| <<<Back 1 day (to 2011/10/26) | 2011/10/27 |
ray_laptop | OK, so all of my diffs were due to the changes in the band height :-( | 02:07.32 |
| I've backed out my band height & BufferSpace changes (temporarily) because they cause too many differences, some of which look serious. One last cluster run before pushing the other changes that fix bug 692618 | 02:32.21 |
AlecTaylor | hi | 03:57.34 |
ghostbot | privet | 03:57.34 |
AlecTaylor | How was the MuPDF .sln generated? | 03:58.30 |
mvrhel2 | night all | 05:51.39 |
kens | OK that's the *basic* 'XML' format txtwrite output compatible with MuPDF. Now I need to try and create a 'block' output. | 09:44.22 |
tor8 | grrr, f*ng buggy p.o.s. framework animation idiocy!!! | 09:59.50 |
kens | ? | 10:00.17 |
tor8 | ios... | 10:00.21 |
kens | Ah.... | 10:00.29 |
tor8 | more specifically, when rotating the device, it sets up an animation and I need to tweak the layout of the page views I have put inside the scrollview and adjust the scroll view so the current page stays centered | 10:01.41 |
kens | :-( | 10:02.16 |
tor8 | but ... in some cases once I've reached the final edge of the scroll view once, it overwrites my adjustment and sends the view careening off to the last page, whenever it scrolls so that the view becomes enlarged | 10:02.50 |
kens | Sounds bizarre. | 10:03.14 |
tor8 | I get all the expected input values, I set the damn value like it should be. and still it flies off completely the wrong way! | 10:03.25 |
| and of course, this is apple so documentation is lacking, all the apple fanboys on the forums have zero clue about how anything works, and there's no stinking source code... | 10:03.52 |
kens | But look at the shiny shiny :-) | 10:05.56 |
chrisl | And they'll still try to patent it......... | 10:06.15 |
Robin_Watts | finally gets mupdf building for android with the latest android tools. Time to rewrite the ReadMe, Again. | 10:06.28 |
tor8 | Robin_Watts: so I should hold off pulling that commit you have waiting for me? | 10:06.46 |
Robin_Watts | tor8: might make sense. | 10:25.38 |
| In my last job, it was me writing the buggy animation frameworks, so I have some sympathy with them :) | 10:40.25 |
tor8 | Robin_Watts: finally figured it out! took me the better part of a day and a half... | 11:06.25 |
| the contentOffset property was animated, the contentSize property was *not* animated, but did clip the contentOffset property during animation... | 11:06.49 |
| juggling the setting of all the various properties between the half dozen callbacks or so I managed to sort it out so everything gets set when expected | 11:07.35 |
Robin_Watts | oh, balls. I got the implication stuff the wrong way around. | 11:13.35 |
kens | Oops :-( | 11:13.55 |
henrys | power went off again yesterday if not obvious... A night like that makes me appreciate the furnace a bit more. | 13:37.42 |
kens | Hmm, electric heating ? | 13:38.23 |
henrys | we have gas but the furnace needs electricity to run. | 13:38.59 |
kens | Umm. Ours needs electricity to run the water pump. | 13:39.17 |
| But we also have a wood-burning stove ;-) | 13:39.30 |
henrys | quite an adventure we put all the refrigerator food in the backyard in the snow. Luckily the racoons weren't around. | 13:41.02 |
| shelly is looking for stuff to do if you have something let me know. IMHO it is good to have someone like him looking/learning about the code, so fixing stuff is not the sole motivation. | 13:44.45 |
| kens:support looks light | 13:51.28 |
