| <<<Back 1 day (to 2012/11/19) | 2012/11/20 |
Robin_Watts | mvrhel_laptop: It is. | 00:03.25 |
mvrhel_laptop | Robin_Watts: well happy birthday | 00:04.58 |
| I think it is 12:05 there yes? | 00:05.16 |
Robin_Watts | mvrhel_laptop: It is, and thanks :) | 00:06.51 |
sebras | Robin_Watts: congrats! :) | 00:16.14 |
| Robin_Watts: so your turning 28 just like me! (but in base 16.5!) | 00:18.04 |
Robin_Watts | 0x29 sounds so much better :) | 00:24.43 |
| Hmm. my ARM coded CMYK -> RGB is wrong in the green plane somehow... | 00:25.16 |
sebras | Robin_Watts: which commit? | 00:39.28 |
Robin_Watts | not committed yet. | 00:39.36 |
| (It's on my private repo and I just found a typo. Trying to build to test it now. | 00:39.58 |
sebras | Robin_Watts: ah. | 00:40.06 |
Robin_Watts | Is there a bit of makefile gubbins to tell a makefile to ignore the timestamps on certain dependencies? | 00:44.13 |
| The ndk appears to make directories in the obj dir for fitz etc. | 00:44.34 |
| Then it makes all the files in there depend on those dirs. But it seems to timestamp the dir AFTER each file within in, so every build ends up building from clean. it's driving me potty. | 00:45.23 |
sebras | Robin_Watts: I guess you you could add this to the end of jni/Android.mk: | 00:48.32 |
| $(linked_module): dummytarget | 00:48.57 |
| dummytarget: | 00:49.00 |
| touch dira dirb | 00:49.09 |
| I have no idea why ndk misbehaves for you though. | 00:49.25 |
Robin_Watts | No, that'd make it worse, I think. The problem is I want it NOT to change the directory timestamps each time. | 00:50.31 |
| OK, better results, but still a yellowish cast. | 00:51.34 |
sebras | I know, but if you can make sure that the timestamp is "correct" relative its dependencies then its fine. | 00:51.39 |
Robin_Watts | Will hunt more in the morning. | 00:51.42 |
sebras | Robin_Watts: good night! | 00:52.08 |
Robin_Watts | sebras: You're not running r8c under windows are you? | 00:52.10 |
sebras | Robin_Watts: r8b under linux. | 00:52.50 |
Robin_Watts | I may try dropping back to r8b while I debug this. | 00:53.12 |
| I need r8c for the profiler, I think. | 00:53.21 |
| ho hum. bedtime. night. | 00:53.35 |
sebras | oh, I though it was just clang making a mess in r8c... | 00:53.42 |
ball | If I want to print to a network printer that runs lpd, is there any way of predicting the queue name? | 02:13.04 |
saper | ball: try "lp" :) | 02:22.20 |
ball | saper: Thanks, I'll try that. | 02:26.06 |
saper | no warranty, though | 02:26.40 |
henrys | ball:lpq | 02:32.43 |
ball | I hope it works but I won't get to test it until tomorrow | 02:32.47 |
| henrys: I'll try that too. | 02:32.55 |
| Does it make a difference if the printer includes postscript support? | 02:33.17 |
henrys | nope | 02:33.37 |
| from the man page lpq shows the current print queue status on the named printer. Jobs queued on the default destination | 02:33.54 |
| will be shown if no printer or class is specified on the command-line. | 02:33.54 |
ball | Will lpq list the queues on the printer? | 02:35.11 |
| (will it include empty queues?) | 02:35.23 |
henrys | nope just the default if no arguments, lpc status all says to see all the queues | 02:41.10 |
| google for lpc | 02:41.32 |
ball | writes that down | 03:02.04 |
kens | tor8 Robin_Watts another Stack Overflow question. I'm fairly sure the answer is 'no' but can one of you confirm ? | 08:09.05 |
| http://stackoverflow.com/questions/13455846/is-it-possible-to-merge-pdf-files-into-one-with-mupdf-on-android | 08:09.05 |
sebras | kens: we only compile the view app for android, not the mutool. | 08:13.57 |
