| <<<Back 1 day (to 2013/02/26) | 2013/02/27 |
Robin_Watts | xreal: I am fairly sure the answer is no. But try back in 12 hours and ask kens. | 00:01.20 |
xreal | Robin_Watts: okay | 00:01.32 |
Robin_Watts | updating chatzilla. back in a mo. | 00:01.45 |
| mvrhel_laptop: If you get anywhere with the ndes thing, let me know... | 00:13.28 |
mvrhel_laptop | Robin_Watts: ok. digging right now. there is some confusion in the shading code with respect to the device which in this case is a pdf14 device, with yes you guessed it a soft mask | 00:16.17 |
Robin_Watts | oh, fun all round then. | 00:19.55 |
mvrhel_laptop | Robin_Watts: ok a simple fix for the ndes issue, but I may do a little code clean up here too | 00:20.37 |
Robin_Watts | mvrhel_laptop: I'm off to bed in a mo. If you don't get to commit the fix before you quit for the day, can you mail me what you get please? | 00:24.33 |
| That'll let me run tests, do profiles etc. | 00:24.45 |
mvrhel_laptop | Robin_Watts: will do | 00:24.45 |
Robin_Watts | Thanks. | 00:24.49 |
mvrhel_laptop | Robin_Watts: pushing my fix in a bit to see if anything explodes | 00:26.51 |
| oh crap | 00:38.27 |
| Robin_Watts: you still there? | 00:38.31 |
Robin_Watts | I a | 00:41.15 |
| I am | 00:41.16 |
| mvrhel_laptop: ping | 00:48.55 |
mvrhel_laptop | oh Robin_Watts no worries. Get some sleep | 00:49.38 |
Robin_Watts | oh, ok. | 00:49.47 |
sebras | tor8: Robin_Watts: http://goo.gl/dZDoK http://goo.gl/St3DT | 02:54.57 |
| we have droidsansfallback(full) from 2012-feb-15, not the one from 2012-sep-24. | 02:54.58 |
| google appears to have stripped korean from the font in the latest version and | 02:55.01 |
| decided to use another font in this case. | 02:55.04 |
| maybe it is worth changing to that derivative font..? | 02:55.39 |
mvrhel_laptop | Robin the issue with ndes being wrong has been fixed. Please feel free to profile | 04:40.31 |
ray_laptop | found (and fixed) a problem with the SMask diddling. I didn't allow for the case where SoftMask is not in the currentdict, but may be a bit deeper on the dictstack. Needed to use: /SoftMask where pop /SoftMask //true put | 05:51.24 |
| this finds the dictionary that contains SoftMask and replaces its value with //true | 05:51.57 |
| I can safely discard the boolean from 'where' since it will always be true in this function (.execmaskgroup) -- otherwise we wouldn't have entered the function to render the SMask | 05:53.14 |
| I'm surprised that I had so few differences with this error. We'll see how the next run goes. | 05:53.53 |
| who would have thunk that something called "Soft"Mask would be so hard ;-) | 05:54.37 |
kens | Robin_Watts : (for the log) can you look at http://stackoverflow.com/questions/15103022/android-mupdf-for-mips-and-x86-arch please ? | 08:11.06 |
xreal | kens: Can I convert fonts to paths and keep copying chars? Acrobat can do I this. | 10:14.27 |
| DAMN, he's off | 10:14.34 |
kens2 | No I'm here | 10:14.42 |
| just a moment | 10:14.46 |
| I'm replying to a bug report | 10:14.59 |
xreal | nice | 10:15.04 |
kens2 | RIght, sorry | 10:16.49 |
| THe answer is 'not with Ghostscript' | 10:16.59 |
| Acrobat does htis by writing the data twice, once as vectors, and tehn as text, but it writes the 'text' with teh text rendering mode which says 'neither stroke nor fill' | 10:17.39 |
| I can't really see the point..... | 10:17.51 |
xreal | kens2: oh, that's a very usefull feature when you're not allowed to embedd the subset of a font. | 10:18.15 |
kens2 | I would think that the best thing is to get a better font :-) | 10:18.50 |
xreal | kens2: Corporate Identity, you know? :) | 10:19.10 |
kens2 | So if you own the font, then you should be allowed to embed it, if you bought the font, then well, should have checked the DRM :-) | 10:19.37 |
xreal | kens2: It's our University's font but licence says: "Do not embedd when sending PDFs outside University" | 10:20.14 |
kens2 | In any event, we don't support it, and have no plans to do so (it would take a lot of time for very little perceived benefit) | 10:20.28 |
| xreal : well that's a corporate thing, not a DRM | 10:20.46 |
| ANd its daft | 10:20.52 |
| subset fonts are useless from a pirate's POV | 10:21.05 |
xreal | kens2: I know, but that's what the license says ( | 10:21.46 |
| ken2: Is there a way to "emulate" the function? | 10:22.01 |
kens2 | xreal : no | 10:22.08 |
