| <<<Back 1 day (to 2012/02/21) | 2012/02/22 |
Robin_Watts | sebras? | 00:14.04 |
| fz_bitmap_s is purely for the output of halftoning. | 00:14.29 |
| Currently that's just 1bpp. | 00:14.44 |
| In theory at some point I could do 1bpc, for cmyk or something. | 00:15.17 |
tor8 | Robin_Watts: I think he's referring to the comment, "Bitmaps have 1 component per bit" | 00:49.47 |
| which is a bit ambiguous | 00:49.59 |
| and backwards to how things are normally stated :) | 00:50.10 |
Robin_Watts | Indeed. | 00:50.15 |
| This lock reworking is nightmarish. | 00:52.08 |
| s/lock/cache/ | 00:52.17 |
tor8 | hashtable trouble? | 00:52.33 |
Robin_Watts | A combination of things. | 00:58.26 |
| Trying to extract fz_obj knowledge from the hashtable. | 00:58.42 |
| Trying to make the hashtable threadsafe. | 00:59.03 |
| I'm currently worried that the fz_obj reference counting isn't protected by a lock. | 00:59.31 |
| which means that if someone goes to update a reference count in the interpreter thread while something else is mallocing in another thread, that might cause an eviction and make a race on the update of the reference. | 01:00.56 |
tor8 | oh. nightmarish. | 01:01.33 |
Robin_Watts | I could wrap the fz_obj refcounting in the alloc lock. | 01:02.21 |
| but then I need to be able to adjust the counts while the lock is held. | 01:02.44 |
| So I might need locked and unlocked variants. | 01:03.03 |
| (of the keep/drop) | 01:03.17 |
| Actually, I think, I can get away with only needing the one variant, so wrapping the ref counts in locking should be OK. | 01:07.41 |
| But I'm going to bed now. Tomorrow I will be doing trip prep, so it might end up being a problem for the plane. | 01:08.20 |
tor8 | good night | 01:09.01 |
Robin_Watts | night | 01:09.40 |
halabund | I have a vector figure made of a large number of shaded polygons (~10,000 polygons, and each vertex might be of a different colour). It takes very long to render in most PDF viewers. Is this normal, or would it be possible to have the same information with a different structure so it's fast to render? | 09:03.43 |
| Does PDF make it possible to render these fast at least in principle? Is it possible that the program I exported the figure from just generates a PDF structure that is inefficient? | 09:04.33 |
chrisl | halabund: It sounds to me like the representation is pretty inefficient. You *might* be able to improve things by using Type 2 shading patterns...... | 09:11.08 |
halabund | chrisl: Are there any programs which can optimize this type of PDF for me? If not, is it easy enough to try to create my own PDF from scratch using the figure data? A long time ago I have experimented a little bit with postscript, so I know that PostScript can be written "by hand" easily enough. But I have no idea how PDF works. | 09:14.17 |
kens | Type 2 shading patterns are a PostScript feature as well as PDF. | 09:14.42 |
| I doubt there are any applications which can analyse an existing mulit-polygon description and produce a sahding pattern from it though | 09:15.17 |
| You owuld have to hand-code it in either PostScript or PDF | 09:15.32 |
halabund | This is a sample figure I have: http://dl.dropbox.com/u/38623/test.pdf.gz | 09:16.36 |
| (It compressed 10-fold with gzip, so I suppose there must be something redundant in the representation) | 09:16.55 |
kens | Can't open it, Firefox tries to open it as a PDF.... | 09:17.33 |
halabund | Oh ... wait | 09:17.53 |
| kens: I re-named it to .gzip but it's the same file: http://dl.dropbox.com/u/38623/test.pdf.gzip | 09:18.46 |
chrisl | halabund: Oh, that's not going to be so easy to make more efficient..... | 09:19.00 |
| halabund: just how scalable do you need this to be? | 09:19.16 |
kens | Not going to be easy to do that as a sahding. | 09:20.15 |
| Would need *many* shadings. | 09:20.23 |
halabund | chrisl: If you think anything else is too much trouble, I am willing to rasterize it. But high resolution rasterized versions take up a lot of space. The main reason I didn't rasterize it so far was that I needed the axes and labels (which are not shown here) to be sharp. | 09:20.24 |
kens | You could render the polygons to image, and incorporate that, while retaining the axes and text as vectors | 09:20.55 |
halabund | Alright, the next plan is: figure out how to rasterize the smooth shaded part and keep the axes vector | 09:21.02 |
| Thanks for the advice! | 09:21.15 |
kens | Well, you couyld use GS to make an image, then curt out the smooth area | 09:21.26 |
chrisl | halabund: I reckon that's your best bet for a speedy rendering | 09:21.27 |