kens | sebras thought so | 08:14.08 |
sebras | kens: and that tool is as far as I know not yet capable of merging two pdf. | 08:14.19 |
| though the tool is developing in the right direction to do it I guess. :) | 08:14.32 |
| but you'd need a gui for it on android too. | 08:14.47 |
kens | Probably. Do you know any android app which cxan do this ? I think not... | 08:14.56 |
sebras | no, never heard of one. | 08:15.04 |
kens | I guess I'll reply... | 08:15.13 |
sebras | but as Robin says in his spiel -- mupdf can do anything, it's mostly just the app that doesn't provide an interface for it. ;) | 08:15.36 |
| kens: I think you can, and if you're wrong I'm sure tor8 and robin will vote you down. :-D | 08:16.10 |
kens | :-) | 08:17.46 |
Viorel | hello! is anything wrong with the current branch? I had an old version, did a "git pull", followed by a "git submodule update --init", then, after a "make clean", when running "make" it gives me an error on "fitz/doc_search.c:3:19: error: expected identifier or â(â before â__extension__â" | 08:50.35 |
| can anyone confirm the issue? | 09:32.12 |
kens | Viorel : You're a bit early in the day, in an hour or so someone should be able to answer your question. However I'm not aware of any build problems. It would be helpful if you could at least state the OS and compiler you are using. | 09:37.06 |
Viorel | yes, I know, it's a bit early. I'm using Win7 x64 and cygwin. | 09:38.02 |
kens | So are you compiling for WIndows, or Cygwin ? | 09:38.21 |
| Or tather, using MS Visual Studion or Gcc | 09:38.35 |
Viorel | I'm compiling from cygwin | 09:38.49 |
| with gcc | 09:38.55 |
kens | Well, if it was me I wouldn't do that, but I'll have to leave it up to someone who knows more than me about MuPDF | 09:39.16 |
| Do you know the SHA for the copy you pulled ? | 09:39.47 |
Viorel | commit b1e3f310637a22c6b4c5e142e3582f90ff467ddf Author: Robin Watts <robin.watts@artifex.com> Date: Mon Nov 19 16:40:51 2012 +0000 | 09:40.49 |
| I guess that's the SHA, right? | 09:40.57 |
kens | Yep | 09:41.01 |
| I'll probably have to wait for Robin_Watts to try that out though | 09:41.24 |
Viorel | ok, thanks :) | 09:41.33 |
kens | I'm updating my checkout but I'm not using Cygwin | 09:41.44 |
kens | goes to fetch a coffee while Git works | 09:42.57 |
paulgardiner | Robin_Watts: ping | 09:50.47 |
kens | I'm guessing he's not here yet Paul | 09:50.58 |
| Since he hasn't said anything about the build problem from viorel | 09:51.13 |
paulgardiner | kens: oh ok. I'll catch him later | 09:51.41 |
chrisl | Viorel: I get the same error on Cygwin - no idea why. You probably will have to wait for someone who really knows about the mupdf build | 09:55.31 |
| Actually, I *do* know why: "tolower" is defined as a macro in Cygwin, but is a function most other places. | 10:00.33 |
kens | Windows build using Visual Studio works fine, the Linux build obviously works OK, as the regression was fine, so it must be something Cygwin-weird | 10:00.53 |
chrisl | kens: see above.... | 10:01.07 |
kens | Yeah, your post was just as I was pressing return | 10:01.23 |
Viorel | I see, so I should switch to either VS or linux to do the build | 10:04.11 |
kens | Well, if you want to build for WIndows I would use Visual Studio | 10:04.29 |
| But it 'should' work under Cygwin, maybe someone will look at it | 10:04.49 |
Viorel | well, as I don't have VS installed, I'l boot up the Linux, and return after, maybe Robin will be up - by the way, what time is it now there? | 10:06.02 |
kens | Currently it is 10:05 am | 10:06.16 |
