| <<<Back 1 day (to 2013/04/04) | 2013/04/05 |
Robin_Watts | hmm. egg door is not racoon proof apparently. | 00:00.13 |
| amazon.com reviews. | 00:00.29 |
| Bedtime for me! night | 00:00.36 |
mvrhel_laptop | good night Robin_Watts | 00:02.04 |
| racoons can be rather agile with their hands | 00:02.27 |
ray_laptop | It seems to me that the cloud printers (at least according to the google cloud print docs) get a URL to the data, where the data type is specified in the accept header of the request made to the printjob URL (the "fileUrl"). | 03:59.29 |
| This says it defaults to PDF data type, but if it supports normal http partial get ("Content-Range : ") access, a cloud printer can read the PDF from cloud and doesn't need to download the entire page | 04:15.51 |
| so, should we develop a cloud virtual web printer (viewer) that displays a page WITHOUT downloading the entire PDF (demo with a 90,000 page PDF that is not Linearized) that just uses HTTP to fetch the job and read using Content-Range gets ? | 04:17.44 |
| seems like this would allow even relatively resource limited (ink jet) printers with an efficient RIP (mupdf) to render in bands with VERY limited memory given some smart caching of PDF Resources. | 04:19.23 |
| about the only large things we'd need to hold are ICC profiles (unless we only use the Alternate tint) or embedded fonts (which might be large). Even image data from the PDF is a "stream" so can be read as needed | 04:20.52 |
| there might be a bit of "seeking" if the PDF wasn't web optimized (Linearized) but to access things like the XREF would usually only take a couple of accesses, then we are off and running | 04:22.34 |
| our "app" could do caching of blocks of the data to minimize back and forth seeking, and a memory limit could be used to tell us when to flush resources that can be reloaded from the PDF (i.e. fonts or other objects) | 04:28.05 |
| and if a job prints slowly, just blame it on the internet access to the cloud ;-) | 04:29.14 |
| (the only ones that can prove it's the printer software are those with a high-speed wired connection to a switched network) | 04:30.32 |
| the fact that cloud printing defaults to PDF gives a lot of advantage to a printer based PDF renderer, particularly if it is able to work using the network rather than needing to download the entire PDF and then start rendering, storing the resources in memory as well as it processes the PDF | 04:33.11 |
| and (IMHO) the less said about XPS on the cloud, the better ! | 04:33.55 |
| enough for tonight. Waiting for the other shoe to drop from cust 532... | 04:35.31 |
soo | Hi everybody, I have a question on mupdf integration with android. My application is working on galaxy tab. But when I used other devices, library seems to be crashing. | 07:28.33 |
kens | You are a little early for the MuPDF developers, can you come back in a couple of hours ? | 07:29.04 |
soo | Sure kens. Thanks | 07:29.38 |
kens | You're welcome | 07:29.49 |
nilli | I'm using a gs command to convert an already existing PDF in RBG color to a copy in CMYK instead, and the weird issue I'm having is that all the pages of the original PDFs are always in landscape orientation while some pages in the copy ends up in portrait orientation. I can't see anything in the command that would cause this behavior, but maybe you guys can. | 08:50.58 |
| the command I'm using is: gs -o output.pdf -sDEVICE=pdfwrite -dUseCIEColor -sProcessColorModel=DeviceCMYK -sColorConversionStrategy=CMYK -sColorConversionStrategyForImages=CMYK input.pdf | 08:51.02 |
kens | -dAutoRotatePages=/None | 08:51.22 |
nilli | I'll add that, thanks | 08:51.39 |
| it works like a charm, thank you :) | 08:57.03 |
kens | You're welcome | 08:57.09 |
Robin_Watts | soo: What other devices? non ARM ones? | 09:17.56 |
kens | I think he's gone Robin | 09:18.06 |
Robin_Watts | so he has. | 09:18.19 |
kens | Hopefully he'll come back | 09:18.23 |
| I didn't feel qualified even to ask questions.... | 09:18.41 |
| chrisl ping | 10:20.22 |
chrisl | kens: on phone - just a minute | 10:20.59 |
kens | NP | 10:21.03 |
chrisl | kens: back..... | 10:24.34 |
kens | Sanity check please... | 10:24.41 |
chrisl | Okay | 10:25.04 |
kens | Am I correct in thinking that in PDF Word spacing (Tw) is only applied to the 0x32 character in *normal* fonts, not 2 byte fonts ? | 10:25.18 |
Robin_Watts | kens: AIUI, yes. | 10:25.44 |
kens | Ah well that seems to be my first bug, the code in pdfwrite is applying it to 2-byte CIDFonts too. | 10:26.17 |