halabund | Interestingly Sumatra does render it very fast, but occasionally some "cracks" appear between polygons at some zoom levels. But that's the only viewer that rendered it fast. (Now I'm trying to coerce Mathematica to separate axes and polygong, rasterize one, then correctly overlay the results) | 09:36.30 |
kens | Sumatra uses MuPDF, you could try that too :-) | 09:37.10 |
| But presumably you want to share the PDF, so it doesn't really help | 09:37.26 |
halabund | Exactly | 09:37.50 |
kens | Can't you just render the image, then place it in Mathematica as a figure ? | 09:37.52 |
| THat is, take your existing PDF, render it to TIFF using Ghostscript, then use imagemagick or something to crop out the axes, and resave. Load the TIFF into Mathematica ? | 09:38.50 |
| Well, the customer isn't going to be happy abou thteir PDF file, there's nothing I can do to affect the way Acrobat is printing it. | 10:07.42 |
| At least, not easily, and not without slowing down everything else | 10:08.22 |
chrisl | kens: do you know what's causing the problem? Another Acrobat bug? | 10:11.01 |
kens | Nope, its just the way the file is made | 10:13.22 |
| The file is drawn by setting a big black rectangle. | 10:13.35 |
| Then a clipping path. | 10:13.41 |
| That path is then filled with a Pattern colour space | 10:13.51 |
halabund | Alright, managed to hack together a version with the axes in vector and the smooth density plot as raster :-) | 10:13.58 |
kens | The pattern does only one thing, it draws an image | 10:14.01 |
| The image is pure white. | 10:14.13 |
| At some resolutions/scaling, the image disappears. | 10:14.27 |
| THe pattern comes from PCL as does the rest of the drawing. | 10:14.47 |
| Looking for a pattern which only uses a single image (and nothing else) and tehn checking to see if the image is a pure black or white is possible, btu expensive to do | 10:15.24 |
| Maybe henrys could spot the pattern higher up and replace with a flat fill, I don't know. | 10:15.57 |
chrisl | Presumably, the result is just fine at 300 and/or 600dpi? | 10:16.08 |
kens | And 1590 | 10:16.14 |
halabund | Do you know of any free programs that are able to optimize the PDF size and work well, other than Multivalent? | 10:16.15 |
kens | 150 | 10:16.17 |
| halabund you could try running it through GS using the pdfwrite device, but it depends what's using up the space. | 10:16.47 |
| chrisl, but at 144 dpi it doesn't work (pure black) | 10:17.00 |
chrisl | So in other words, at resolutions valid for PCL it works - it's just like the old SciText Postscript - "garbage" in, "garbage" out....... | 10:18.11 |
kens | Yes, pretty much. | 10:18.52 |
chrisl | kens: (and Robin_Watts) BTW, checkin is open, but you might want to leave it a little while - the site seems to be having a bad hair day...... | 10:19.07 |
kens | I'm about to go out, I'll do it later, but thanks | 10:19.23 |
chrisl | That's one reason I mentioned it - If you wanted to get it done before you went out, you might be frustrated! | 10:20.22 |
kens | No time after debugging that problem, but there's no great rush | 10:20.40 |
| I guess that's probably it work-wise for the week, I'll need to finish loading up the laptop and pack after lunch | 10:21.09 |
| Early start tomorrow :-( | 10:21.17 |
chrisl | I'm checked in, but I can get a boarding pass - "Sorry, we've encountered a problem, please try later....." :-( | 10:21.37 |
kens | Oh great.... | 10:21.44 |
chrisl | What time do you have to leave, then? | 10:22.21 |
kens | I'll have to leave the house around 5:50 am | 10:22.34 |
| Coach from Gatwick at 6;20 | 10:22.48 |
| Get to Heathrow around 7:50 | 10:22.56 |
| (assuming the M25 is OK) | 10:23.03 |
| Ho ho | 10:23.09 |
| Anyway, time to go, be back later | 10:23.55 |
chrisl | have fun! | 10:24.01 |
Robin_Watts | Well, I have a boarding pass. | 10:26.57 |
chrisl | I'm now waiting on it being e-mailed - fifth attempt's a charm, apparently | 10:27.33 |
sebras | Robin_Watts: have a nice trip and say hello to tor8 when you meet him. | 10:40.25 |
| Robin_Watts: if he's tired it's because I kept him awake reviewing documentation until 3am last night. ;) | 10:40.59 |
Robin_Watts | sebras: hehe. | 10:41.13 |
| Morning paulgardiner | 10:41.44 |
paulgardiner | morning | 10:41.58 |
| Received a second email from Curry's. so looks like the order is going through ok. | 10:43.57 |
chris2 | just found this possibly corrupt pdf file which crashes gv, mupdf, zathura/poppler-glib yet xpdf and mupdf can open it and render it correctly: http://leaf.dragonflybsd.org/mailarchive/kernel/2011-11/pdf00001.pdf | 11:09.14 |
| not mupdf, evince can open it, sorry | 11:09.28 |
chrisl | chris2: no crash here, just a load of warnings about the pdf being broken | 11:11.54 |