Viorel | thanks! | 10:06.31 |
kens | But Robin_Watts and tor8 start late and work into the night | 10:06.31 |
chrisl | It's easy to fix - open fitz/doc_search.c, search'n'replace "tolower" with "my_tolower" and the build should work - it doesn't like the X11 viewer, through....... | 10:06.34 |
Viorel | ah, that's good to know, I don't need X11 | 10:06.57 |
kens | chrisl remember I said there was a problem with COmpatibilityLevel=1.3 and SMask ? SOmeone has found it :-( | 10:09.29 |
Viorel | chrisl: thank you, that worked | 10:09.40 |
chrisl | kens: yeh, I noticed that - it was inevitable :-( | 10:09.56 |
kens | :-( | 10:10.05 |
| Probably going to have to get Micahel involed with it, eventually | 10:10.18 |
chrisl | Viorel: I'm going to put that patch up for Tor to consider | 10:10.29 |
| kens: would it worthwhile putting it off until your colour improvements are done? | 10:10.56 |
kens | I think it probably is, yes | 10:11.06 |
Viorel | I'm glad I could help by bringing up a bug | 10:11.07 |
kens | chrisl I'm just quickly haking at it to prove its what I think it is | 10:11.36 |
chrisl | I'm just assuming it's to do with colour conversions, since Smask handling normally works - unless it's the issue Alex is working on | 10:14.25 |
kens | THe Smask is OK normally | 10:14.37 |
| Its pdfwrite-specific | 10:14.46 |
| But I don't think its exactly colour conversions, the SMask just seems to be ignored | 10:15.04 |
| (as far as transparency goes) | 10:15.11 |
| Of course, the pages calls a form, whcih calls a form, which executes the image.... | 10:15.36 |
chrisl | Yeh, I thought it *might* be the SMask problem we know about, it just *happens* that with pdfwrite involved, it exposes problem in this case | 10:15.39 |
kens | And its not even Cairo this time | 10:15.43 |
chrisl | Who made it this time? | 10:16.01 |
kens | Quartz | 10:16.06 |
chrisl | Erm, Apple? | 10:16.15 |
kens | Yeah the OS/X built-in thing | 10:16.25 |
| THere is colour conversion going on, since the original image is in a ICCBased (calibrated RGB) space and we're trying to convert to CMYK | 10:17.41 |
| But I'm not sure its the problem, I think its just not applying the SMask | 10:17.57 |
chrisl | I wondered if we were colour converting the Smask | 10:18.20 |
kens | I don't think so, at least, not yet, we may be after I get finished ;-) | 10:18.38 |
| Yeah, if I remove the SMask from the original PDF I get exactly the same effect as with pdfwrite. It very much looks like we are 'dropping' the SMask | 10:20.40 |
chrisl | Hmm, very odd | 10:20.55 |
kens | Yes, especially since we recognising the transparency and rendering to an image | 10:21.21 |
chrisl | Oh, this isn't one of the places where the PDF interpreter does something different for pdfwrite, is it? | 10:21.38 |
kens | Probably. pdfwrite normally handles transparency itself. | 10:22.01 |
| So the pdf14 compositor never gets pushed | 10:22.12 |
chrisl | I assumed pdfwrite would push the compositor itself, when needed | 10:22.47 |
kens | When we want to downgrade to non-transparent output (PDF 1.3, PDF/X, PDF/A etc) then we push the compositor in pdfwrite | 10:22.49 |
chrisl | :-) | 10:22.56 |
kens | WHich I believe we are doing, as the resulting PDF contains an image. I guess I'll have to check | 10:23.17 |
| Like I said, may need Michael to help on this | 10:23.31 |
| Well we definitely create a compositor | 10:31.24 |
| and that appears to be it. | 10:31.42 |
| Never seem to push a smask | 10:31.49 |
| Which would sort of explain the problem | 10:31.59 |