| Or rather, not with GS and pdfwrite | 10:22.28 |
xreal | ken2: Okay, then I'll keep on creating PDFs with Ghostscript and converting them with Acrobat :) | 10:22.38 |
kens2 | LOL | 10:22.50 |
xreal | kens2: Most of the software can perfectly parse Ghostscript PDFs, like Corel, but they can't parse Acrobat documents | 10:23.19 |
| parse => import | 10:23.39 |
kens2 | Possibly because Acrobat is exporting them at a higher level, GS+pdfwrite only uses PDF 1.4 | 10:23.45 |
| Recent versions of Acrobat default to higher levels for better compression (compressed xref and object streams) | 10:24.11 |
xreal | kens2: can Ghostscript to PDF-A ? | 10:24.47 |
kens2 | yes | 10:24.51 |
xreal | ehm, pdfwrite | 10:24.52 |
kens2 | Sttill yes ;-) | 10:24.58 |
xreal | kens2: Nice. Need to look it up. | 10:25.02 |
kens2 | PDF/A-1b | 10:25.05 |
| Not pdf/a-1A | 10:25.13 |
xreal | kens2: I'm using PDF/X-4 a lot now, I like it because of transparency support. | 10:25.32 |
| kens2: PDF/A-1 is more strict, isn't it? | 10:26.09 |
kens2 | There are too many PDF standards :-( | 10:26.11 |
xreal | PDF/A-1a | 10:26.18 |
kens2 | PDF/A-1 is much more strict yes | 10:26.20 |
| PDF/A1-a includes accessibility and formatting which we can't guarantee to be present | 10:26.43 |
xreal | ah okay. | 10:26.51 |
| PDF/X-3 is standard for printing here, but the "transparency rendering" (?) annoys me. | 10:27.05 |
kens2 | I have to rush off, got to be somewhere else in 30 minutes. I'll be back in about 2.5 hours. | 10:27.34 |
xreal | okay, much fun | 10:27.46 |
kens2 | bye all | 10:27.49 |
Robin_Watts | paulgardiner: Hey. | 10:52.40 |
paulgardiner | hi | 10:52.50 |
ghostbot | que tal, paulgardiner | 10:52.50 |
Robin_Watts | How hard would it be for you to do mips and x86 builds of android v8 ? | 10:53.00 |
paulgardiner | Don't know. Was long enough ago that I can't remember what was involved. You may remember better than I, since I think you more recently built v8 | 10:58.39 |
| Ah you built the linux libs? | 10:59.28 |
| Robin_Watts: my annotation creation stuff causes segvs, btw. | 11:01.11 |
| Disappointing: "history | grep v8" doesn't find anything on my current server or the spare | 11:03.15 |
Robin_Watts | I built the linux/mac libs. | 11:04.07 |
| I never built (as far as I know!) the android ones. | 11:04.20 |
paulgardiner | I have v8 checked out under Windows and Linux. I guess it was the Linux one I used, but I have no recollection of what's involved. Probably if I try make somewhere I'll get the ARM ones. | 11:05.40 |
Robin_Watts | paulgardiner: One commit on robin/master | 11:05.47 |
| paulgardiner: You use gyp to make a makefile, then make that, I think ? | 11:06.16 |
paulgardiner | Commit looks fine. I'll have a look later - see if I can suss how to build the other versions. I don't think I've already built them by accident because there are only one set of .a files and no .dll files | 11:15.08 |
Robin_Watts | ok. I just tried a build of MuPDF for all arch's and of course it falls over without having non-arm libs. | 11:15.56 |
| So no hurry, but it'd be nice to support it. | 11:16.06 |
Srini | how to cut off the pdf till the cropbox? | 11:23.49 |
| -DuseCropBox -FittoPage? | 11:24.08 |
| The printers mark will be outside the cropbox right? | 11:24.45 |
tor8 | Robin_Watts: I forgot to mention last night, but I put zip files of windows build of mupdf 1.2 release candidate on mupdf.com/download/archive/ | 12:40.30 |
| mind checking if everything is okay with those, before I upload to google code? | 12:40.49 |
Robin_Watts | I will look now. | 13:01.54 |
| tor8: We should probably put the v8 stuff in there too. | 13:02.36 |
| http://ghostscript.com/~robin/ <- v8* | 13:03.27 |
| looks OK to me. | 13:04.37 |
tor8 | Robin_Watts: still the same as on mupdf.com/download/v8/ ? | 13:04.52 |
| those also list 3.9 as version | 13:05.00 |
Robin_Watts | tor8: yes, nothing has changed. | 13:05.14 |
tor8 | good! I hate updating software these days... | 13:05.28 |
| apple and android have really started annoying me with their incessant upgrade nagging, and the upgrades only make things worse if they change things at all... | 13:06.02 |
kens | Srini -dUseCropBox will render a PDF using hte CropBox instead of the MediaBox | 13:06.06 |
Robin_Watts | tor8: I haven't updated android for ages. | 13:07.36 |
