's the makefile and arch.h I ended up with to build that: http://www.ghostscript.com/~chrisl/gs-makefile_arch_h-x86.tgz | 09:10.12 |
| You should be able to just tweak the compiler name and executable name to your needs | 09:11.03 |
richud | right ill have a fiddle with that | 09:11.12 |
chrisl | BTW, did you get cups to cross compile? | 09:11.24 |
richud | yep | 09:11.33 |
| everything else is ok | 09:11.39 |
chrisl | because gstoraster links to the cups libs, amongst other stuff | 09:12.07 |
richud | well fingers crossed if i can get gs working then that should still all be ok :) | 09:12.42 |
chrisl | But as I said, you really shouldn't need to build gstoraster, just replacing the "gs" exe should be sufficient | 09:12.49 |
richud | hmm some differences in the arch file | 09:14.36 |
chrisl | Like? | 09:16.30 |
richud | #define ARCH_ALIGN_LONG_MOD 8 #define ARCH_ALIGN_PTR_MOD 8 | 09:16.48 |
| #define ARCH_LOG2_SIZEOF_LONG 3 | 09:16.59 |
| #define ARCH_SIZEOF_PTR 8 | 09:17.13 |
| those 4 | 09:17.15 |
chrisl | So you didn't create a platform appropriate arch.h file then - clearly, that's not going to work..... | 09:17.51 |
richud | i was under the impression genarch did that? | 09:18.08 |
chrisl | genarch runs on the host, how can it know what the requirements of the target are? | 09:18.39 |
richud | hmm | 09:18.53 |
| look like my ignorance then is the problem | 09:20.01 |
chrisl | And it is mentioned in the gs FAQ: "generate arch.h manually and modify gs/base/lib.mak to copy your arch.h rather than generate a new one" | 09:20.07 |
| One of the "new features" we added to aid cross compiling was to allow genarch to pull in a pre-defined arch.h instead of generating one | 09:21.19 |
| I must admit, I thought we'd pointed you at that before...... | 09:22.01 |
richud | to be honest I read the faq about that but hadnt really understood | 09:22.26 |
| let me fix and retry everything | 09:23.08 |
chrisl | Note that the FAQ is a little out of date, since more recent versions (including 9.07) have a "TARGET_ARCH_FILE" in the Makefile: you set that to your predefined arch.h and genarch will use that (no need to modify lib.mak). If you search the Makefile I gave you, you can see its use | 09:25.21 |
richud | i used target_arch_path, it now runs but segfaults, i am just going to check what i have done , carefully. | 09:40.29 |
| segfaulting , lcms2/src/cmsio0.c no such file or directory (line 857) | 09:45.38 |
kens | Missing iccprofiles directory by the sound of it | 09:46.30 |
richudd | sorry | 09:47.56 |
| should probably use a proper irc client :) | 09:48.08 |
| ken: shouldnt that get all compiled into the romfs by default? | 09:48.29 |
kens | If you get it right..... | 09:48.39 |
richudd | chris: could you paste me your compile line? | 10:00.01 |
chrisl | compile line? "make" | 10:00.31 |
richudd | hehe | 10:04.18 |
| configure | 10:04.25 |
| what options? | 10:04.51 |
chrisl | --without-ijs --without-libpaper --without-libidn --with-libiconv=no --with-local-cups --disable-dbus --disable-fontconfig --disable-sse2 --disable-contrib --enable-little-endian | 10:05.45 |
| But remember I then tweak the Makefile after configure runs | 10:06.02 |
richudd | ok thanks (presume thats from 9.1x if has endian switches) | 10:06.10 |
chrisl | Er, oh, just a sec - wrong directory..... | 10:07.17 |
| 9.07 was the same, but without the --enable-little-endian | 10:09.08 |
richudd | ok thanks, i will have another fiddle in a bit , AFK | 10:09.28 |
chrisl | richudd: it occurs to me that you might be overcomplicating this a bit, though...... | 10:10.02 |
richudd | chris: i ran genarch on the target and one other thing needed changing however its still segfaulting in same place, at lcms2/src/cmsio0.c line 857 (which is return Icc -> ColorSpace; | 12:37.16 |
