| <<<Back 1 day (to 2013/04/10) | 2013/04/11 |
soo | Can anybody please tell me if mupdf for android handles clickable links? | 05:35.10 |
| If it does, what do I need to enable it? | 05:36.14 |
| Sorry for disturbing everybody with a question before, which was how to enable links in a pdf in android app. I was able to figure this out some moments ago. Its as simple as setting mLinksEnabled=true in MuPDFReaderView.java. On doing this should do the trick. But I got a NullPointerException when doing quick swipes to navigate across pdf pages. For this thr quick fix is to put a try catch in LinkInfo hitLink(float x, float y). re | 06:42.38 |
| Special thanks to Robin_Watts in giving timely help during his busy schedule :-) | 06:44.14 |
kens | chrisl ping | 08:13.20 |
chrisl | kens: pong | 08:30.03 |
kens | chrisl I was looking into high level colour because of henrys valgrind problems, was that what prompted you to look aty gx_hld_saved_color_equal ? | 08:31.23 |
chrisl | kens: yes, it was | 08:31.50 |
kens | ah, would you like me to take it over ? | 08:32.15 |
chrisl | I was just about to push the fix - just tweaking the comment | 08:32.47 |
kens | Or just apply your patch ? | 08:32.49 |
| OK that will get rid of a lot of the log issues I think | 08:33.01 |
chrisl | Okay, pushed it. | 08:33.36 |
kens | I don't think that the 'problems' are really anything to do with 'high level colour' at all, because the interpreter shouldn't have to do anything, patterns are a different thing though. | 08:33.57 |
chrisl | I didn't look at the PCL ones - I've been tracking down a couple of other issues | 08:35.29 |
kens | I suspect your fix will solve most of the valgrind problems | 08:35.32 |
chrisl | Due to the way Marcos assigned them, I suspect there will be quite a lot that going on | 08:36.16 |
kens | Adding high level patterns is going to need me to understand patterns in PCL more I think.... | 08:36.20 |
| yes, I think a lot of thse will turn out to be the same problem manifesting in different ways | 08:36.37 |
chrisl | Understanding PCL is like understanding quantum physics - if it doesn't give you a headache, you haven't understood it! | 08:37.29 |
kens | Well I can't even see where the PCL interpreter does patterns at the moment. TH PXL interpreter I can find it.... | 08:37.58 |
chrisl | They're probably not called "patterns" in PCL...... | 08:38.31 |
kens | I think I'm going to put a note in the bug that henrys has open saying that I think your fix will resolve these valgrind problems. | 08:38.42 |
chrisl | I think, at some point, we're just going to have to get Marcos to re-run the tests, and start over. Too much stuff moves around as we fix/change things. | 08:40.00 |
kens | Yes, and I think a lot of the problems will be gone with this fix, my change for hashing and a few that Alex and Robin (possibly others) have put in. | 08:40.33 |
| wow, thre's a lot of code ot handle PCL patterns :-( | 08:57.44 |
| I thuink I may just giv eup on PCL patterns and stick with just PXL | 09:04.26 |
ceroman | hello | 09:53.37 |
ghostbot | hey, ceroman | 09:53.37 |
ceroman | may I ask a question? | 09:53.53 |
ghostbot | may i ask a question is probably Don't ask to ask, just ask. | 09:53.53 |
kens | Hmm ghostbot getting dicatorial | 09:54.11 |
ceroman | nice bot | 09:54.20 |
| well I want to convert a pdf to tiff, and preserve the color profiles of the pdf, I tried http://pastebin.ubuntu.com/5697952/ , identify -verbose tells me " Artifex CMYK SWOP Profile" though | 09:57.27 |
| gs is 9.05 | 09:58.04 |
kens | Our colour expert is on vacation this week, but I think basically you can't do that. I'm not sure it even makes sense | 09:58.52 |
ceroman | pdf is a pdf1.4 made with scribus, all fonts to curves, just cmyk colors in vector, and all images in the right color profile; well the printing-facility wants to have that profile - do you say that does not matter with tiff? | 10:00.47 |
kens | If the PDF has multiple profiles they ewill relate to hte input colour spaces, not the output, so you couldn't really sue them (and if they are multiple, which will you use ?) | 10:01.00 |