tor8 | I've just started living with the "google play" icon in the status bar being permanently on | 13:09.21 |
| Robin_Watts: okay, -v8.zip version uploaded as well | 13:10.31 |
| but running that with calc.pdf gives me weird results | 13:11.22 |
Robin_Watts | tor8: oh, you mean your device constantly wanting to upgrade apps? | 13:12.59 |
| calc.pdf: that's probably not good. Wierd in what way? | 13:13.20 |
tor8 | oh, nvm. I can't read today. I'm dropping the PDF on mujstest instead of mupdf :D | 13:13.42 |
Robin_Watts | phew! | 13:14.05 |
tor8 | Robin_Watts: what do you think about putting both the regular and the -v8 .exe in the same zip? | 13:16.11 |
Robin_Watts | I'd have no objection. | 13:19.25 |
Srini | kens is it physically reducing the mediabox to cropbox? | 13:20.22 |
kens | Srini, no its using the cropbox | 13:20.39 |
| Anythig outrside teh cropbox is removed, and the image size will be the area of teh cropbox | 13:20.58 |
Srini | how can we physically resize the pdf to crobox ? | 13:21.06 |
kens | I don't know what you mean | 13:21.18 |
Srini | kens, while we try to extract the pdf , it includes the printers marks outside the cropbox ... I am trying to avoid that | 13:21.48 |
| by cutting off everything outside the crobox | 13:22.03 |
kens | Right, so render using -dUseCropBox (assuming the marks are outside the crop box) | 13:22.07 |
Robin_Watts | chrisl: I think something is broken with the profile build on windows. | 13:22.40 |
Srini | kens, :) Thanks... I am not good at pdf.. so I had ask the same question again and again... sorry to bother people here! | 13:23.00 |
kens | NO Srini | 13:23.07 |
| NP | 13:23.12 |
chrisl | Robin_Watts: I've never done a profile build on Windows..... | 13:23.12 |
Robin_Watts | chrisl: It seems that somewhere in our makefile nightmare there is another profile: target defined. | 13:23.39 |
| other than at the top level in msvc.mak | 13:23.51 |
chrisl | Okay, do you need this fixed urgently? I already have the build in pieces just now..... | 13:24.38 |
Robin_Watts | chrisl: I am looking at it now. Just thought I should mention it in case you said "ah, that'll be..." | 13:25.02 |
chrisl | Nope, nothing springs to mind, sorry. | 13:25.44 |
Robin_Watts | I can't find another profile: target. I've renamed the target to profilewin: and it seems to be working. Maybe it's a VS2008 oddness. | 13:28.05 |
Srini | kens, is this the correct usage? http://pastie.org/private/izphalalejccveb6cajw | 13:31.50 |
kens | Srini looks like it, but you should check the documentation | 13:32.12 |
| It is case-sensitive | 13:32.21 |
Srini | oh! I will double check.... | 13:32.40 |
kens | However, I do not thisnk that will have anby effect with pdfwrite | 13:32.45 |
| I was referring to rendering | 13:32.52 |
| pdfwrite specifically preserves the Boxes from the input PDF file | 13:33.06 |
henrys | Robin_Watts: are you just turning off optimization and using sleepy or does the windows c compiler now have performance options? | 14:02.12 |
kens | Visual Studio has had profiling for some time | 14:03.12 |
| I usually can't get it to work | 14:03.26 |
paulgardiner | Robin_Watts: a fix for the annoation-creation crash is on paulg/master | 14:03.37 |
henrys | kens:in the free version? | 14:04.05 |
Robin_Watts | henrys: The Profile target that I set up ages ago in vs, does a specific profile build. | 14:04.13 |
| It turns DEBUG off, and hence has optimisations on. | 14:04.25 |
kens | henrys , no in the professional version I think | 14:04.25 |
Robin_Watts | but it turns DEBUGSYM on. | 14:04.34 |
| so we still get meaningful symbols out. | 14:04.44 |
| VS2008 Team Edition has profiling in. | 14:05.00 |
| Very Sleepy has become unreliable for me, hence I now use VS2008 Team Edition to profile. | 14:05.21 |
| kens: Next time you want to do profiling, talk to me, and we'll see if we can get it sorted. | 14:05.56 |
kens | Robin_Watts : I usually get there,eventually, its not something I do often enough to remember teh magic runes | 14:06.21 |
henrys | I didn't realize you got the team edition | 14:06.22 |
kens | Artifex has it | 14:06.44 |
Robin_Watts | I use VS2005 most of the time, just use 2008 for the profiler. | 14:06.53 |
kens | Which is why we are all using it (at least I think so) | 14:06.58 |
Robin_Watts | The profile shows that michaels fix has removed the shadings as being a hotspot, so that bug is closed now. | 14:07.58 |