chrisl | Ah, sort of.... In a multibyte font it is applied *if* the 0x32 code is defined as a single byte value. If it defined as a multibyte value (0x0032) it is not applied. | 10:26.57 |
kens | OK this is definitely a multi-byte font | 10:27.12 |
| 2 bytes 0x00 0x32 | 10:27.22 |
| So it should *not* be applied | 10:27.30 |
chrisl | Right so Tw should *not* be applied | 10:27.36 |
kens | OK well that gets rid of some of the problem :-) | 10:27.48 |
| The metrics errors turn out to be rounding differences from the TrueType 2048x2048 as opposed to PDF/PS 1000x1000 design grid | 10:28.22 |
| Since they are all supposed to be ints I can fix that by adding 0.5 and taking the floor(). | 10:28.50 |
| THe differences are minute but they make the resulting PDF file bigger and trigger this movement clamp | 10:29.18 |
| I think I'll do a quick cluster check and see what that change does to our tests | 10:30.13 |
| reboot :-( | 10:38.19 |
soo | Robin As I mentioned before am still having difficulty running on any other devices other than samsung galaxy tab | 11:17.29 |
Robin_Watts | soo: such as? | 11:17.58 |
soo | I tried it on Samsung galxy pro. it didnt work | 11:18.13 |
Robin_Watts | But normal mupdf as downloaded from google play does? | 11:18.34 |
soo | neither in any emulators. | 11:18.41 |
| nope. it wont work. normal app from playstore wont work on 2.2 devices | 11:19.15 |
Robin_Watts | right. The galaxy pro is an armv6. | 11:19.39 |
soo | you see, there are couple of methods somewhere which works on 3.0 and above. so i tweaked | 11:20.13 |
Robin_Watts | There are 2 issues here. | 11:20.46 |
soo | but its running fine on tab. not on any other | 11:20.49 |
Robin_Watts | Firstly, we use API calls that only exist in 2.3 I believe - hence it won't run on 2.2 or earlier. | 11:21.17 |
soo | btw. i have to say a special thanks to you, as all of those updates from our previous chat worked :-) | 11:21.47 |
| robin: no, tab is running on 2.2 | 11:22.11 |
Robin_Watts | Secondly, we only build the lib for armeabi and armeabi-v7a | 11:22.21 |
| oh, well then we work on 2.2, but nothing earlier (details, details...) | 11:22.37 |
| Emulators may well be running as x86 things. | 11:23.04 |
| (Certainly that's the way that the iOS emulators work). | 11:23.18 |
soo | yes. i need just 2.2 and above. i checked the methods for compatibility issues. its alright | 11:23.19 |
| so, are you sure that if i run on an actual device, it would work? | 11:24.00 |
Robin_Watts | soo: I'm not swearing to anything. | 11:24.14 |
| If you look in android/jni/Application.mk there are various APP_ABI lines. | 11:24.32 |
| All but one are commented out. | 11:24.47 |
| try changing the ones that are commented out and rebuild - in particular, try APP_ABI := all | 11:25.05 |
| That may be enough to make it work under the emulator. | 11:25.15 |
| You said: App seems to be stopping while doing this. " Trying to loas lib /data/dat/com.artifex.mupdfdemo/lib/libmupdf.so" | 11:25.35 |
soo | the issue is as I said am only doing this as my project, i wont be having lots of devices to test with. :-( | 11:25.37 |
| yes | 11:25.43 |
| exactly | 11:25.46 |
Robin_Watts | libmupdf.so is the shared library with all the native code in. | 11:26.00 |
| I'd expect that to fail exactly there if the emulator was an x86 based thing. | 11:26.23 |
soo | ok | 11:26.39 |
Robin_Watts | and I'd expect that rebuilding with APP_ABI := all would fix that. | 11:26.43 |
| at the expense of making the apk bigger cos it has to have more platforms in. | 11:27.16 |
soo | I can open that mk file with any text editor right? | 11:28.09 |
| Application.mk or android.mk? | 11:29.31 |
Robin_Watts | android/jni/Application.mk - yes, with a text editor | 11:32.27 |
soo | checking. . . | 11:33.21 |
| not yet working on galaxy pro | 11:38.19 |
| You said something about uncommenting something | 11:42.49 |
Robin_Watts | soo: In android/jni/Application.mk, what did you change? | 11:43.13 |
soo | Can you please tell me which line was that | 11:43.25 |
| APP_ABI := all | 11:44.09 |
Robin_Watts | Right, and that's the only APP_ABI line that doesn't have a # at the start ? | 11:44.29 |
soo | yes | 11:44.46 |
| its not commented | 11:44.59 |
Robin_Watts | right, that's what I was talking about. | 11:45.02 |
| so I don't have any other immediate suggestions. | 11:45.14 |
soo | this was supposed to make it work on galaxy pro and emulators right? | 11:46.36 |