kens | Which suggests 'icc' is not a valid memory address. Which tells you something went wrong prior to that. Time for the debugger | 12:37.29 |
richud | ken: afraid im just a beginner with gdb, could you kindly suggest what i should be looking for? (incidently strace giaves the si_code=SEGV_MAPERR , which is something to do with malloc?) | 12:47.03 |
kens | icc is a pointer to a structure, the fact that icc->ColorSpace causes a seg fault tells you that the process cannot access the memory pointed to by 'icc'. Somewhere prior to this poitn 'icc' should have been allocated and teh problem is likely to be that it has not been (something went wrong). In order to figure out what is wrong, you need to find where 'icc' is allocated, which might be quite a trial, but you can track backward | 12:48.03 |
| s through the call stack to see where it was passed in. THen you need to grep/read teh source to find out where its allocated. Set a breakpoint at the allocation, and see why it didn't get allocated. | 12:48.03 |
| Given that Chris' binary worked for you, it seems to me that you still have a build problem. | 12:48.21 |
richud | hmmm , his was static. I will have to do some learning of gdb as this is all rather over my head (not a c programmer) | 12:51.52 |
kens | If you're not a C programmer, then debugging this is likely to be a 'challenge'. Ghostscript is difficult even for those of us with plenty of experience in C, PostScript and Ghostscript itself | 12:51.58 |
richud | :) | 12:52.00 |
| the irony is all i want to ultimately do is print a label out | 12:52.07 |
| bt shows previous function as gscms_get_input_channel_count | 12:53.53 |
| default_grey.icc is the profile | 12:54.14 |
| gray even on the other side of the pond | 12:57.54 |
kens | Its not really going to be possible for us to debug this over an irc channel.... | 12:59.02 |
richud | you are more than welcome to ssh into the box and look, its running from ram | 12:59.16 |
kens | Not me, I don't like gdb. | 12:59.31 |
richud | hehe | 13:00.07 |
kens | where don't like == can't use | 13:00.07 |
| I use ddd under protest, or VisualGDB | 13:01.32 |
| Which reminds me, I must pay for VisualGDB | 13:01.53 |
tor8 | Robin_Watts: the jpeg_finish_decompress to jpeg_abort change ... might that not cause us to go wrong on inline jpeg image data if we don't consume the image trailer? | 13:18.21 |
| Robin_Watts: RunLengthDecode is bits-per-pixel agnostic, it is not guaranteed to be 8-bpc | 13:20.57 |
| couldn't you just test image->bpc == 8 instead of using buffer_is_8bit? | 13:22.02 |
richud | ken: i just tried 9.14 and it worked | 13:24.38 |
kens | OK well from what chri said, you should be OK | 13:26.27 |
richud | could you paste what he wrote I am afraid i lost it when the browser irc thing closed | 13:29.23 |
| gsraster is C not a wrapper script? | 13:29.42 |
kens | chrisl said that gstoraster works with gs from the command line, so if yo uhave a working gs 9.14 gstoraster should work | 13:29.48 |
| Its an executable, I assume written in C | 13:30.05 |
richud | sorry i didnt explain myself properly - is there a way to create gsraster without needing to build 9.07 along with 9.14? | 13:31.25 |
kens | gstoraster is (as far as I cna see) Nothing to do with GS< except that it runs the Ghostscript executable. | 13:31.43 |
| So if you have gstoraster, and a working GS, then you have all you need | 13:31.59 |
chrisl | gstoraster is a completely stand alone exe | 13:32.54 |
Robin_Watts | tor8: RunLengthDecode is always 8 bits, AIUI. | 13:33.18 |
richud | chris: ok, so i should be able to build it out of the 9.07 cups folder without building anything else? | 13:33.55 |
chrisl | Yes | 13:34.04 |
richud | right, ill get crasking with that | 13:34.13 |
| incidently when i just did the test with 9.14 a label shot out of the machine to my amazement :) | 13:34.31 |
| so i guess the rest of it all worked | 13:34.43 |