henrys | kens:right I think ray distributed it, I didn't know if people picked it up. The free stuff seems functional to me. Is there something other than profiling I'm missing out on? | 14:11.37 |
kens | No that's pretty much it I think | 14:12.08 |
Robin_Watts | henrys: The only reason to use VS2008 Team instead of an express edition is profiling. | 14:12.26 |
Srini | kens I corrected the syntax referring to the document... it is working as expected... thanks for all the help | 14:12.53 |
Robin_Watts | (or if you want to use Visual Basic, or .net or ...) | 14:12.59 |
kens | Srini OK thanks for letting me know | 14:13.07 |
henrys | kens:oh no, they are using ps2txt now, what on earth are they up to? | 14:21.33 |
sebras | tor8: Robin_Watts: you saw the messages about Google opting for a separate font for Hangul instead of relying on DroidSansFallbackFull..? | 14:27.09 |
kens | henrys I wish I knew. THey will get better results from txtwrite though | 14:47.09 |
| henrys I think they are doing 'format conversion' | 14:47.27 |
Robin_Watts | sebras: I did not see that. Where were the messages? | 14:55.24 |
kens | In IRC logs Robin_Watts | 14:55.54 |
Robin_Watts | henrys: It used to be that 'standard' editions of VS didn't do optimisation. | 14:56.15 |
| At least, that was the case for Visual C 6. | 14:56.30 |
| but that may have gone by the wayside by the time VS2005+ rolled around. | 14:56.55 |
| 8.64? Crumbs. Upgrade already. | 14:57.50 |
kens | If only people would.... | 15:06.56 |
Robin_Watts | sebras: OK, so there is a newer font, but it's missing a particular subset of glyphs because google are using another font... I can't see that other font in the git though... | 15:09.59 |
| kens: "8.64 is too old. We don't support it. Upgrade already." | 15:10.26 |
kens | Look at the customer, Marcos will do them a patch | 15:10.45 |
Robin_Watts | ... and thus they won't upgrade. | 15:11.02 |
kens | Until we say 'we can't' yes | 15:11.19 |
| THough I think this might be such a case | 15:11.26 |
| I suspect we are using a different JPEG2000 decoder | 15:11.38 |
| So if the new one works and the old one doesn't, a 'patch' would be humungous | 15:12.01 |
chrisl | JBIG2..... | 15:12.09 |
kens | Hmm, well I guess that is feasible. I thought they said JPX I must have misread | 15:12.30 |
chrisl | "Dark bakground when transforming JBIG2 image" | 15:13.09 |
kens | Yeah, I misremembered | 15:13.17 |
| Lol looks like a stochastic screeen | 15:13.40 |
chrisl | is leaving it to Marcos..... | 15:14.04 |
sebras | Robin_Watts: oh..? I never checked whether the new one has made it to git yet... | 15:17.07 |
| Robin_Watts: I guess the commit message is a heads up about us not wanting to merge the next version of droid without thinking first though... | 15:19.05 |
Robin_Watts | The other one is supposed to be in "external/naver-fonts", but I can't find that | 15:19.06 |
| https://android.googlesource.com/platform/external/naver-fonts/ Aha | 15:19.35 |
| It sounds like we'll be getting to the point of having to provide 2 fallback fonts, which I don't like. | 15:20.09 |
| And what's the license on the second font ? | 15:20.23 |
| BSD_LIKE, I think. | 15:21.06 |
sebras | yes. | 15:21.17 |
| there is a derivative font based on droid that we might want to use instead... | 15:21.49 |
| Robin_Watts: http://en.wikipedia.org/wiki/WenQuanYi#Micro_Hei | 15:22.52 |
| I don't really understand why google decided to use a separate font..? does it all come down to chinese characters requring a too big font when all you want to write is korean..? | 15:23.59 |
Robin_Watts | It seems an odd decision to me. | 15:26.31 |
| but only Raph knows the answer I guess. | 15:27.03 |
| sebras: If they had split the font into several fonts, one per unicode 'block', so people could import the sets they wanted, that I could have understood. | 15:43.16 |
tor8 | Robin_Watts: mupdf.com/news2 | 15:59.15 |
| Robin_Watts: sebras: naver-fonts would most likely be the nanum font | 16:00.10 |
Robin_Watts | Prototype transition support should be in new features? | 16:00.22 |
tor8 | Robin_Watts: yes, you're right. | 16:00.37 |
Robin_Watts | but otherwise, looks great. | 16:00.44 |
tor8 | with so many bullet points, I figured we should categorise them a bit :) | 16:01.03 |
Robin_Watts | It mentions the AGPL change. henrys ? | 16:01.06 |