Robin_Watts | henrys: So how long were you without power for? | 14:00.39 |
kens | henyrs, yes, nothing new today, I'm grateful ;-) | 14:00.54 |
Robin_Watts | A few years ago, the village I used to live in was without power for 3 or 4 days. (Fortunately, I'd moved out by then) | 14:01.27 |
henrys | there we 2 outages 1 am - 11 pm then 3 pm - 5 am | 14:02.40 |
| s/we/were | 14:03.32 |
Robin_Watts | ouch | 14:04.49 |
henrys | I don't know why these machines won't come up automatically after a power failure, I had to manually push the power button each time. Is there a bios option for that? | 14:05.50 |
Robin_Watts | It's a soft power thing. | 14:06.11 |
| The power switch isn't on/off. It's a momentary contact thing. | 14:06.28 |
arthurf | tor8: Re "grrr, f*ng buggy p.o.s. framework animation idiocy!!!" - so I guess that means I should hold off on iOS testing :D | 14:07.11 |
tor8 | arthurf: :) | 14:08.09 |
| I've got it mostly sorted, just need to sweep clean in the corners and commit then it's ready for a first bash of testing | 14:08.43 |
henrys | Robin_Watts:yes but there must be some way to have them restart unattended. | 14:09.20 |
| attaching a monitor to henrysx6 to see what happened. | 14:09.53 |
| it seems to be down | 14:10.07 |
Robin_Watts | Professor google says that supposedly some computers have it as a BIOS option. | 14:10.49 |
arthurf | tor8: Thanks. Developing for iOS can be a big challenge - and you're right - there is no source code to the frameworks - and they seem to have these little undocumented exceptions. I have a bunch of notes to myself about how such and such appears to *really* work | 14:11.04 |
Robin_Watts | henrys: but it sounds risky to me. | 14:11.52 |
| If I have a 'bouncing' power supply, I'd rather it didn't repeatedly try and restart. | 14:12.10 |
henrys | Robin_Watts:good point. | 14:13.33 |
Robin_Watts | tor8: kammerer did some timings on mupdf recently and found that for his (image heavy) files, 90%+ of the time was spent in the image scaling stuff. | 14:15.10 |
| The cores admit of fairly easy ARM coding. | 14:15.37 |
henrys | henrysx6 is back just a long fsck | 14:17.10 |
Robin_Watts | I'm going to have a play with that while my planar tests run. | 14:17.25 |
tor8 | Robin_Watts: sounds like a plan. | 14:42.17 |
| hopefully something we can leverage on the iphone | 14:42.27 |
Robin_Watts | tor8: was just typing that :) | 14:42.51 |
| OK, now to find some appropriate PDFs to test with. | 15:04.25 |
ray_laptop | unexpected, but many things are fixed with my bandsize changes (don't seem to recur elsewhere, either). halftone stitches in text on tests_private/customer_tests/md311dwf.pcl.ppmraw.600.1 for example | 15:20.28 |
kens | sounds positive | 15:21.40 |
henrys | tor8:when do we get mupdf 1.0 btw? | 15:21.48 |
tor8 | february if we stick to the current release schedule | 15:22.45 |
henrys | mvrhel2:did you get my mail about the customer you visited? If you want me to followup I can. Let me know. | 16:00.54 |
kens | I have to be off, night all | 16:01.11 |
henrys | bye kens | 16:01.18 |
Robin_Watts | night kens | 16:01.31 |
mvrhel2 | henrys: oh I missed that one. I will get an email off to them today | 16:01.47 |
Robin_Watts | mvrhel2: My planar tests just passed, I think. | 16:01.53 |
mvrhel2 | I thought I sent them an email | 16:01.59 |
| way back | 16:02.03 |
henrys | np thanks | 16:02.11 |
Robin_Watts | All the differences are now down to either clist banding things (causing phase shifts) or 'different blacks'. | 16:02.53 |