| evince and xpdf also dump a load of warnings about it being broken. | 11:13.06 |
chris2 | /usr/bin/mupdf is owned by mupdf 0.9-1 | 11:13.11 |
| http://sprunge.us/HcdF | 11:13.38 |
chrisl | You said "crashes gv", I tested on ghostscript | 11:14.01 |
chris2 | ah | 11:14.09 |
| http://sprunge.us/DPSC | 11:14.36 |
| /usr/bin/gv is owned by gv 3.7.3-1 | 11:14.44 |
Robin_Watts | chris2: We are familiar with mupdf here :) | 11:15.16 |
chris2 | the mupdf page told that mupdf people are here too | 11:15.31 |
| chrisl: ok, funny | 11:15.34 |
| chrisl: pdftops works, and evince can render the resulting ps | 11:15.44 |
| but gv crashes | 11:15.47 |
chrisl | gv version doesn't help much, it's just a "skin" over ghostscript. gv works for me, too | 11:16.01 |
| gv 3.7.2 | 11:16.20 |
chris2 | mupdf-git works | 11:16.28 |
| /usr/bin/gs is owned by ghostscript 9.05-1 | 11:16.46 |
| yes, plain gs works too | 11:17.39 |
chrisl | Well, gv isn't ours, so we'll need a Ghostscript command line in order to drive gs in the same way that gv does in order to reproduce the problem | 11:18.11 |
| chris2: what distribution are you running? | 11:18.38 |
chris2 | haha | 11:18.38 |
| gv crashes just on starting | 11:18.46 |
| i'm sorry | 11:18.49 |
| Arch | 11:18.52 |
chrisl | Ugh, can't help - report it to the gv maintainer. They should be able to determine if it's a gv or gs problem, and report it to the relevant upstream project | 11:19.44 |
chris2 | i'll try recompiling and then report it | 11:19.58 |
| but its nice to see what crap libreoffice can generate | 11:20.10 |
| https://bugs.archlinux.org/task/28516?project=1&string=gv ah | 11:20.30 |
chrisl | ;-) Personally, I'd also report the broken PDF to LibreOffice...... | 11:21.15 |
chris2 | yes | 11:21.44 |
chrisl | chris2: BTW, (just a pet hate of mine) what you've got there is not a "crash" in MuPDF, it's a graceful exit on encountering broken input...... | 11:22.34 |
chris2 | true | 11:22.59 |
chrisl | It's just a crash is *very* important to resolve ASAP, whilst a graceful exit is rather less so (assuming the problem can be worked around at all) | 11:24.12 |
chris2 | yeah | 11:25.13 |
chrisl | bbiaw..... | 11:25.14 |
paulgardiner | tor8: I saw your post about moving the background. That's good. I have no idea why I had it on the docview (although maybe I added the background before there was an outer layout to include the buttons). Background clearly makes more sense on the outermost view. | 11:27.47 |
tor8 | paulgardiner: if I'm not imagining things, I think that sped up the scrolling speed too | 11:28.14 |
paulgardiner | tor8: Did changing from fade to slide-on help at all with the animation problems? | 11:28.16 |
tor8 | the fade vs slide didn't affect the flickering (moving the background did that) but it sure looks nicer and animates much more smoothly | 11:28.48 |
paulgardiner | Oh good. It was the smoothness I was most wondering about. | 11:29.18 |
| Are there still some signs that the scrolling may benefit from a different drawing algorithm? | 11:31.13 |
tor8 | not sure without an fps meter :) | 11:32.04 |
| it looks like some of the browsers have the same choppiness | 11:32.23 |
| of course, if we could get it to 60fps using opengl es that'd be even sweeter :) | 11:32.35 |
| sebras noticed a few bits in the search ui that could use some more polish | 11:33.08 |
| like clearing the page search hit highlights when you start editing the search field | 11:33.29 |
paulgardiner | Oh ok. That should be easy enough. | 11:33.58 |
tor8 | and indicating that it's searching somewhere. I cheated there in the ios version by just disabling the search field while it's searching. | 11:34.00 |
| something else that's bothered me is the links on the page, you can't really see where they are | 11:34.18 |
| perhaps we could highlight those regions in a different color if the buttons layer is visible, do you think that would be a good design? | 11:34.49 |
| so that if you want to follow the links, you tap to show the buttons and see the hyper links in red highlights or something like that | 11:35.10 |
| or would it be better to have them show permanently? | 11:35.28 |
paulgardiner | Not sure. If always highlighted, it becomes impossible to see the doc in it's intended form. | 11:36.31 |
tor8 | yeah. too bad we can't detect finger hovering :) | 11:36.56 |
paulgardiner | :-) | 11:37.03 |
| Maybe links should be visible only when buttons layer is up, and also active only when the buttons layer is up. | 11:38.02 |
tor8 | paulgardiner: I tweaked (i.e. massacred) the icons a bit in the branch on casper. the icons you'd picked looked dreadful on android 3.2 with its default dark gray button backgrounds | 11:38.51 |