alexcher | chrisl: Dark background is reproduced in v. 8.64. Newer versions have white background. The customer's TIFF was produced by v. 8.64. | 16:01.19 |
chrisl | alexcher: so, echoing Robin_Watts's comment: they should upgrade..... | 16:01.49 |
tor8 | Robin_Watts: more than one CJK fallback font is easy to add to mupdf, it doesn't need to change anything in the API | 16:02.46 |
henrys | Robin_Watts: it's fine for release notes just not something I want to send in the newsletter to customers. | 16:02.48 |
tor8 | but doing that (ie updating to the stripped droidsansfallbackfull and adding in nanumgothic) will add 500k to the data size | 16:03.58 |
| sebras: probably because the nanum font is a better looking font to korean sensibilities. or because it looks like Roboto... | 16:04.36 |
chrisl | Erm, so "droidsansfallbackfull" is no longer "full"? | 16:05.03 |
tor8 | chrisl: indeed... | 16:05.09 |
| they took out the korean script from it | 16:05.23 |
chrisl | So, surely it shouldn't be called "full" any more...... | 16:05.45 |
tor8 | chrisl: couldn't agree more | 16:06.06 |
Robin_Watts | DroidSansFallbackAsFullAsWeNeed | 16:06.15 |
chrisl | Bad google! | 16:06.22 |
tor8 | or we should use some font editor and mash together droidsans, droidsansfallbackfull, and nanumgothic into one big fat fallback font which really is FULL! | 16:06.50 |
Robin_Watts | DroidSansFallBackCrammed | 16:09.13 |
Hooks | whats goin on fellas | 16:10.32 |
| wondering if anyone can give me a hand installing ghostscript on a server2012 terminal server, or if it is even possible, one of my software vendors appears to require this to be installed for an invoicing aspect | 16:11.23 |
kens | It shoudl be possible, what is the problem ? | 16:11.38 |
Hooks | during install when i set the printer name and select next it come right back to the same screen almost instantly | 16:12.03 |
| from what i can gather its a driver signing issue, i have removed the requirement for signed drivers through group policy and update policy / restarted | 16:12.34 |
kens | Hmm, don't know about that, do you really need a virtual printer ? | 16:12.40 |
chrisl | Ghostscript doesn't install a printer | 16:12.53 |
kens | Yeah I thougth that | 16:13.02 |
Hooks | Ghostscript is the driver for a "new print to file" printer being added | 16:13.17 |
kens | Err, what ? | 16:13.31 |
chrisl | Ghostscript isn't a printer driver | 16:13.36 |
kens | Ghostscript is not a pritner driver | 16:13.37 |
chrisl | Hooks: are you trying to use ghostpdf.inf or something? | 16:14.15 |
Hooks | yes i am | 16:14.25 |
chrisl | Right, that's not Ghostscript, as such | 16:14.40 |
Hooks | Postscript and PDF interpreter/renderer Ghostscript 9.07 | 16:15.22 |
tor8 | Robin_Watts: do you have a release APK for mupdf 1.2 somewhere? | 16:15.26 |
Hooks | not the same thing? | 16:15.26 |
kens | Not at all | 16:15.35 |
chrisl | Ghostscript is an application, not a virtual printer, nor a printer driver | 16:15.57 |
kens | Windows has a PostScript printer driver, you can create PostScript files by sending them to disk. | 16:16.12 |
| Ghostscript consumes PostScript files and produces 'aomething else'; | 16:16.24 |
| something* | 16:16.35 |
| THe 'something else' depends how you configure Ghostscript | 16:17.22 |
chrisl | Huh, I was just trying to find something that might help Hooks, but no point now...... | 16:17.41 |
kens | I guess not, no | 16:17.51 |
Robin_Watts | tor8: Do I have an apk built from vanilla 1.2? No. | 16:18.33 |
tor8 | Robin_Watts: or should we juts point people at Google Play? | 16:19.02 |
Robin_Watts | i think that's best. | 16:19.08 |
tor8 | let's do that then | 16:19.14 |
| Robin_Watts: the "technology preview release" section on the front page, that's a bit out of date by now :) | 16:23.11 |
Robin_Watts | tor8: lose it! | 16:24.22 |
| or update it to say "the latest release" instead. | 16:24.49 |
| tor8: A tiny review on robin/master for you. | 16:27.01 |
tor8 | Robin_Watts: mupdf.com front page, is that wording okay? | 16:28.46 |
Robin_Watts | perfect. | 16:29.15 |
| Is it Affero GPL version 3? | 16:29.39 |
| It is. perfect. | 16:30.01 |
tor8 | I also removed the section about third party and added the "--recursive" flag to "git clone" | 16:30.11 |
Robin_Watts | sales@artifex.com could be a mailto: link ? | 16:30.59 |
tor8 | Robin_Watts: good catch. the IRC link, is there a web irc client we could link to? | 16:32.06 |