| So, I guess I should make a patch, and pass it to you so you can see what's involved in fixing your color halftoning stuff ? | 16:03.36 |
ray_laptop | Robin_Watts: mvrhel2: I also have a list of differences due to band height differences (and some are banding no banding). I will be opening bugs on them. Some are in PDF 1.7 FTS trans | 16:13.36 |
mvrhel2 | Robin_Watts: yes that sounds great | 16:15.06 |
ray_laptop | Robin_Watts: Some others change which lines are drawn in ps3cet files, eg. tests_private/ps/ps3cet/11-11.PS.pgmraw.300 | 16:15.08 |
Robin_Watts | mvrhel2: http://ghostscript.com/~robin/0001.patch | 16:17.18 |
mvrhel2 | Robin_Watts: ok. let me get finished up with this XPS bug (I am close to having this resolved) and then I will work on HT | 16:18.22 |
Robin_Watts | mvrhel2: Sure. No worries. | 16:19.10 |
mvrhel2 | henrys: what ever happened to the project to get rid of c++ key words in our code | 16:28.06 |
| drives me crazy when I want to look into a variable named template for example during debug | 16:28.28 |
henrys` | ray_laptop:I think we can skip the meeting today. | 16:30.33 |
mvrhel2 | oops henrys: you may have been off when I made my above comment | 16:33.51 |
ray_laptop | mvrhel2: oh, you mean 'template' and 'this' etc. | 16:34.48 |
mvrhel2 | yes | 16:34.53 |
henrys` | that was on alexcher's desk | 16:34.58 |
ray_laptop | mvrhel2: yeah, they really confuse VS display of variables | 16:35.27 |
mvrhel2 | exactly | 16:35.35 |
Robin_Watts | template, this, what else ? | 16:35.57 |
mvrhel2 | I generally run into template | 16:36.18 |
ray_laptop | Robin_Watts: I don't this "what" and "else" cause problems ;-) | 16:36.22 |
| s/this/think/ | 16:36.30 |
Robin_Watts | ray_laptop: :-p | 16:37.30 |
henrys` | do check with alexcher before fixing it. | 16:38.06 |
Robin_Watts | henrys: You probably missed the conversation I had with kens yesterday... | 17:09.26 |
| pxlasm/pxldis don't work under python 3. | 17:09.41 |
| They only work under python 2. | 17:09.46 |
| I made as many changes as I could to make pxldis work, and got it compiling, but it hit a runtime error. | 17:10.18 |
mvrhel2 | henrys: so, we should probably NOT be doing image interpolation in XPS if the output resolution is sufficiently high. According to the spec, interpolation SHOULD (their words) be used (weaker than MUST). Right now we are always doing interpolation | 17:26.58 |
| or it should be a command line option. I don't want us to be penalized time wise for doing interpolation | 17:41.25 |
Robin_Watts | We need a maximum interpolation resolution figure. | 17:42.21 |
| i.e. interpolate unless the image is greater than a given resolution already. | 17:42.49 |
mvrhel2 | yes | 17:42.55 |
henrys` | mvrhel2:which bug is this related too? | 17:44.55 |
mvrhel2 | henrys: no particular bug, but in my fix for 692505, which involved patterns with transparency, I also had an issue that cropped up in the image interpolation code and how it was interacting with the pdf14 device | 17:47.08 |
henrys` | python 3 eh - damn I was only interested in python for a short time, to bad the asm and disass work came about at the same time. | 17:47.11 |
mvrhel2 | which made me notice that we are always doing interpolation | 17:48.38 |
| for all the image rendering | 17:48.49 |
| certainly on a 2400dpi halftone device we should not be doing that :) | 17:49.24 |
henrys | I have pcl so it comes on with -dDOINTERPOLATE | 17:51.00 |
| and is off by default | 17:51.45 |
mvrhel2 | we should probably do the same in xps | 17:52.50 |
| to help when people are evaluating performance | 17:53.16 |