kens | In PostScript, RUnLengthDecode consumes bytes (i 8-bit data) | 13:34.43 |
tor8 | Robin_Watts: it's like the png predictors ... it works on bytes, doesn't care where the bits of the colors are | 13:34.55 |
chrisl | richud: well, that's a good start! | 13:35.12 |
Robin_Watts | tor8: From the PDFRM: If the image stream uses a filter, the value of BitsPerComponent must be consistent with the size of the data samples that the filter delivers. In particular, a CCITTFaxDecode or JBIG2Decode filter always delivers 1-bit samples, a RunLengthDecode or DCTDecode filter delivers 8-bit samples, and an LZWDecode or FlateDecode filter delivers samples of a specified size if a... | 13:35.13 |
| ...predictor function is used. | 13:35.13 |
tor8 | it doesn't make much sense, but there is nothing to enforce it | 13:35.47 |
| in any case, checking image->bpc should be both easier and independent of the actual compression format of the buffer | 13:38.36 |
| and as such, less prone to future breakage | 13:38.57 |
Robin_Watts | Page 340. | 13:45.24 |
tor8 | Robin_Watts: yes, I have read that section. I don't trust it! (or rather, I don't trust anybody who writes pdf creation software to follow that rule) | 13:46.10 |
Robin_Watts | I agree that RunLengthDecode *could* work with other depths. | 14:01.33 |
richud | chris: oh and btw my arch file had #define ARCH_ALIGN_DOUBLE_MOD 4 when i generated it on other machine and looking around on google I found a list of arch''s somone1 made and their i386 had that too (yours had 8 if you are bothered) | 14:01.56 |
Robin_Watts | Certainly our code assumes that RLD => 8 bits currently, I think. | 14:01.56 |
| Cos we don't store bpc for the RLD case in the compressed_buffer params. | 14:02.16 |
| tor8: As for the jpeg_decompress_finish -> jpeg_abort change... | 14:02.20 |
| I think we're safe, because where it matters we always consume the full amount. | 14:02.28 |
| It's just that we now *can* break out within consuming the whole thing. | 14:02.31 |
tor8 | Robin_Watts: I don't believe that's the case. We only need the BPC in the compressed buffer for flate and lzw because of the predictors | 14:02.46 |
chrisl | richud: that shouldn't be a critical setting, it should just mean we use a bit more memory..... | 14:03.17 |
richud | ok | 14:03.18 |
chrisl | But obviously preferable to get it right.... | 14:03.20 |
richud | chris/ken, many thanks for all your help with getting this working! | 14:03.29 |
chrisl | richud: does that mean you're up and running? | 14:03.31 |
kens | hopes so.... | 14:03.32 |
richud | lol ,well the label came out, i just have to find a nice way of getting gsraster out but i know it wil work one way or the other , so yes | 14:03.34 |
chrisl | richud: actually, IIRC, you might need to fiddle with cups.mak as it normally uses "cups-config" to get the libs, header paths etc...... | 14:03.38 |
richud | yeah i am just looking at that, i am intending to put it back into cups 9.14 source i think | 14:03.40 |
Robin_Watts | tor8: Sure, but if we believe that RLD can have bpc != 8 we need to fix the code that's there already. | 14:03.44 |
| hmm. maybe not. | 14:03.48 |
| OK, so we could just amend my commit to exclude RLDs? | 14:03.49 |
| tor8: Oh, we have image->bpc? | 14:03.52 |
| Yeah, ok. checking image->bpc ==8 instead of buffer_is8bit seems fine. | 14:04.00 |
tor8 | Robin_Watts: we have image->bpc already! it's what I've been trying to get across :) | 14:04.07 |
Robin_Watts | tor8: yes. yes, I see that now. You're right. | 14:11.10 |
t4nk462 | Hi | 14:13.17 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 14:13.17 |
t4nk462 | I have sent an email regarding ReaderView with Vertical scroll in MuPDF. If you have the source code of that vertical ReaderView, please send it to kparunjyothis@gmail.com | 14:14.44 |
Robin_Watts | t4nk462: Hi, yes, we spoke yesterday. | 14:15.01 |