Robin_Watts | tor8: yeah, hold on... | 16:32.18 |
| http://webchat.freenode.net/?channels=ghostscript&uio=d4 | 16:32.37 |
| or maybe just http://webchat.freenode.net/?channels=ghostscript | 16:32.59 |
robin_test | seems to work. | 16:33.33 |
tor8 | fab! | 16:34.10 |
Robin_Watts | Artifex Software Inc in the bottom could be a link ? | 16:34.35 |
| and you could link to the FSF etc too? | 16:34.55 |
tor8 | too many links, need to keep it to the relevant parts :) | 16:35.09 |
Robin_Watts | fair enough | 16:35.23 |
tor8 | a link to the AGPL might be good though | 16:35.27 |
sebras | tor8: if you want the crammed droid, isn't that this one? http://en.wikipedia.org/wiki/WenQuanYi#Micro_Hei | 16:53.46 |
ray_laptop | morning, all | 16:55.40 |
Robin_Watts | sebras: Is it? It's not clear to me from that description how the merge has been done. | 16:56.17 |
tor8 | sebras: there are some licensing issues with that one (not to mention the difficulty in actually locating the font files) | 16:56.22 |
| broken effing website. javascript popup "Please browse the font page first." when I followed the link from said page... | 16:57.08 |
Robin_Watts | Is it 'droidsans with any missing glyphs overridden by merging from the other font' or is it 'droidsans with any glyphs defined in the other font replacing the droidsans one' | 16:57.18 |
sebras | tor8: heh... yes the files are really hard to find. even on their sf.net site... that's unexpected. | 17:00.16 |
kens | goodnight all | 17:03.27 |
tor8 | chrisl: right, so mupdf 1.2 is released. the files are in mupdf.com/download/ if you want to copy them to the ghostscript.com download page | 17:06.49 |
chrisl | tor8: Thanks, I'll probably do it tomorrow morning | 17:07.34 |
sebras | tor8: does that mean a new apk as well? | 17:07.50 |
tor8 | sebras: nope. now we point to google play. | 17:08.02 |
sebras | tor8: yes, that's what I meant. | 17:08.18 |
Robin_Watts | sebras: I wasn't planning to put a new apk up on play until we got the reflow stuff agreed. | 17:08.59 |
sebras | Robin_Watts: ok. | 17:09.11 |
chrisl | tor8: so, you switched to zip? Ick :-( | 17:14.01 |
malc_ | Robin_Watts: false alarm | 17:14.18 |
Robin_Watts | with the transform thing? | 17:14.30 |
| And what was the question you asked a few days ago that I didn't answer? | 17:15.42 |
malc_ | Robin_Watts: a. yes | 17:16.18 |
sebras | tor8: you write "javascript" on the front page and "Javascript" on the news page. I believe the latter is more common... | 17:16.51 |
malc_ | Robin_Watts: b. rendering tile which is 4096 pixels high takes aproximately the same time as the one that is mere 512 | 17:16.52 |
Robin_Watts | malc_: Right. It depends largely on the content of the tile :) | 17:17.15 |
malc_ | Robin_Watts: well, in my experience it makes zooro difference | 17:17.41 |
Robin_Watts | If the content has an interpolated image on, say, then that might be the slowest thing to render. | 17:17.54 |
| If the tile at 256 high only has the same number of pixels full of such an image as the one at 4096 does, then the time for the extra 4096-256 pixels is probably negligible. | 17:18.46 |
malc_ | Robin_Watts: let's say the pdf just contains an embedded jpeg | 17:18.55 |
Robin_Watts | ok. | 17:20.20 |
| And how big is that embedded jpeg when it's rendered? | 17:20.52 |
| i.e. are we changing the zoom level between rendering at 4096 high, or 256 high ? | 17:21.21 |
malc_ | Robin_Watts: nope | 17:21.49 |
Robin_Watts | Ok, so under our thought experiment, what size does our embedded jpeg render to? | 17:22.18 |
| 256x256 ? 4096x4096 ? something else ? | 17:22.48 |
malc_ | Robin_Watts: sorry i don't quite get the question | 17:23.26 |
Robin_Watts | We have a PDF which has a page which has an embedded JPEG on. | 17:24.00 |
malc_ | Robin_Watts: right | 17:24.08 |
Robin_Watts | You have asserted that when you render a 'tile' from that page at a given zoom, it takes the same time to render a tile that is 256 pixels high as a tile that is 4096 pixels high. | 17:24.58 |
| I am asking: what height is the rendered image of the jpeg? | 17:25.22 |
ray_laptop | well, my fixes for bug 693115 now show a clear progression on sumatra/1485_-_image_not_displayed.pdf, trivial differences in shading on 586_-_missing_images_gs_SMask_not_applied.pdf | 17:25.40 |