henrys | or just turn it off completely and not fool with it. | 17:53.33 |
mvrhel2 | I would think you would want the option. the windoze viewer uses interpolation of images. if people are doing screen comparisons it is good to have | 17:55.31 |
Robin_Watts | -dMAXINTERPOLATERES | 17:55.54 |
| and set MAXINTERPOLATERES to 600 by default. | 17:56.26 |
henrys | can we just do it if it is under 200 or something and not have an option? | 17:58.59 |
Robin_Watts | henrys: That's risky, as people often abuse the resolution to do scaling. | 17:59.22 |
henrys | XPS is inherently unrisky at this time. | 18:00.49 |
Robin_Watts | I was thinking of this as being useful for all products, not just XPS. | 18:01.42 |
henrys | okay, I don't feel strongly about it. Seems like DOINTERPOLATE has done the job fine, but I can see your new option would be a marginal improvement over just that. | 18:04.29 |
| bbiab | 18:04.41 |
Robin_Watts | DOINTERPOLATE is a blunt instrument (but a very useful one) | 18:05.14 |
mvrhel2 | ok. so digging into this xps bug appears to have resulted in some pdf rendering progressions with interpolated images that have transparency | 18:05.17 |
| for gray devices | 18:05.21 |
| cool looks like a lot of progressions in pdf and xps with this one | 18:09.06 |
Robin_Watts | always nice | 18:09.19 |
ray_laptop | I had hopes that I could (on x86 platforms) use the 'cpuid' with eax set to 3 in order to make a serialnumber. doesn't seem to work on peeves :-( | 18:29.08 |
henrys` | wonders why we should have progressions in pdf regression test files. | 18:30.33 |
ray_laptop | but when I installed the 'cpuid' it displays the same thing I get: | 18:33.12 |
| eax in eax ebx ecx edx | 18:33.14 |
| 00000003 00000000 00000000 00000000 00000000 | 18:33.16 |
| but somehow it displays: Processor serial: 0001-06A4-0000-0000-0000-0000 | 18:33.44 |
| this more like a UUID that is the all zeroes s/n from cpuid and the first 32 bits are the return of stepping, model, and family information from cpuid when eax-in is 1 | 18:38.23 |
mvrhel2 | henrys: yes, I was surprised to see this too | 19:02.46 |
| 692512 is an interesting one. | 19:03.14 |
ray_laptop | henrys: is 'macpro' yours ? That cluster machine frequently goes down so the jobs have to go to other nodes | 19:19.49 |
henrys | I wasn't aware macpro went down frequently other than the power failures over the last couple days. | 19:21.52 |
| there was a brief networking problem just now, but you say this is frequent? | 19:24.06 |
ray_laptop | henrys: sorry -- wasn't paying attention. I often see just 'macpro' doing that. Don't you have another cluster machine there too ? | 19:30.48 |
| if henrysx6 is also yours, it should behave the same if it is power or network | 19:31.21 |
| I don't think the cluster master logs when that happens, but we can ask marcosw when he finishes doing fun, but foolish, activities | 19:32.35 |
ray_laptop | doesn't even think it sounds that fun | 19:33.08 |
henrys` | yes let's ask him. | 19:33.16 |
| I'd have 0 interest in racing around mexico. | 19:34.02 |
ray_laptop | but, to each, their own choice. Some people consider skiing fast down a steep hill foolish and not fun (but not me) | 19:34.50 |
henrys` | yep | 19:35.38 |
Robin_Watts | as for those nutters who go running without being late for anything, or being chased by anything... | 19:36.42 |
ray_laptop | luckily, living where I do, it is sort of self limiting by distance and $. If I lived in Mammoth, CA, I'd probably have had a terminal tree event by now | 19:36.50 |