| yeah, I think that's probably the best solution for links | 11:39.12 |
Robin_Watts | Maybe have another button in the buttons layer that toggles links on/off in the normal view. | 11:39.53 |
paulgardiner | I didn't like the buttons either, and hoped someone would pick some better ones. | 11:40.05 |
tor8 | yeah, we've got space for another button (or two) in the normal view | 11:40.27 |
Robin_Watts | We could have a button with 3 states. 1) Links aren't shown or followed, 2) Links are followed, but not shown, 3) Links are shown and followed. | 11:41.21 |
tor8 | Robin_Watts: I don't see the point of (2) to be honest | 11:41.43 |
| that's just tapping blindly :) | 11:41.50 |
Robin_Watts | Well, lots of documents have hyperlinks shown in a different color anyway. | 11:42.09 |
| (i.e. html pages that have been turned to PDFs). | 11:42.23 |
| Having an extra layer of highlighting from the client would be bad in those cases. | 11:42.35 |
paulgardiner | Yes. Hadn't considered that. | 11:42.57 |
Robin_Watts | 1) could be a greyed out 'chain' icon, 2) would be a not-greyed out chain, and 3) would be a not-greyed out chain with a highlight on it. | 11:43.15 |
tor8 | or crossed over chain icon, normal chain icon, and highlighted chain icon | 11:43.54 |
| though three state buttons are problematic by themselves, not sure how cumbersome android will make it | 11:44.25 |
Robin_Watts | yes, that may be better as greyed out might look like 'button is disabled'. | 11:44.29 |
tor8 | paulgardiner: http://somerandomdude.com/work/iconic/ this is the icon set (CC-BY-SA) that I used | 11:44.59 |
paulgardiner | Nice. Is the use of those committed to the branch? | 11:46.04 |
tor8 | paulgardiner: yup. | 11:46.13 |
| and a new app icon too :) | 11:46.19 |
| they look a bit bad on my phone, since the white silhouette doesn't show half as well on android 3.2's white-colored buttons. *sigh* | 11:47.12 |
paulgardiner | I shall update and have a look. | 11:47.16 |
tor8 | they really do make it difficult to build good looking apps | 11:47.25 |
| I did play around with making a custom button graphic layout using some xml shape definitions | 11:48.15 |
| Robin_Watts: did you get your new device yet? | 11:49.15 |
Robin_Watts | tor8: It's being delivered to Paul tomorrow afternoon. | 11:50.02 |
tor8 | fab | 11:50.09 |
Robin_Watts | I'll be on the plane of course. | 11:50.19 |
paulgardiner | Henry has given me the go ahead to do more work on the app, but I want to be sure that the draw algorithm is still an issue before looking into changing that. Could be a lot of work just to end up with no improvemnt. | 11:52.17 |
tor8 | paulgardiner: how about just doing a benchmark sample app that scrolls one bitmap over a tiled background and see if you can replicate the slowness and if opengl speeds it up? | 11:53.14 |
| it may be that android 4 has solved the problem altogether | 11:53.40 |
| and the current state of the app is okay, even though it's not quite as smooth as iOS | 11:54.09 |
paulgardiner | One easy thing I can check is if removing the background still speeds it up. Also I can remove the low quality ImageView see how fast it goes with just the HQ patch | 11:55.44 |
| The problem with the opengl approach is that it might require the a complere rewrite and with none of the existing architecture helping. | 11:56.37 |
tor8 | yeah, you'll have to reimplement the whole scroller won't you? | 11:57.14 |
paulgardiner | Yes, plus I don't there isn't a framework to plug into (no existing classes to subclass). | 11:58.05 |
| Haven't thought it through enough yet. May be not as much work as I think. Or perhaps there's a trick. | 11:58.51 |
| It may be possible to just bring in OpenGL in the draw methods, and still use the existing control architecture | 11:59.23 |
| Actually, that might work. | 11:59.33 |
tor8 | worth looking into, I think | 11:59.46 |
| but Robin and henrys will have to have their say as well :) | 11:59.53 |
paulgardiner | Of course :-) | 12:00.04 |
| And as you suggested yesterday, use the z buffer to avoid overrendering. | 12:02.31 |
tor8 | overdraw shouldn't matter on powervr based chips, but I think the androids mostly use the nvidia tegra | 12:04.04 |
| but once the framework is in place and the pages loaded into textures, experimenting with optimal opengl drawing order is trivial :) | 12:05.35 |
| gotta head out for a bit, back in an hour or so | 12:08.43 |
paulgardiner | cyl | 12:09.06 |
Robin_Watts | paulgardiner, tor8: The Transformer prime is indeed Nvidia Tegra based. | 13:14.54 |
ahfeel_ | hi :) | 13:22.59 |
| I'm having issues with imagemagick and ghostscript on ubuntu 11.04 (natty) | 13:23.28 |