| And via email. | 14:15.06 |
t4nk462 | Yes | 14:15.12 |
Robin_Watts | And i think I made it quite clear that we don't have such a ReaderView. | 14:15.19 |
| I said that if you came here, we might be able to give you some pointers, but we haven't suddenly come up with a new implementation overnight. | 14:16.03 |
t4nk462 | Yes, I have sent an email with changes I made, it works somewhat, if somebody can check that quick & give some pointers, it would be helpful | 14:16.44 |
Robin_Watts | t4nk462: Yes. I forwarded that mail to paulgardiner yesterday. | 14:17.15 |
t4nk462 | Yes | 14:17.29 |
Robin_Watts | The question now, is if he has time to look at the issue. | 14:17.35 |
t4nk462 | Ok | 14:18.00 |
paulgardiner | Sorry not looked yet. Caught up in problems with a release at the moment. Intend to look as soon as that's done. | 14:18.11 |
t4nk462 | Thank you | 14:18.26 |
paulgardiner | ReaderView is definitely the right place. My belief is that nowhere else in the app is there code that assumes a specific page scrolling direction. | 14:19.40 |
t4nk462 | i changed direction now it works but scrolls only if i zoom a little | 14:25.21 |
Robin_Watts | The important stuff looks to be the last couple of clauses in onLayout, right? | 14:27.58 |
Robin_Watts | must *not* get distracted by this. | 14:28.45 |
t4nk462 | yes i changed in onLayout & onFling | 14:28.50 |
| Please refer my email, you can see the changes which I made | 14:29.09 |
| it works vertically now when I made the changes | 14:29.29 |
| but vertical scroll works if I zoom a little | 14:29.43 |
paulgardiner | t4nk462: I think you are close. There is some special casing with being completely zoomed out. We then stop any scrolling in the vertical direction. You'll want to make that stopping in the horizontal. | 14:29.51 |
| Can't remember at the moment where that comes in | 14:30.01 |
t4nk462 | is this in ReaderView itself ? | 14:30.44 |
| I am leaving to office Paul, if you are able to point some place ( which I have not done changes in my email ), please reply to my email. Thank you for your time & help | 14:35.51 |
paulgardiner | okay will do | 14:36.17 |
Robin_Watts | paulgardiner: } else if (cv.getMeasuredHeight() <= getHeight()) { that line ? | 14:40.20 |
paulgardiner | Sounds very much like it | 14:40.39 |
tor8 | Robin_Watts: more demands from the test-device folks just dropped in :( | 14:42.03 |
Robin_Watts | I saw. | 14:42.09 |
kens | Me too.... | 14:42.13 |
| Moon on a stick, right now! | 14:42.23 |
| Maybe you should ask them to suggest ways to improve the performance, they are supposed to be developers. | 14:44.53 |
Robin_Watts | I suspect those use JPX, and that decodes the whole image. There is nothing we can do about speeding them up. | 14:44.56 |
tor8 | developer: n. person who pastes chunks of code he googled on stack overflow and asks random strangers to fix it for him. | 14:46.09 |
kens | :-) | 14:46.19 |
| I'm not sure this customer has reached that certification level yet | 14:47.22 |
tor8 | Robin_Watts: and on JPX we can't even cheat since the PDF colorspace is optional (and often in contradiction with the image data) | 14:47.23 |
Nice_ | d | 14:48.00 |
kens | You can pull the JPX colour space out (GS dos this) without decoding the image, but it doesn't get away for not trusting it | 14:48.09 |
tor8 | for separation colorspaces, we could sample it 256 times and check those values rather than each image sample | 14:48.44 |
| likewise for indexed colorspace images | 14:48.49 |
| but none of that applies to JPX | 14:49.02 |
chrisl | With JPX, in theory, you could decode one resolution level at a time - but the vast majority only seem to have one resolution level anyway..... | 14:50.42 |
tor8 | we should hook it up to the app store and charge in-app purchase money each time we have to decode a JPX :) | 14:51.27 |