Robin_Watts | ray_laptop: When you went for a cruise, what ship were you on, btw? | 19:37.55 |
ray_laptop | uhh, Carnival Sensation, iirc | 19:38.14 |
henrys` | my thing is coming down these canyons on my bike - a guy just got killed about a month ago... it is pretty dangerous.... but I like the speed thing. | 19:38.30 |
ray_laptop | it was not too bad | 19:38.33 |
Robin_Watts | just saw a programme on "The Allure of the Seas". | 19:39.03 |
| That thing is just mind blowingly big. | 19:39.10 |
ray_laptop | henrys: well, be careful. The world's knowledge 'meme' level would drop by a significant percentage if anything happened to you | 19:39.49 |
| Robin_Watts: my wife wants to do a Disney cruise with the kids next (before they are too old for it) | 19:40.34 |
henrys` | I did the cruise thing with my kids when they were little, be prepared to gain 5 lbs - it seems we were always eating. | 19:41.20 |
Robin_Watts | I've never cruised. | 19:41.52 |
| Maybe to Antarctica, or around the north of norway, or the inside passage. | 19:42.13 |
ray_laptop | henrys: chasing around after 3 kids and trying to also stay up after the kids went to bed and do grown-up activities, I actually didin't gain anything. | 19:42.41 |
| but then it was only 4 days. | 19:42.54 |
henrys` | ray_laptop:oh you watch them and all... ;-) | 19:43.14 |
ray_laptop | henrys: well, they LOVE to swim, and I love the water, too, so I got _some_ exercise | 19:43.55 |
| those big boats look like they're going slow, but they're hard to keep up with when you're swimming ;-) | 19:45.53 |
henrys` | ;-) | 19:47.21 |
ray_laptop | if anyone has any ideas on "automatic" serial numbers, let me know. Otherwise we'll just leave it up to our customers | 19:48.20 |
| errand time... bbiaw | 19:48.35 |
henrys` | I guess I need to do some bisecting I'll be glad when marcos gets back. | 19:49.53 |
Robin_Watts | OK, I have the common cases for the scaler ARM coded. | 20:04.58 |
| I'll do some timings and sort out a commit tomorrow. | 20:05.22 |
mvrhel2 | Robin_Watts : you still around? | 21:02.32 |
| bbiab | 21:08.34 |
Robin_Watts | mvrhel2: Just checking in... | 21:27.48 |
| mvrhel2: ping? | 23:10.58 |
mvrhel2 | hi Robin_Watts | 23:11.56 |
| I was going to ask you a drawing question | 23:12.03 |
Robin_Watts | go for it | 23:12.10 |
mvrhel2 | So one of the XPS bugs that I tracked down, I have made a simple example with the issue | 23:12.24 |
| it basically is a series of strokes of width = 1 | 23:12.35 |
| horizontal lines | 23:12.43 |
| starting at 0 0 to 10 0 | 23:12.52 |
| then 0 1 to 10 1 | 23:13.05 |
| etc | 23:13.07 |
| this goes on for about 10 lines | 23:13.17 |
| when this is rendered at 150 dpi, we end up with missing lines | 23:13.33 |
| well the lines are not missing | 23:14.12 |
| but in the shuffle to scale them, we end up missing a fill area | 23:14.39 |
| let me throw up an image for you to see | 23:14.52 |
Robin_Watts | mvrhel2: What is fill adjust for xps ? | 23:15.08 |
mvrhel2 | hold on phone | 23:15.50 |
Robin_Watts | And are those coords in device space. | 23:15.54 |
| no worries, I'll wait. | 23:15.58 |
mvrhel2 | sorry I am back | 23:17.57 |
| so a stroke at line zero should be half off the page | 23:18.57 |
Robin_Watts | mvrhel2: Right. | 23:19.08 |
mvrhel2 | a pixel is considered to be from X to X+1 | 23:19.13 |
Robin_Watts | ok. | 23:19.19 |
| And do we paint a pixel if any part of it is covered? Or if its pixel centre is covered? Or some other rule? | 23:19.43 |