| identify somefile.pdf crashes (segmentation fault) | 13:23.52 |
| gs version is 9.01 | 13:24.01 |
kens | Try upgrading, current version is 9.05 | 13:24.16 |
ahfeel_ | any idea of a similar reported issue ? any trick ? | 13:24.18 |
| kens: mh okay, 9.05 package isn't available on repositories, i'll try to find an unofficial package | 13:24.50 |
kens | There are crashes reported, I have no idea if they are the same as yours without seeing the file. THe 'trick' is to upgrade and if taht doesn't work, report a bug | 13:24.51 |
ahfeel_ | allright | 13:25.00 |
| i'll keep you informed | 13:25.09 |
kens | I'm sure there is a Ubuntu package for 9.04 at least | 13:25.09 |
| Till issued a patch release. | 13:25.19 |
| 9.05 is due to be packaged in teh next release of Ubuntu (I forget the name , Oneiric ?) | 13:25.49 |
ahfeel_ | oneiric is the current last, but it's not the LTS one | 13:26.08 |
| it's 11.10, LTS is the 11.04 | 13:26.13 |
| next one has a strange name, can't remember, like sharp penguin | 13:26.28 |
kens | No idea, I assume you are right. I don't try to keep up with distro versions | 13:26.36 |
chrisl | Perfect Pangolin | 13:26.43 |
ahfeel_ | chrisl: thanks :) | 13:26.47 |
kens | I htought Pangolin was hte next LTS ? | 13:27.04 |
chrisl | 12.04 is an LTS version | 13:27.42 |
kens | gives up on versionis | 13:27.57 |
chrisl | Sorry, it's "Precise Pangolin" - I was close! | 13:28.11 |
ahfeel_ | hehe | 13:28.25 |
| yeah, and it's going a super long term supported one, 5 years | 13:28.33 |
chrisl | It looks like official Ubuntu 11.04 tops out at GS 9.01.... | 13:29.33 |
ahfeel_ | yeah | 13:29.38 |
| seems like upgrading to 11.10 solves the problem | 13:29.45 |
| pff :< | 13:29.51 |
kens | Well, you could build from source (and get the proper 3rd party libraries that way too) | 13:29.58 |
ahfeel_ | How the hell can they deliver a broken version and not backport on the lts | 13:30.40 |
| anyway, I think I have no other choice than upgrading | 13:31.05 |
| :) | 13:31.05 |
chrisl | Could be that no one has reported the problem | 13:31.09 |
kens | chrisl, looks like someone is bitching about FreeType's rendering, the latest 'PLP' one. | 13:40.10 |
| He's complaining that 'the bottom row of pixels' is missing | 13:40.26 |
chrisl | kens: yes, looking at it - *really* stupidly big and complex file, and Acrobat won't reset the bbox! | 13:40.56 |
kens | Right, I was looking at the output, a single pixel line font..... | 13:41.11 |
| Oh yes, massive drawing like all their files. | 13:41.52 |
chrisl | Yeh, with weird construction, layers and all kinds of crap in there :-( | 13:42.20 |
kens | 48x30 inches and he's complaining about one row of pixels.... | 13:42.27 |
| Yes, all their files are weird | 13:42.38 |
| Easy to delete the main content though | 13:43.19 |
chrisl | I should get a debuggable file today, but I doubt I'll do any "real" work on it until next week | 13:43.31 |
kens | Acrobat crops it OK for me | 13:44.10 |
| Though I can't actually find the offending data | 13:44.36 |
chrisl | Hmm, Acrobat told me "cannot reduce page size"..... | 13:44.36 |
| It's very small, at the far right of the document | 13:44.56 |
kens | Document->Crop Pages, then play with Margin controls ? | 13:45.00 |
chrisl | I tried that, and strange things happened, then I just tried to reduce it to A4, and I got the error message | 13:45.48 |
kens | Oh, is it in the seal ? | 13:45.52 |
| I see it is | 13:46.03 |
| Want a reduced file ? | 13:48.25 |
Robin_Watts | chrisl, kens, tor8: Weather in SF is forecast to be 14 degrees or so during the day, but only 5ish at night; I bet Alcatraz may feel colder. You might want to pack something warmer for that. | 13:48.42 |
chrisl | kens: I've got it down to just the seal | 13:49.05 |
kens | Thanks Robin I had intended to check, I'll wear a jacket | 13:49.13 |
tor8 | Robin_Watts: so basically the same clothing as here :) | 13:49.16 |
kens | chrisl, me too | 13:49.21 |
| I can delete bits of it too | 13:49.36 |
chrisl | Yeh, I'm being circumspect in case there is a clip in there | 13:50.05 |
kens | I have just the text now. | 13:50.27 |
| There's something funny about hte Acrobat disply | 13:50.43 |
| Bottom of the text is 'faint' | 13:50.51 |
| Well, its OK at 1200% zoom | 13:51.14 |
| ahfeel I can't see the message you are sending me | 13:52.06 |
ahfeel_ | ah | 13:52.16 |
| I was saying | 13:52.22 |
| So the problem seems that my gs version segfaults on a given pdf file, I tried another one and it works great | 13:52.26 |
| can I give you the file so you can try and validate that it's fixed ? | 13:52.30 |
kens | Umm.... | 13:52.37 |
| I don't see how I can do that. | 13:52.50 |