| It looks like pdfwrite is expecting to get a PDF14_BEGIN_TRANS_MASK and it never does | 10:32.37 |
| So I think I need Micahel now. I'll leave it untilo the meeting | 10:32.55 |
Robin_Watts | viorel: Load doc_search.c. Change "static inline int tolower(int c)" on line 3 to be "static inline int (tolower)(int c)" | 10:39.17 |
| Does that solve it ? | 10:39.22 |
| paulgardiner: Morning. | 10:40.23 |
paulgardiner | Robin_Watts: hi | 10:40.29 |
| That fix for the async task wont work. If the task isn't running then the first attempt to show an alert will stall. | 10:41.06 |
Robin_Watts | oh. | 10:41.22 |
| why not ? | 10:41.30 |
paulgardiner | How do you mean? | 10:42.02 |
Robin_Watts | I mean, what does the fix do wrong to stop it working? | 10:42.15 |
paulgardiner | Maybe I've read it wront. I thought it called the method only if present, with no back up case for if it doesn'e exist. | 10:43.09 |
Robin_Watts | Instead of calling foo.bar(baz), my code says "if object foo has a method bar, then call it with baz" | 10:43.09 |
| Right, that's exactly what it does. | 10:43.21 |
| I thought you said that that code basically set up the default method for threading for the AsyncTask. | 10:43.42 |
paulgardiner | Yeah, so if the method doesn't exist then the task doesn't get started. | 10:43.48 |
| Ah no. | 10:43.57 |
Robin_Watts | And in cases where that call did not exist, we were using the standard default method anyway. | 10:44.00 |
paulgardiner | It actually starts the task | 10:44.02 |
Robin_Watts | Right. I misunderstood. | 10:44.08 |
| Do you want to pull my change in, fix it, and put it up for review? | 10:44.34 |
| It sounds like we need an "if (m == null) ... " but | 10:45.14 |
| s/but/bit/ | 10:45.17 |
paulgardiner | Well, even if we call execute in the case that executeOnExecuter doesn't exist, we might be on an Android version where all async tasks run on a single thread, and then it still wont work | 10:46.15 |
sebras | Viorel: may I ask what you are using mupdf in cygwin for? I haven't heard of anyone using it in cygwin before...? | 10:46.23 |
paulgardiner | I think the only fix is to import the latest asynctask into the app. | 10:46.34 |
sebras | and we're always eager to understand how people use mupdf. | 10:46.41 |
paulgardiner | ... I have the android source here so I should be able to find it | 10:47.07 |
sumek | hi everyone! I'm using ghostscript to convert pdf to png and getting a segmentation fault. What would be a good first step of troubleshooting? | 10:59.25 |
Robin_Watts | sumek: Give us the exact command line you are using, and put the PDF file somewhere we can get it. | 11:01.18 |
sumek | the command is gs -dBATCH -dNOPAUSE -sDEVICE=png16m -dTextAlphaBits=4 -r300 -dGraphicsAlphaBits=2 -sOutputFile=%03d_big.png chp%3A10.1007%2F978-3-642-31454-4_14.pdf | 11:02.38 |
| I don't think I can send you the pdf though... :/ It's a copyrighted material | 11:03.09 |
| any hints you could give me without the pdf? | 11:03.28 |
| ok, got a go ahead to send the pdf | 11:05.27 |
| could you give me your email? the pdf is not publicly available | 11:05.45 |
Robin_Watts | robin@wss.co.uk | 11:06.02 |
Robin_Watts | will be out for a bit, so maybe someone else can look. | 11:06.52 |
Viorel | sebras, I'm using it as I'm on Windows, don't have VS installed, and I already need cygwin for android native development | 11:09.43 |
| Robin_Watts, I took chrisl's advice and changed it to my_tolower, and it worked that way | 11:11.56 |
chrisl | sumek: what version of gs are you using? And from where did you get it? | 11:18.55 |
sumek | it's 9.04 | 11:19.06 |