kens | chrisl one resolution level and 1 tile :-) | 14:52.46 |
| Again, they could use multiple instances of MuPDF to check groups of pages | 14:53.31 |
chrisl | Can you direct mutool to specific pages? Then they could just just run multiple executables...... | 14:54.29 |
kens | ^^ :-) | 14:54.40 |
| I asked that the other day I think | 14:54.50 |
chrisl | Oh, when I think "instances of mupdf" I think threads....... | 14:56.01 |
kens | I guess maybe threads would work too | 14:56.32 |
chrisl | Threads very much relies on customer having some degree of competence, knowledge, etc, etc | 14:57.12 |
kens | I'm not convinced this bunch will be able to handle multiple processes | 14:57.32 |
Robin_Watts | VectorColorBigSize.pdf looks to be weighed down by sheer volume of line art. | 14:57.43 |
richud | chris: btw i added it back into 9.14 and all ok :) | 14:57.47 |
Robin_Watts | We read in all the paths etc, only to then ignore them. | 14:57.58 |
| Baling using an error would be much faster. | 14:58.11 |
| Also, there look to be some indexed colors in there. | 14:58.18 |
chrisl | richud: fair enough...... | 14:58.42 |
richud | thanks again | 14:59.41 |
rayjj | Robin_Watts: don't you bail once color is detected (or are these monochrome pages -- the worst case) | 15:01.52 |
richud | im off for a beer, ill have one on yours and ken's behalf in thanks | 15:02.08 |
Robin_Watts | currently, we set a flag when color is detected. | 15:02.12 |
| and then if that flag is set don't bother to check anything else. | 15:02.30 |
rayjj | and skip to the next page ? | 15:02.44 |
Robin_Watts | but the interpreter still slaves through all the subsequent page objects, only to have the device ignore them instantly. | 15:02.54 |
| I did have code that set the flag then threw an exception. | 15:03.13 |
rayjj | ah -- so the interpreter doesn't actually bail on the page | 15:03.20 |
Robin_Watts | That worked, but it didn't give a speedup for the file I was testing at the time. | 15:03.24 |
| and it gave a nasty error message on every exception. | 15:03.39 |
| I might reinstate that at some point. | 15:03.48 |
| tor8: Updated version of fix pushed. | 15:05.16 |
henrys | Robin_Watts: feel free to share the work with tor8 if it is getting tiresome. Let's spread the hell around evenly ;-) | 15:58.09 |
Robin_Watts | henrys: No, no, SOT is tiresome. | 15:58.27 |
| The test-device stuff is a nice stress release. | 15:58.41 |
henrys | wonders how we could get tor8 to work on SOT, it would be entertaining to try... | 16:00.14 |
Robin_Watts | I'm closer to the fallout zone than you... | 16:02.15 |
mvrhel_laptop | good morning | 16:04.27 |
davesleep | oh... | 16:09.15 |
| i got a question.. custom fonts when converting PDF to png.. is there a way to ignore the font and replace it? | 16:09.42 |
Robin_Watts | mvrhel_laptop: Morning | 16:09.44 |
| davesleep: That's what happens by default. | 16:09.58 |
| You just get a warning. | 16:10.02 |
kens | davesleep : ignore the font embedded in the PDF ? | 16:10.06 |
davesleep | no, just use a font that's available | 16:10.14 |
kens | THen right now that is indeed what happens | 16:10.25 |
davesleep | hrm.. certain PDFs is just doesnt convert | 16:10.38 |
kens | If you want to use a specific font, set a rep[lacement ujp in fontmap | 16:10.39 |
| davesleep : GS or MuPDF ? | 16:10.55 |
Robin_Watts | If you try and get gs to render a file and it needs a font that isn't there, gs will pick a replacement. | 16:11.01 |
davesleep | exec('convert -density 220 -background white -alpha flatten '.$filedest.' -resize 50% '.$newfileloc."/".$timefile.".png"); | 16:11.11 |
Robin_Watts | No guarantee is made that gs will pick a good replacement :) | 16:11.13 |
kens | THs is neither GS nor MuPDF< it is Imagemagick | 16:11.27 |
davesleep | i dont mind about the style it replaces.. but it seems to pick and choose which PDFs it'll convert | 16:11.28 |