ahfeel_ | the only thing I run is identify file.pdf (identify calls gs underneath) | 13:53.04 |
kens | Without fully debugging the file with an old version of software, which I'm not keen on | 13:53.06 |
| I'd need a full GS command line | 13:53.18 |
| I can't use Imagemagick | 13:53.25 |
ahfeel_ | arf | 13:53.35 |
| you'd just have to run "identify file.pdf" | 13:53.45 |
kens | On Windows ? | 13:53.56 |
ahfeel_ | linux | 13:53.59 |
kens | :-) | 13:53.59 |
ahfeel_ | :/ | 13:54.01 |
| anyone with gs >= 9.04 running linux ? :) | 13:54.19 |
kens | Almopst everyone :-) | 13:54.28 |
| But not me | 13:54.33 |
chrisl | Yes, if you send the file to me, I'll run it for you | 13:54.40 |
ahfeel_ | chrisl: ahh great thanks | 13:54.49 |
| chrisl: http://technofeel.com/~ahfeel/28eac32d-4bab-4eb1-8475-ab782ecfbd91.pdf | 13:54.52 |
| it segv here, $ identify ~archivezen/28eac32d-4bab-4eb1-8475-ab782ecfbd91.pdf | 13:55.16 |
| Segmentation fault | 13:55.19 |
| identify: Ãchec délégation Postscript `/home/archivezen/28eac32d-4bab-4eb1-8475-ab782ecfbd91.pdf': @ error/pdf.c/ReadPDFImage/645. | 13:55.21 |
| It works great with gs 8.71 | 13:55.41 |
chrisl | Same here - seg fault | 13:55.49 |
ahfeel_ | oh | 13:55.54 |
| chrisl: gs 9.05 ? | 13:55.59 |
chrisl | hang on a sec | 13:56.17 |
| *our* 9.05 seems to work fine | 13:56.59 |
kens | Aha. | 13:57.12 |
ahfeel_ | so which version segfaulted ? | 13:57.15 |
chrisl | 9.04 | 13:57.21 |
ahfeel_ | lolz. | 13:57.26 |
| :((( | 13:57.33 |
kens | Could build from source | 13:57.43 |
ahfeel_ | means that even by upgrading my whole system to 11.10 and having 9.04 it will still segv | 13:57.49 |
chrisl | *but* I'm using 9.05 with the third party libraries that *we* test and ship with, whilst 9.04 is using whatever random library versions got installed with it on Ubuntu | 13:58.21 |
ahfeel_ | chrisl: do you have any hint on what's invalid in this pdf file ? if there's something invalid, maybe it's just these version of gs that have a bug | 13:58.23 |
| ah | 13:58.34 |
chrisl | hah! *Looks* like a broken or problematic ICC profile | 13:59.14 |
ahfeel_ | maybe it's because it's a PDF/A ? | 14:00.04 |
chrisl | I know we fixed quite a few problems (including crashes) in LittleCMS (even with valid profiles) which weren't taken upstream, and weren't used by Ubuntu | 14:01.54 |
ahfeel_ | chrisl: it segfaulted with a vanilla 9.04 on ubuntu 11.10 ? | 14:11.11 |
kens | Ubuntu versions of GS are not quite the same as 'vanilla' | 14:11.37 |
| We keep our own versions of 3rd party librareis, which we know work | 14:11.51 |
ahfeel_ | lol, what I mean is with the version delivered by ubuntu :) | 14:11.56 |
chrisl | ahfeel_: no, Ubuntu's installed version crashed | 14:12.14 |
kens | Ubuntu insists on using shared libraries, which may not work. Hence Chris's comments about LCMS version 1 | 14:12.16 |
ahfeel_ | chrisl: allright | 14:12.33 |
| well, thanks for testing ! | 14:12.41 |
chrisl | Basically, with our patched lcms, 9.04 renders the file successfully, with the Ubuntu "system" lcms (without our patches) it seg faults - the problem is in lcms | 14:13.00 |
kens | OK, checked in. Now I need to go and print all these pieces of documentation. | 14:26.49 |
| Mail sent to customer and henrys about the PCL problem. | 14:53.04 |
henrys | kens:I'll look at the pattern thing thanks. | 15:30.08 |
kens | Thanks henrys | 15:30.17 |
henrys | alexcher:do you have enough to work on for 692851? | 15:56.36 |
| is it clear what needs to be done given the 2 messages from Luratech? | 15:57.35 |
patrisk | I'm just going to go on a limb here. Can I use ghostscript to convert a bunch of .mht-files to pdf? | 16:01.12 |
| If so, can anyone point me in a direction where I can read more about it? | 16:01.21 |
alexcher | henrys: yes, I've just fixed another bug and switching to 692851. | 16:01.57 |
Robin_Watts | patrisk: No, ghostscript does not read .mht files. | 16:02.02 |
patrisk | Ah, well thanks for the quick answer. | 16:02.14 |
Robin_Watts | Your best bet is probably to load them into a browser (or something that does support them) and then print to pdf. | 16:02.34 |
henrys | alexcher:great | 16:02.39 |
patrisk | Yeah I was worried I have to do that. I have 100+ so it will take time, but I'll probably end up doing that. Thanks anyway. | 16:03.06 |
mvrhel | bbiab | 16:09.58 |
henrys | going through the bugs before the meeting - I've sent now 4 sales leads to scott, we are missing them I'll put that on the agenda. | 16:11.56 |
kens | Hmm, I hadn't realised we were missing so many | 16:13.13 |
henrys | they aren't great leads but they are possible support contract opportunities. | 16:14.59 |
kens | Fair enough, worth trying anyway | 16:15.10 |