| but... several of the pdfwrite tests show regressions :-( | 17:26.03 |
Robin_Watts | if we were rendering an entire bitmap of the page, how high would the section of the page bitmap that came from the bitmap be? | 17:26.05 |
malc_ | Robin_Watts: understood, let me collect the information | 17:26.06 |
Robin_Watts | malc_: I don't need an exact figure, just an approximate one so I can get it in my head. | 17:26.24 |
malc_ | Robin_Watts: the only non exact information i remember is that the embedded jpeg is BIG | 17:27.14 |
Robin_Watts | IF the embedded jpeg was (say) 16x16 in the top corner, then clearly the time taken to render that onto a 256x256 tile will be more or less the same as the time taken to render it onto a 4096x4096 tile, as clearly the rate determining step is the painting of those 16x16 pixels. | 17:27.39 |
| Ah, right. | 17:27.41 |
| So it's a big jpeg. | 17:27.50 |
| If it's a BIG jpeg, then you may be seeing issues with the store. | 17:28.06 |
malc_ | Robin_Watts: actually scratch that, it's a PDF without jpegs.. it's the pango-view output | 17:28.24 |
| which i can share | 17:28.33 |
sebras | Robin_Watts: is that the encoded source image, or is it scaled up/down when rendered on the page..? | 17:28.41 |
Robin_Watts | malc_: Is that a page of bitmaps? | 17:29.03 |
| Or 1 big bitmap? or something else ? | 17:29.16 |
malc_ | Robin_Watts: nope, just some text | 17:29.18 |
| single font | 17:29.26 |
Robin_Watts | How are you drawing each tile? | 17:29.49 |
| Do you run the page to a displaylist, then run the displaylist for each tile ? | 17:30.06 |
malc_ | Robin_Watts: yep | 17:30.21 |
Robin_Watts | OK. I have no idea why the times are the same. | 17:30.39 |
malc_ | Robin_Watts: not the same, 4kx4k takes 0.06 seconds 512x512 0.057 (for each tile) | 17:31.10 |
| or thereabouts | 17:31.15 |
| comparable, contrary to expectations that is | 17:31.25 |
Robin_Watts | right, but I have no idea why they are comparable. | 17:31.30 |
| Can you run profiles? | 17:31.38 |
malc_ | Robin_Watts: boblycat.org/~malc/main.pdf | 17:32.19 |
| Robin_Watts: how? | 17:32.20 |
Robin_Watts | malc_: If it's a platform specific thing, then it's quite possible I won't be able to reproduce it here. Better to actually look at where time is being spent on your machine. | 17:32.59 |
| malc_: Using a profiler :) | 17:33.06 |
| You're working on linux, right? | 17:33.12 |
malc_ | Robin_Watts: aye | 17:33.47 |
Robin_Watts | Does your kernel have oprofile support built in? | 17:33.58 |
malc_ | Robin_Watts: oprofile is so last century, real idiots user perf these days | 17:34.20 |
Robin_Watts | malc_: OK, so... using whatever tool you prefer... run a job using 256x256 tiles. Then run again using 4096x4096 tiles. | 17:35.00 |
| Then we can compare the outputs. | 17:35.08 |
| I'd strongly suspect that we'll see one thing dominating the timings. | 17:35.50 |
| Either that, or the page must be MUCH bigger than 4096x4096 and our time is being dominated by the time taken to run through the displaylist. | 17:37.04 |
malc_ | Robin_Watts: i posted the pdf, you can confirm or reject this theory yourself.. (the pdf is ~75K.. single page (although a tall one)) | 17:38.10 |
Robin_Watts | malc_: So, you want me to download the file, write a test harness for it, try to reproduce the problem locally, then profile, analyse the profiles, and fix the problem. | 17:40.46 |
malc_ | Robin_Watts: not really, what you can do is see the dimensions yourself in whatever space you want | 17:41.34 |
| Robin_Watts: and sorry if i offended you in anyway, if i did it wasn't intended | 17:42.04 |
Robin_Watts | malc_: Until I see profiles, anything else is guesswork. | 17:42.14 |
| And I don't want to head off down a blind alley. | 17:42.31 |
ray_laptop | Robin_Watts: isn't that the way we work at Artifex -- guess what the problem is, then go and implement changes based on our guesswork ;-) | 17:43.02 |
Robin_Watts | profiles, or a simple test app I can run here to see the same problem. | 17:43.02 |
| ray_laptop: I like to pretend I have more than guesswork to go on generally, but shhh! there might be commercial customers listening. | 17:44.05 |
malc_ | Robin_Watts: now you are asking me to write a simple app that should run on windows.. grumble grumble | 17:44.08 |
ray_laptop | if I had a dollar for every example of "premature optimization" I've seen in Ghostscript, I could buy Artifex :-) Of course, mupdf didn't start off quite as badly | 17:44.14 |