Robin_Watts | I was hoping so. | 11:46.45 |
soo | ok | 11:48.14 |
| issue is still there in emulators as well | 11:49.31 |
| :-( | 11:50.02 |
Robin_Watts | Googling, I can't see any reason for this not working. | 11:50.04 |
soo | ok | 11:50.12 |
Robin_Watts | I don't have a Galaxy Pro, nor can I get one. | 11:50.16 |
| so I'm at a loss. And I don't have time to go on a snark hunt at the moment, sorry. | 11:50.32 |
soo | atleast if I could run it in a 2.2 simulator it would have been great | 11:52.38 |
Robin_Watts | mupdf runs under the standard android emulator. | 11:53.31 |
soo | 2.2 as well? | 11:54.02 |
Robin_Watts | Don't know. I use Froyo here, which I think is 2.3 | 11:54.36 |
soo | btw APP_ABI := ALL or APP_ABI := all ? | 11:55.36 |
Robin_Watts | lower case, I believe. | 11:56.11 |
soo | 2.3 is gingerbread :-) | 11:56.34 |
Robin_Watts | OK, so I routinely use a 2.2 emulator. | 11:56.54 |
soo | ok | 11:57.21 |
| also there was only one app_abi in my mk file | 11:57.57 |
| just making sure if I edited the right file | 11:58.21 |
Robin_Watts | This pdf_image -> fz_image change has turned into a bit of a trip down the rabbit hole. | 14:49.31 |
tor8 | Robin_Watts: you're in a maze of hairy code, all alike. | 14:52.40 |
Robin_Watts | I have been eaten by a grue. | 14:54.29 |
tor8 | There's a lamp in the living room! :) | 15:00.21 |
| I need to grab some food. Not grue stew, mind you :) I'll check in later in case you want to chat more about the image stuff. | 15:00.37 |
Robin_Watts | tor8: I think it's coming together slowly... | 15:01.23 |
mhfan | hi, Robin | 15:15.13 |
Robin_Watts | hi | 15:15.30 |
ghostbot | what's up, robin_watts | 15:15.30 |
mhfan | i found u pushed a new branch(robin/progressive) to hold progressive loading patch recently. i'm very interested in this feature. but i failed to compile on the progressive branch. the compiler complains there are some "undefined references". so i guess the patch is not complete, mayby u foget to add new file into the git? such as fitz/stm_progressive.c mensioned in win32/libmupdf.vproj. | 15:15.36 |
Robin_Watts | mhfan: That's entirely possible, yes. | 15:15.54 |
| Just a mo. | 15:15.58 |
| mhfan: New version of the branch there. Does that solve it? | 15:19.21 |
mhfan | I'll try it. many thanks! | 15:20.09 |
Robin_Watts | hmm. Not working for me. | 15:20.23 |
| ok, better version there that actually compiles :) | 15:26.31 |
mhfan | got it. thanks. | 15:30.05 |
Robin_Watts | Do we really have 10 aws nodes idling at the moment ? | 16:21.05 |
henrys | Robin_Watts: something is amiss - my job is running but everything is idle. | 16:31.15 |
Robin_Watts | Your job has just finished... | 16:32.05 |
| ok, so those aws nodes are actually in use. | 16:33.07 |
henrys | okay oh wow it is using those nodes | 16:33.13 |
| maybe that was triggered because I submitter 4 commits? | 16:34.08 |
Robin_Watts | maybe. | 16:34.20 |
henrys | yup my test ran on 29 nodes | 16:37.16 |
| have to be getting pretty far along on the "more-nodes" diminishing returns curve | 16:38.22 |
kens | G'night all | 17:01.51 |
Robin_Watts | have a good weekend kens | 17:02.04 |
| YES! Got in before he disconnected :) | 17:02.12 |
kens | :) | 17:02.17 |
marcosw | henrys: yes, if there are 3 or more cluster jobs pending we launch 10 amazon nodes. They reduce the total run time for a full cluster run from ~20 minutes to ~14 minutes at a cost of $1.15/hour (total, not per node). Pending mupdf cluster jobs aren't counted, but user cluster jobs are. | 18:21.43 |
| I have code set up to launch 20 amazon nodes; I figure we won't need those until the week before the staff meeting :-) | 18:22.29 |
henrys | marcosw:okay sounds good. | 18:23.11 |
| marcosw:did you use PACIFY VALGRIND for the pcl bugs also? | 18:27.14 |
Robin_Watts | henrys: I didn't think that marcosw had used PACIFY_VALGRIND or ENABLE_VALGRIND for any of the memento runs so far (not least because ENABLE_VALGRIND caused problems until just a day or two ago) | 18:35.35 |
| s/memento/test/ | 18:35.47 |
henrys | thanks | 18:37.13 |
marcosw | henrys: Robin_Watts is correct, I'm re-running the valgrind tests with PACIFY_VALGRIND and ENABLE_VALGRIND enabled at the moment. I'll close the valgrind bugs that don't show problems. | 18:37.47 |
henrys | okay marcosw | 18:38.38 |
| marcosw:I think 693835 will go away | 18:40.11 |
| and 834 | 18:40.42 |
| Forward 1 day (to 2013/04/06)>>> | |