ceroman | nope, it is just one profile for the pictures, or do you mean the vector color and pattern colors ? | 10:02.21 |
kens | THe 'progile for the pictures' is an input profile, and describes how to go form the colour space of the image into a device-indepndent CIE space | 10:02.57 |
| There isn't any good reason to preserve that | 10:03.16 |
| And it has no bearing on the CMYK colourants which GS has put in the output, the output profile descibes how the device-independent space was converted into CMYK (or RGB or whatever) | 10:03.50 |
ceroman | so.. the reason for all of this conversion I try to do is that the pdf I produced uses transparency, but the printer-facility does not support that; I also tried to flatten the thing to pdf 1.3 http://pastebin.ubuntu.com/5695249/ | 10:07.31 |
| but then the Image's cut out backgrounds (the area which should be transparent) are not transparent any more... | 10:09.07 |
| and as Scribus is unable to produce a tiff with cmyk colours, I thought I would convert it like this. | 10:11.44 |
| looks like I have to dig a lot deeper to understand the whole thing... | 10:12.39 |
kens | ceraoman if the result of rendering the transparency to tiff is incorrect, then that is a bug, what version of GS are you using ? | 10:23.56 |
| But in any event, I don't hitnk youo really want teh profile from teh input PDF file attached to the output TIFF file | 10:24.41 |
ceroman | ok, the version I have is9.05 | 10:25.59 |
kens | Well that is a little old but should be OK | 10:26.09 |
| You could try 9.07 but if you think there is a bug, please report it | 10:28.25 |
ceroman | so my settings would be ok? | 10:28.55 |
kens | Loks ok yes | 10:29.13 |
ceroman | I am on ubuntu 12.04, i386 .. will try to update to 9.07 and try again, meanwhile, here is tha screenshot of wat its spits out http://pix.toile-libre.org/?img=1365676746.png | 10:39.56 |
kens | Sorry, not really sure what that is telling me | 10:40.55 |
ceroman | well the girl was actually cut out in front and background pictures before I converted... | 10:46.26 |
kens | well helps to know what to expect :-) | 10:46.58 |
| LIke I sadi, if you think its a bug please do report it | 10:47.22 |
ceroman | shure I will.. | 10:48.13 |
tor8 | Robin_Watts: paulgardiner: a commit on tor/android for review | 11:03.23 |
kens | lunches | 11:34.22 |
paulgardiner | tor8: all looks fine. I don't know enough about the build to understand whey the first one is needed, but it does what it says on the tin. | 12:01.37 |
tor8 | paulgardiner: tor/android, not tor/master (but those needed review too) | 12:03.32 |
paulgardiner | Doh! | 12:03.47 |
| tor8: I'm not sure FileName should be mapped. It's just a key against which info is stored in a DB. | 12:07.43 |
tor8 | paulgardiner: that's why I removed it from the mappings | 12:10.35 |
| what if the user changes language, then it can't find the same key | 12:10.59 |
paulgardiner | Doh2! | 12:11.13 |
| Right. Nicely spotted. | 12:11.22 |
tor8 | :) | 12:11.24 |
paulgardiner | I must have captured that one by accident. | 12:11.38 |
Robin_Watts | tor8: 2 of the 4 on robin/master are up for review too. | 12:25.46 |
tor8 | Robin_Watts: already checked those. they look good. | 12:26.53 |
Robin_Watts | tor8: Thanks. | 12:27.00 |
| tor8: pushed your review + my 2. | 12:35.55 |
| tor8: We should think about transfer functions. | 12:38.22 |
tor8 | Robin_Watts: could you give the three on tor/master an eye as well? | 12:39.25 |
Robin_Watts | will do. | 12:39.35 |
tor8 | would it be possible without too much trickery to bake the transfer function into the colorspace? | 12:39.44 |
Robin_Watts | tor8: I can't remember :( | 12:40.17 |
| I keep looking and it keeps falling out of cache again. | 12:40.27 |
| Anything which involves color mapping changes should perhaps be held off on until we can discuss it with mvrhel - it would seem silly to change everything twice. | 12:41.37 |
| OK, those 3 look fine to me too. | 12:58.06 |
| I'll push them too. | 12:58.12 |
tor8 | Robin_Watts: thanks. | 13:10.13 |