kens | davesleep : try using GS directly, not IOmageMagick, then we can talk to you sensibly | 16:11.46 |
Robin_Watts | davesleep: So you're using imagemagick. We don't speak ImageMagick here. If you want us to help, you'll need to give us a gs commandline. | 16:11.56 |
davesleep | how can i do that inside of my php script? | 16:12.12 |
Robin_Watts | leaves this to kens, as we appear to be saying the same things in different words. | 16:12.15 |
kens | davesleep : I have no idea, I don;t spek PHP | 16:12.31 |
| speak* | 16:12.35 |
davesleep | so im fkd basically lol | 16:13.14 |
| im sure i'll figure it out | 16:13.29 |
kens | I would imagine gd -sDEVICE=png16m -sOutputFile=out%d.png input .pdf woudl be a start | 16:13.32 |
Robin_Watts | s/gd/gs/ | 16:13.49 |
kens | Damn,I mean 'gs' not gd | 16:13.51 |
chrisl | davesleep: can't you just call gs on the command line for testing (not in php)? | 16:14.10 |
davesleep | im on windows | 16:14.21 |
kens | THen use gswin32 or gswin32c | 16:14.33 |
davesleep | so you guys dont know of cases where it would fail to convert some pdfs? | 16:14.55 |
| or will i get a specific error if i use commandline | 16:15.21 |
chrisl | What error are you getting now? | 16:15.48 |
kens | How can we possibly tell ? YOu have given us *nothing* to go on. We do not know if the error is in ImageMagick or Ghostscript, you don't say what version of GS you are using, and you can't give us a command line. We are fumbling in the dark. | 16:15.53 |
davesleep | none | 16:15.54 |
kens | So use Ghostscript, at the very least it will give you some feedback | 16:16.21 |
| Steps to follow: | 16:16.44 |
| 1) Find a PDF file which didn't work | 16:16.44 |
| 2) Use the GS command line above on it | 16:16.54 |
| 3) COme back and say what happened | 16:17.00 |
davesleep | sorry one moment | 16:17.20 |
| yeah im really really tired, friday etc. give me a bit | 16:17.34 |
henrys | 0) take a nap, then 1, 2 and 3 | 16:18.14 |
| ;-) | 16:18.22 |
kens | Good idea, I'll be gone for the weekend by then :-) | 16:18.36 |
mvrhel_laptop | brb | 16:21.45 |
rayjj | great -- I have a PDf from cust 532 (no Creator info) that doesn't show an image in gs, but shows it in Adobe. mupdf totally honks up on the file, however (at least gs is close) | 17:00.22 |
chrisl | rayjj: does it happen just on the sim, or on "normal" gs? | 17:02.43 |
rayjj | chrisl: normal gs. | 17:05.25 |
chrisl | rayjj: well, if you need me to look at it, let me know - I know you've got a lot going on just now | 17:06.58 |
rayjj | HEAD, at that. I've narrowed it down to just the image, and if I keep the form before it, with most of the Contents blanked out, but keeping the Matrix and the 'cm' of that form, then the image is missing. If I zap either the Matrix _or_ the initial cm, the image is OK | 17:07.08 |
| the image is about the only thing that mupdf shows -- it's missing the entire data from the form preceding the one with the image. | 17:08.29 |
Robin_Watts | I'd suspect something being broken about the form then. | 17:09.04 |
rayjj | hmm... the TR in the graphics state seems to be throwing gs off. It has a transfer function that is {1 exch sub} and if I change that to { } then gs is happy, but {0 exch sub} fails ??? | 18:17.23 |
| strange | 18:17.49 |
| even stranger, changing the function to: { exch exch} fails while { } is OK. | 18:30.45 |
| time for RAW_DUMP I guess :-( | 18:31.08 |
Robin_Watts | rayjj: mupdf doesn't handle transfer functions at all. If the form requires a TR to look right, we'll render it wrong | 18:42.06 |
| tor8: For the logs: 2 commits on robin/master. | 22:35.31 |
| They fix the problems you spotted, plus they use errors to let us bale out. Gets the 6300page file that used to take 4 minutes down to 1 minute | 22:36.51 |
| Forward 1 day (to 2014/09/06)>>> | |