henrys | Robin_Watts:something to ponder on the plane how difficult would it be to change the path code to not remove unneeded moveto's and rmoveto's. 692809 is the bug. | 16:24.43 |
Robin_Watts | oh, right, so HPGL is like Draw. | 16:25.45 |
| It's probably dead easy. | 16:25.55 |
henrys | What's Draw? | 16:26.10 |
Robin_Watts | The question is, how to expose the different drawing mode. | 16:26.13 |
| Draw is Acorns vector format. | 16:26.18 |
henrys | that's easy just gs_set* and gs_current* operators. | 16:27.21 |
| that sets a boolean in the gstate | 16:27.59 |
Robin_Watts | I think I'd rather see it per path. | 16:28.08 |
| Maybe not. | 16:28.18 |
| I think it's doable. | 16:28.26 |
henrys | okay I'll put it on the agenda if you want or we can just leave it for now, let me know. | 16:30.15 |
| bbiab | 16:30.33 |
Robin_Watts | Stick it on the agenda, just so it's not forgotten. | 16:31.06 |
| tor8: ping | 16:31.08 |
tor8 | Robin_Watts: yes? | 16:37.47 |
Robin_Watts | I've just pushed an updated images branch to my repo. | 16:38.11 |
| I've done all sorts of nasties with the store/hash table etc. | 16:38.22 |
| If you have time to cast your eye over it, that'd be great. | 16:38.41 |
| On the positive side, I believe I've broken the stores dependence on fz_obj in as nice a way as possible. | 16:39.12 |
| sebras: So, I've pulled the doc branch onto my laptop. | 17:06.43 |
sebras | Robin_watts: ok. I still have a series of spelling fixes and other grammar that tor8 mentioned last night. | 17:08.17 |
| I'll try to fix most tonight, but I haven't got that far yet. | 17:08.33 |
Robin_Watts | If tor8 is OK with the changes in the images branch, I can merge that to master, then merge doc to that too, and then update it with the changes I've made. | 17:09.02 |
| OK, so I'll repull the doc branch before I fly tomorrow. | 17:09.14 |
sebras | yup, that's probably easiest. | 17:09.39 |
ahfeel_ | chrisl, kens: Just to keep you informed, I installed 9.05 using the tarball, everything went very smooth and it now works :) | 17:25.42 |
chrisl | ahfeel_: great! Also, so you're up to date: 9.05 on the coming Ubuntu release uses lcms2, so the problem should be resolved there. | 17:28.09 |
kens | Glad to hear it is working | 17:28.29 |
| Off to finish packing, see you all Thursday or Friday. | 17:29.30 |
ahfeel_ | chrisl: lcms2 is bundled with the tarball distribution ? | 17:31.35 |
| chrisl: I guess is it, because I havn't installed anything :) | 17:31.58 |
marcosw | alexcher: are you around? | 17:32.31 |
chrisl | ahfeel_: our tarball comes with everything you need to build GS - there's just a few optional features which depend on external libs. | 17:32.58 |
ahfeel_ | chrisl: okay great :) thanks anyway guys ! | 17:33.41 |
chrisl | ahfeel_: np, thanks for letting us know you've got it working | 17:34.23 |
Robin_Watts | Gawd. The openjpeg documentation fairly sucks. | 17:37.45 |
| It looks to me like I have to feed it a whole buffer full of data - I can't feed it small blocks at a time. | 17:38.56 |
| Anyone here had any experience of it? | 17:40.48 |
| OpenJPEG 1.5 is out. | 17:43.21 |
chrisl | Ooh, alexcher should probably check that - I have a nasty feeling it still won't contain everything GS needs, though..... | 17:44.22 |
Robin_Watts | 'many bugfixes' apparently. | 17:44.35 |
chrisl | Yeh, I think it's missing features we need - probably not due until 2.x | 17:45.55 |
mvrhel_laptop | Robin_Watts: when do you depart? | 17:52.25 |
Robin_Watts | mvrhel_laptop: Leave the house tomorrow at 6am. | 17:52.45 |
mvrhel_laptop | you chrisl and kens all on the same flight | 17:54.30 |
Robin_Watts | yeah. | 17:54.35 |
mvrhel_laptop | good deal. I get in around 10pm or so Thursday | 17:55.04 |
Robin_Watts | I suspect we'll all be tucked up in bed by then :) | 17:55.34 |
mvrhel_laptop | yes | 17:55.40 |
W8uiE5 | gs -dQUIET -dNOPAUSE -sPAPERSIZE=a4 -sDEVICE=tiffpack -dBATCH -sOutputFile=a.tiff -r300 a.ps creates a tiff with a kintelmark. how can i create one with a kmotorolamark? | 18:00.45 |
chrisl | W8uiE5: add -dBigEndian=true might work | 18:06.04 |
W8uiE5 | chrisl: works. (-: how did you investigated this? | 18:09.12 |
chrisl | W8uiE5: I looked at the source - it happens I've been working in the tiff devices recently, and I remembered seeing something about setting endianness | 18:11.01 |
W8uiE5 | chrisl: ok. thanks | 18:11.43 |
Robin_Watts | tor8: ping | 18:12.09 |
chrisl | W8uiE5: some, but not (apparently!) all of the tiff options are documented in Devices.htm which ships with GS | 18:12.54 |
alexcher | Robin_Watts: currently, the whole compressed image is buffered in memory. JP2 file consists of tables that require direct access. | 18:13.39 |