Robin_Watts | tor8: Should I remove the fz_load_jpeg function now it's not used? | 13:22.47 |
henrys | kens:pcl uses gsptype1.c:gs_makepixmappattern | 13:24.36 |
| kens:I assume you were looking in the pcl code. | 13:24.51 |
kens | oh really ? There seemed ot be a lot of other pattern-related code | 13:24.58 |
| henrys yes, I was | 13:25.04 |
Robin_Watts | tor8: 2 of the 3 reviews on robin/master are good to go. Last one is still a work in progress. | 13:25.56 |
henrys | kens:I'll make an easy pattern file shortly. | 13:26.18 |
kens | Thanks henrys, the PXL code does seem to use the pattern things I expected to see :-) | 13:26.37 |
henrys | kens:PCL does too, just in a different place. it is all make/render pattern based. | 13:27.26 |
kens | Ah, fair enough, I couldn't see the 'PainProc' bit, which is what I need to work on. | 13:27.52 |
| I'm just looking at pstype1.c now | 13:28.09 |
tor8 | Robin_Watts: yes, no point in keeping dead code around. | 13:28.43 |
Robin_Watts | tor8: Cool. That's the second of the 2 reviews :) | 13:29.07 |
| kens: PainProc. Freudian. | 13:29.24 |
kens | :-) | 13:29.30 |
tor8 | how do you feel about putting the fz_image functions in res_image.c? | 13:29.39 |
Robin_Watts | tor8: That would probably make sense. | 13:29.58 |
tor8 | instead of having it cozy up to fz_pixmap (even though they are related) | 13:30.01 |
Robin_Watts | Even allowing for the lack of transfer functions, we're handling something in that bug wrong :( | 13:31.55 |
henrys | kens:note that PCL is restricted to mono mask and color bitmaps pattern-wise. | 13:32.25 |
kens | Well that's a simple subset of regular patterns | 13:32.43 |
| a mask is just an imagemask in a pattern, and an image is just an image in a pattern | 13:33.07 |
| So it ought to be easy enough to recode. | 13:33.20 |
henrys | kens:right - I just didn't want you searching around for other types of tests. | 13:33.28 |
kens | Ah :-) | 13:33.34 |
| I've 'half done' the PXL case, but I need a real test case for it so I can see what I need to do for the remainder, I think I will have to reasd the pixmap pattern stuff some more. Does anything other than PCL use this code ? | 13:34.38 |
| aha, and I just found the bits I need in the pixmap patterns too. | 13:35.22 |
Robin_Watts | tor8: Updated patches online with res_image.c | 13:39.13 |
| lunch. | 13:41.19 |
henrys | kens:no just PCL | 13:41.48 |
| kens:oh darn I made the PCL test file first. | 13:42.40 |
kens | :-) | 13:43.02 |
| No matter, I'd like both | 13:43.08 |
henrys | kens:okay both sent | 13:56.38 |
kens | thanks henrsy | 13:56.46 |
| henrys that is | 13:56.58 |
| got them, | 13:57.04 |
henrys | chrisl:thank for fixing the indeterminism I misdiagnosed. | 14:15.00 |
chrisl | henrys: np. As I said to kens, I suspect there are quite number that overlap like that | 14:15.44 |
Robin_Watts | PDF question... | 14:21.59 |
| Imagine that I've got an alpha softmask here. | 14:22.56 |
| The content stream draws half the mask in white, and half in black. | 14:23.19 |
| Is the mask half transparent, or fully opaque ? | 14:23.36 |
kens | I'd have thought half is 100% transparent, half is 100% opaque | 14:23.54 |
| Have you tried this ? | 14:24.26 |
Robin_Watts | kens: I'd have thought the same. I don't think that's what I'm seeing in acrobat though, and reading page 552 of the PDF 1.7 spec makes me wonder. | 14:24.36 |
kens | Reading the spec ? You must be desperate | 14:24.50 |
Robin_Watts | â¢If the subtype is Alpha, the transparency group XObject G is evaluated to compute a group alpha only. The colors of the constituent objects are ignored and the color compositing computations are not performed. The transfer function TR is then applied to the computed group alpha to produce the mask values. Outside the bounding box of the transparency group, the mask value is the result of... | 14:25.01 |
| ...applying the transfer function to the input value 0.0. | 14:25.02 |
| OK. Colors are indeed ignored. Both black and white regions count as being set. | 14:32.26 |
| And interestingly, mupdf gets it right. I didn't expect that :) | 14:32.41 |