Robin_Watts | malc_: Doesn't have to run on windows. (But it would be great if it did) | 17:44.39 |
| ray_laptop: Oh yes. I'm sure we've all been guilty of it. | 17:44.53 |
malc_ | Robin_Watts: lack of enthusiasm for *nix/windows is what unites us.. too bad we are in the different camps though | 17:45.35 |
Robin_Watts | malc_: No, I like linux a lot. I just don't like debugging on linux :) | 17:45.59 |
ray_laptop | even the design of the fitz display list had a lot of "we think this will be better", but that was mostly based on knowledge of how bad the clist was | 17:46.23 |
sebras | malc_: would this be of any help to you? http://goo.gl/vFzKW | 17:47.06 |
Robin_Watts | tor8: 2 small reviews on robin/master for you. | 17:49.11 |
malc_ | sebras: good question that should have been addressed to Robin though :) | 17:50.24 |
| Robin_Watts: will callgrind satisfy you? | 17:50.33 |
| or cachegrind for that matter | 17:50.44 |
Robin_Watts | Anything that gives me results that I can read here, sure. | 17:50.55 |
malc_ | maybe i should do this tomorow at work.. after all machine there has 8x cores, and 30x memory | 17:52.29 |
Robin_Watts | http://bugs.ghostscript.com/show_bug.cgi?id=693618 <- Anyone know what Photon is in this instance? Is this commercial usage ? | 17:52.33 |
malc_ | Robin_Watts: photon is the name of QNXes windows system if my memory serves me | 17:53.16 |
| -s | 17:53.22 |
sebras | Robin_Watts: according to wikipedia "Photon microGUI, a windowing system of the QNX operating system" | 17:54.09 |
Robin_Watts | So I wonder if that needs to be forwarded to Scott then ? | 17:57.30 |
ray_laptop | Is sumatra/1439_-_color_softmask_fails_to_draw_jpx_image.pdf an Adobe bug? Ghostscript looks right, and Adobe doesn't show the picture in the backgroun | 18:30.17 |
Robin_Watts | softmasks are supposed to be greyscale, not color. | 18:36.16 |
| right? | 18:36.23 |
| ISTR some code in mupdf that turns color softmasks to greyscale ones. Maybe because of that file in fact. | 18:37.13 |
ray_laptop | Robin_Watts: are you talking to me? But, yes, softmask in some modes are mapped to grayscale (Luminance) | 18:41.11 |
| Robin_Watts: and a SoftMask can also be developed that is only the alpha channel from a group | 18:41.55 |
| i.e., throw away all the color/gray info | 18:42.14 |
Robin_Watts | There is a special case for jpx images with softmasks too. The softmask can be taken from alpha values in the JPX data too. | 18:44.54 |
ray_laptop | 586_-_missing_images_gs_SMask_not_applied.pdf through pdfwrite is broken both before and after my "fixes" with different things missing :-( | 18:47.37 |
| Robin_Watts: correct | 18:47.50 |
| Robin_Watts: so does this relate to my question about sumatra/1439... ? | 18:48.18 |
Robin_Watts | ray_laptop: It was intended to. Not sure my comments were helpful in any way though :( | 18:48.41 |
ray_laptop | Robin_Watts: Have you looked at that file (1439) with Acrobat? Does it look right to you ? | 18:49.52 |
Robin_Watts | oh. this is hivemind.pdf | 18:50.40 |
ray_laptop | Robin_Watts: right. Ghostscript looks mostly the same as Foxit, and with my changes is even closer in w.r.t. the colors | 18:51.21 |
Robin_Watts | right. gs looks good. | 18:53.09 |
| mupdf complains with "error: softmask must be grayscale" | 18:53.24 |
ray_laptop | Robin_Watts: sorry -- was on the phone. So mupdf probably is not doing the Luminance conversion | 19:15.59 |
| That's specified in 7.4.2 of the spec | 19:17.09 |
Robin_Watts | I think we do do that. | 19:18.42 |
| There is something "odd" about this file. | 19:18.50 |
ray_laptop | Robin_Watts: BTW, AFAICT VOX.pdf and 586_-_missing_images_gs_SMask_not_applied.pdf are the same | 19:21.40 |
| of course, that's just visually | 19:21.55 |
| Well, I went through the set of differences from my fixes, and some are really minor, and some are clear progressions. 586... and VOX have problems before and after, so I'll make sure there's a bug on it. | 19:26.19 |
| this should get bug 693115 off our list. | 19:26.36 |
| Robin_Watts: 1x1 images with Interpolate true ??? What bozo app generated this PDF (Cairo ?) | 19:34.18 |
| early lunch today... | 19:38.42 |
Robin_Watts | ray_laptop: yeah. But it lead to a larger bug being found. | 19:39.40 |
| Forward 1 day (to 2013/02/28)>>> | |