henrys | mvrhel_laptop:we really should separate them on these flights, if one went down the company would be okay ;-) | 18:13.40 |
mvrhel_laptop | yes | 18:13.47 |
W8uiE5 | chrisl: locate Devices.htm shows no match | 18:13.58 |
Robin_Watts | alexcher: Ah, ok, thanks. | 18:14.23 |
chrisl | W8uiE5: it's available here (latest version): http://www.ghostscript.com/doc/current/Devices.htm | 18:14.25 |
mvrhel_laptop | hmm the linux open printing summit is at apple in Cupertino this year | 18:17.31 |
W8uiE5 | thanks | 18:18.29 |
mvrhel_laptop | something is screwy with xps and linearity in our opacity masks | 18:19.21 |
Robin_Watts | mvrhel_laptop: Tempt the geeks in with shiny things, and don't let them leave... | 18:19.27 |
| mvrhel_laptop: Are you looking at the bug I raised? | 18:19.37 |
| bug 692818 ? | 18:20.17 |
mvrhel_laptop | Robin_Watts: no, I fixed the crash and I now seem an issue here. I do recall you pointing out this issue early. My fix may have made it more prevalent. Trying to track down the issue now | 18:20.23 |
| likely something to do with linear vs. nonlinear gray representation | 18:20.42 |
Robin_Watts | 692818 gives some cases where it's really obvious. Don't know if that will help. | 18:20.49 |
mvrhel_laptop | ok cool | 18:20.55 |
| I will add those into my testing | 18:21.01 |
| oh this is a premultiplied and not premultiplied issu. | 18:21.35 |
| yet another potential problem | 18:21.46 |
| the case I am looking at, we are creating an opacity mask from a gradient | 18:21.59 |
| the icc profiles get set to ensure things stay linear but somewhere things go afoul | 18:22.25 |
| anyway. I need to head out for a bit. bbiab | 18:22.44 |
Robin_Watts | tor8 has a note in muxps that says something like "alpha = x is supposed to mean non-premultiplied, but all files seen in the wild lie" | 18:22.45 |
mvrhel_laptop | interesting | 18:23.04 |
| ok | 18:23.19 |
| I will look into that at the same time | 18:23.25 |
| thanks | 18:23.28 |
tor8 | Robin_Watts: yeah. I remember writing the comment, but I don't remember all the details that led me to the conclusion | 18:24.26 |
Robin_Watts | tor8: Ah. | 18:24.44 |
| tor8: Did you see my burblings about the images branch ? | 18:24.54 |
tor8 | Robin_Watts: yes, but I haven't had the time to look through your latest changes yet | 18:25.15 |
| was hoping to do that on the airport tomorrow | 18:25.21 |
Robin_Watts | Because air sickness isn't nauseating enough? :) | 18:25.40 |
diverdude | Hello. is there some tool which allows me to edit eps figures? | 18:29.53 |
Robin_Watts | emacs? | 18:30.51 |
| or a vector editor, like Xara, maybe. | 18:31.36 |
diverdude | i mean some graphical editor | 18:31.46 |
Robin_Watts | Xara, or inkscape or something ? | 18:32.02 |
| inkscape will apparently import eps if ghostscript is installed. | 18:34.38 |
| but it may want to save svgs out :( | 18:34.47 |
diverdude | xara-gtk? | 18:36.01 |
sebras | diverdude: mmm, that's the name of the package in Ubuntu/Debian. | 18:37.36 |
Robin_Watts | henrys: Is the agenda finalised enough to send yet? I'd like to print it out before I shut down for the trip. | 20:54.17 |
henrys | yeah I'll have it out in about 10 minutes | 20:54.42 |
Robin_Watts | ah cool. | 20:54.59 |
sebras | Robin_Watts: hm... is it ok if you pull sebras/doc when you land? | 20:56.29 |
| Robin_Watts: or did you intend to read through it on the plane? | 20:56.51 |
henrys | Robin_Watts:sent I'm pretty sure that wasn't one of my best researched agendas, much shorter than usual. | 21:11.16 |
Robin_Watts | sebras: I was planning to pull it at 5am :) | 21:16.50 |
| or maybe even at the airport. | 21:17.06 |
sebras | Robin_watts: good. then I have ample time... :) | 21:18.15 |
Robin_Watts | henrys: If we do a bug sweep at the meeting, we are bound to spot anything important you've missed. (Though I can't think of anything) | 21:20.54 |
henrys | I've often considered just dumping most of the agenda items in bugzilla and having the meeting "bugzilla centered". | 21:22.24 |
| specifically all agenda projects could be entered as P3 then each meeting we would review all bugs and enhancements >= P3, we'd probably want to lower the priority of many of the enhancements before trying that. | 21:28.44 |
mvrhel | ok now back to screening.... | 21:43.57 |
henrys | out for a run or maybe more, 60 to 70 mph gusts predicted for today. | 21:52.17 |
mvrhel | wow | 21:52.26 |
henrys | seems to have calmed down a bit so I'm going to head out now. Lots of flights cancelled today. | 21:53.18 |
mvrhel | are you coming in thursday evening? | 21:53.38 |
henrys | yes, well maybe | 21:54.19 |
mvrhel | ok. have a good run | 21:54.31 |
| Forward 1 day (to 2012/02/23)>>> | |