| And so does gs. | 14:33.16 |
cinch | kens, i have two odd PDFs where it segfaults with -dNOGC on 9.07 (but not on 9.06, might be a regression) | 15:07.32 |
kens | yes sounds like it, seg faults are ok I can find those ;-) | 15:08.04 |
cinch | on one of those PDFs: on 9.06 -> segfault | 15:14.41 |
| 9.06 with NOGC -> floating point exception | 15:14.55 |
| :) | 15:14.59 |
kens | Have you opened a bug report ? | 15:19.48 |
cinch | na, was thinking of doing it | 15:19.57 |
kens | Please do, it means I won't forget it, and will have the relevant files | 15:20.12 |
cinch | ok^ | 15:20.25 |
kens | It is conceivable that the recent burst of activity fixing Valgrind reports will have solved it of course | 15:21.07 |
chrisl | Don't forget, NOGC doesn't completely disable garbage collection | 15:21.51 |
Robin_Watts | --debug=gc-disable does. | 15:22.56 |
chrisl | Robin_Watts: but only on a debug build? | 15:23.20 |
Robin_Watts | No, always, I think. | 15:23.28 |
chrisl | Oh, right. Still, BAD IDEA! | 15:23.46 |
Robin_Watts | chrisl: Eh? It's a great idea for debugging problems. | 15:24.22 |
| It's a bad idea for an end user :) | 15:24.30 |
chrisl | Robin_Watts: that's what I meant - only for use when you know what you're doing | 15:24.51 |
Robin_Watts | right, yes. | 15:24.56 |
ray_laptop | darn, my peeves git has master off to the side and about 10 commits back, What's the git command to just abandon the change that was in master and move it up to the origin/master ? | 16:04.41 |
Robin_Watts | ray_laptop: There are several possibilities. | 16:05.16 |
| Firstly, if you: "git pull --rebase origin master", it should pull in the new master from origin, then attempt to rebase your changes onto it. | 16:06.06 |
| Lots of them will hopefully resolve away to nothing, and you'll be left with just the outstanding ones on the end. | 16:06.33 |
| IF you get conflicts along the way you can either resolve them (edit the files to remove the conflicts then git add them, then git rebase --continue) | 16:07.03 |
| or you can choose to just skip that commit completely (git rebase --skip) | 16:07.18 |
| or you can decide that the whole rebase idea was a bad one and backtrack to where you were (git rebase --abort) | 16:07.42 |
| another way to work would be to drop a marker where you are (git branch marker), then to fetch the changes from the remote (git fetch origin), then to forcibly change your current master to match that one (git reset HASH). | 16:09.30 |
| where HASH = the hash of the place you want it to be. I generally get that from looking at git logg or looking on the git web view. | 16:09.51 |
| ray_laptop: Making sense so far? | 16:10.13 |
ray_laptop | sorry. phone call | 16:10.52 |
Robin_Watts | no worries. I'll wait til you finish. | 16:11.02 |
ray_laptop | Robin_Watts: OK, so what I wanted was git rebase --skip | 16:35.24 |
Robin_Watts | ah, ok. | 16:35.41 |
| If you have a modern enough git installed, then it tells you what your options are at each stage. | 16:36.05 |
| And by modern enough I mean "not stone aged" :) | 16:36.19 |
ray_laptop | thanks. I don't often have to do that, and usually I have had to manually edit the failed merge, so I hadn't used --skip | 16:36.31 |
Robin_Watts | So, I clusterpushed a mupdf job, and got a message back telling me that it had got SEGVs while running gs!?!? | 16:37.11 |
chrisl | Robin_Watts: that might be related to my previous push - I left some debug code in by mistake, hence why I aborted it | 16:41.31 |
Robin_Watts | ray_laptop: Ping | 17:57.17 |
ray_laptop | Robin_Watts: sorry -- I wasn't paying attention earliere | 19:11.18 |
| wow! the cmd_write_unknown is REALLY stoopid. It does cmd_write_ctm_return_length(cldev, &ctm_only(&cldev->imager_state)); which does all of the checking to see how much will be written, THEN it later calls cmd_write_ctm to actually copy the data :-( | 19:43.51 |
| where a smarter approach would be to just write the CTM to a buffer, then memcpy it to the command list (however much there was) | 19:45.09 |
| Forward 1 day (to 2013/04/12)>>> | |