chrisl | Oh, so first suggestion would be to try 9.06 | 11:19.26 |
sumek | compiled from source, no special switches | 11:21.05 |
chrisl | Well, like I say, grab 9.06, and try that | 11:21.25 |
sumek | worked with 9.06. Thanks! | 11:25.00 |
chrisl | sumek: okay, that's good. Be aware, there was a bug in interpolation in 9.06 which can also cause a crash | 11:25.44 |
| sumek: the patch for that interpolation problem is here: http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=d527eade | 11:27.27 |
| tor8: cygwin build fix for your perusal in my mupdf repo | 12:20.37 |
tor8 | chrisl: ah, thanks | 12:22.49 |
paulgardiner | Robin_Watts: Here's a new apk with the app using AsyncTask from the latest android source: http://intranet.glidos.net/~paul/MuPDF20121120-forms.zip | 12:41.10 |
| Robin_Watts: I've pushed the changes too. | 12:41.23 |
Robin_Watts | paulgardiner: Fab. | 12:42.31 |
tor8 | Robin_Watts: paulgardiner: whitespace formatting OCD patch on tor/master for review | 12:51.10 |
paulgardiner | tor8: that sounds easy enough that I can do it. :-) | 12:51.44 |
tor8 | paulgardiner: :) | 12:52.28 |
paulgardiner | ... assuming it's whitespace changed in the code, and not subtle changes to the whitespace in displayed documents, that is! | 12:53.25 |
Robin_Watts | tor8: hmm. Can we not change the whitespace in the ARM code please? | 12:54.13 |
tor8 | Robin_Watts: I left the indentation, just fixed some inconsistent space/tabs | 12:54.52 |
Robin_Watts | oops. Had I committed the ARM code in res_colorspace yet ? | 12:55.04 |
tor8 | Robin_Watts: the #if 0 code yeah. I got a compiler warning there about an unterminated string and that's how I got to obsessing about whitespace again... | 12:55.43 |
Robin_Watts | oh, bugger. I removed that once, then backtracked in git and forgot it. Sorry. | 12:56.05 |
| I've *almost* got the ARM code working right. | 12:56.31 |
paulgardiner | tor8: Looks like you've fixed an unclosed quote in one place. Surprised that didn't lead to build failure. | 12:57.35 |
tor8 | paulgardiner: it was in a #ifdef chunk that wasn't used. but it did give build warnings. | 12:57.58 |
paulgardiner | Looks fine | 13:00.32 |
sebras | tor8: do we usually test compiling mupdf under cygwin? | 13:05.42 |
tor8 | sebras: no, I've never done that :) but I think you can build the android stuff in cygwin | 13:06.19 |
Robin_Watts | Reverting to r8b appears to have solved the dependency madness. | 14:20.07 |
| paulgardiner: And your fix seems to work. | 14:20.43 |
paulgardiner | Fab | 14:21.10 |
Robin_Watts | So shall I push all 3 of those? | 14:22.24 |
paulgardiner | Yep. | 14:25.39 |
sebras | tor8: oh, I bet that this is a build environment not officially supported by google. | 14:28.00 |
Robin_Watts | sebras: hmm? | 14:28.41 |
| The official way to be able to build android on windows is to use cygwin. | 14:29.00 |
| it's what I use (and possibly it's the cause of the dependency brokenness in r8c) | 14:29.24 |
sebras | Robin_Watts: oh, I see. now you know I always build in linux. :) | 14:33.12 |
Robin_Watts | sebras: Me windoze boi. | 14:33.52 |
| hmm. For some reason, I'm getting errors while trying to compile pdf_js.c | 15:05.06 |
| oh. I see why. Missing { | 15:05.47 |
henrys | FYI mupdf is unable to open US 1040 tax form, the default pdf viewer on the nexus works. That's the main tax form in the US.. something we should have working. | 15:18.01 |
paulgardiner | henrys: Yes, that sounds like a good one to have working. Do you have a copy easily to hand? | 15:21.22 |
henrys | http://www.irs.gov/pub/irs-pdf/f1040.pdf | 15:23.42 |