mvrhel2 | if the center is covered we should fill the whole pixel according to XPS rules | 23:20.05 |
| so, with that rule, and the above commands we should never have any drop outs | 23:20.25 |
Robin_Watts | right enough. | 23:20.38 |
| So, are those coords in device space then ? | 23:20.54 |
mvrhel2 | the coordinates that we have coming in are at 96dpi native. The xps interpreter sets the CTM to get mappings to other dpi in the graphics library | 23:21.39 |
Robin_Watts | Oh, I guess you'd argue that the transformation matrix is irrelavent, cos all the lines should abut. | 23:21.39 |
mvrhel2 | yes | 23:21.44 |
Robin_Watts | OK, so it's probably floating point rounding issues then. | 23:21.57 |
| but I can look into it, sure. | 23:22.07 |
| Send me the file, and I'll have a look tomorrow. | 23:22.22 |
mvrhel2 | ok. | 23:22.35 |
| will do | 23:22.37 |
Robin_Watts | The fill adjust bletch is a nightmare :( | 23:22.47 |
mvrhel2 | yes, I started to look at it and with this cold I am fighting thought you might be a better candidate since you have been into it before | 23:23.23 |
| the example I have is very simple. | 23:23.29 |
| I need to step out for a bit. | 23:23.36 |
| I will get it to you for your morning | 23:23.42 |
Robin_Watts | ok. I'm off to bed in a mo. | 23:23.45 |
| tor8: ARM versions reduce the total rendering time by 25% for my test pdf. | 23:49.52 |
tor8 | Robin_Watts: cool! | 23:50.04 |
| I've got a nasty test file for my iPad, it takes a couple of seconds per page | 23:50.22 |
| full screen bitmap backgrounds | 23:50.31 |
Robin_Watts | I'll commit them tomorrow and we'll see how iOS works with it. | 23:50.41 |
tor8 | Robin_Watts: the ios branch should be in serviceable shape if you want to give it a spin | 23:50.59 |
Robin_Watts | I'm not set up for ios dev :( | 23:51.14 |
tor8 | the rudimentary bits are in place, now it just needs a few more features, XPS support and a lot of spit, shine, and polish | 23:51.21 |
| I've got smooth scrolling/page flipping with page rendering in a background thread | 23:51.50 |
| a table of contents view | 23:52.01 |
mvrhel2 | Robin_Watts; one other question for you | 23:52.03 |
Robin_Watts | mvrhel2: Go for it. | 23:52.37 |
tor8 | Robin_Watts: I could send you an .ipa file if you want to give it a spin on your iPod touch | 23:52.48 |
mvrhel2 | any idea why when I run bmpcmp on the xps files as a cluster test, the files show up with red and blue swapped | 23:52.53 |
tor8 | but I'll leave that till tomorrow | 23:52.56 |
mvrhel2 | that is the ppm files | 23:53.13 |
Robin_Watts | mvrhel2: That's very strange. | 23:53.24 |
| It uses ppmraw ? | 23:53.36 |
mvrhel2 | yes | 23:53.42 |
Robin_Watts | Other ppms look ok in bmpcmp. | 23:54.01 |
mvrhel2 | yes | 23:54.06 |
Robin_Watts | How can it be different from xps? | 23:54.08 |
| Got a link? | 23:54.13 |
mvrhel2 | I wonder the same thing | 23:54.16 |
Robin_Watts | sorry, it's your cluster bmpcmp of course. Stupid question. | 23:54.27 |
mvrhel2 | yes, I think maybe look at the last page | 23:54.52 |
| hold on | 23:54.58 |
Robin_Watts | 397 for instance. | 23:55.25 |
mvrhel2 | exactly | 23:55.43 |
Robin_Watts | That's just bonkers. | 23:55.53 |
mvrhel2 | hehe | 23:55.58 |
Robin_Watts | I *will* look into that tomorrow, cos I want to understand it :) | 23:56.05 |
mvrhel2 | ok. cool thanks | 23:56.11 |
| they come out fine when I run to the ppmraw device stand alone | 23:56.29 |
Robin_Watts | I must go to bed. See y'all tomorrow. | 23:57.06 |
mvrhel2 | good night | 23:57.10 |
| Forward 1 day (to 2011/10/28)>>> | |