| or it is the first entry googling 1040 tax form | 15:24.55 |
Robin_Watts | Aha! | 15:27.04 |
| I'd missed blue out of the conversion. | 15:27.17 |
sebras | henrys: unable to open or unable to use the form? I succeed in using mudraw on that file. | 15:29.01 |
paulgardiner | henrys: what goes wrong with f1024.pdf? I seem to be able to open it on windows and fill in some fields at least. | 15:29.39 |
sebras | paulgardiner: could this be on the form-preview for android? | 15:32.43 |
| since he mentions nexus... | 15:32.54 |
paulgardiner | I'll try it on my s3 | 15:33.06 |
Robin_Watts | paulgardiner, tor8: ARM code review on my repo. | 15:33.09 |
henrys | yep 1st time I downloaded it from the browser it said can't open the file, now it opens strange. | 15:33.36 |
paulgardiner | Yeah odd | 15:35.10 |
Robin_Watts | Corrupt download? | 15:37.03 |
henrys | I'm trying to repeat the procedure but now I'm having difficult downloading the pdf in droid chrome. | 15:39.57 |
ray_laptop | just blame it on the IRS their specialty is collecting, not distriibuting | 15:42.32 |
Robin_Watts | paulgardiner: How hard would it be to sort the list of files in the menu? | 15:46.50 |
henrys | it is reproducible I erased it downloaded it nexus gives me a choice of 2 pdf viewers, I choose mupdf and "Unable to open document" Dismiss | 15:47.06 |
Robin_Watts | sebras: IRT3000.pdf runs very fast on my phone with the latest version. | 15:47.59 |
henrys | once i open it in the other viewer then I can start mupdf and it appears in the list of pdf files and it opens fine. | 15:48.30 |
Robin_Watts | Ah. Looks like it's a heavy user of CMYK to RGB. | 15:48.43 |
paulgardiner | henrys: can you open it from the download directory? Have you had other downloaded pdfs work? | 15:48.51 |
| henrys: it just worked for me on my s3, downloading directly from the irs site. I guess it has to be nexus specific | 15:49.38 |
Robin_Watts | Some versions of android do wierdass things when invoking applications on downloaded files. | 15:50.35 |
henrys | oh right nothing from the downloads opens directly. I have to download then open mupdf and it appears in the mupdf list. | 15:51.12 |
Robin_Watts | Rather than passing in the real filename (/sdcard/Downloads/foo.pdf) they pass in some other string, like /blah/123 | 15:51.13 |
paulgardiner | Yes, I think the transformer prime does, but we worked out how to handle it | 15:51.15 |
Robin_Watts | And in order to handle filenames like /blah/123 you need to have a particular permission on the app. | 15:51.48 |
| BUT you can only grant that permission if your app is part of the inbuilt software on the device. | 15:52.08 |
| So unless we want to start building our own android images, we are out of luck. | 15:52.36 |
paulgardiner | Some forms of the weirdness can be handled. We handle some forms of content:// url | 15:53.49 |
| If we could log what we're being asked to open, we might be able to find a workaround. | 15:54.30 |
henrys | seems very strange that all these mupdf viewers would behave that way and we wouldn't be getting a load of user complaints | 15:54.48 |
Robin_Watts | paulgardiner: I went through all that a while ago. | 15:54.59 |
paulgardiner | Robin_Watts: which device? Does it still work on the transformer prime? | 15:55.50 |
Robin_Watts | The transformer was the device in question, I think. | 15:56.08 |
paulgardiner | I had that working... maybe before the upgrade to ICS though | 15:57.26 |
Robin_Watts | Maybe it was Helens S3. | 15:57.54 |
paulgardiner | commit 150e7f9 is the one that made the transformer prime work | 15:58.37 |
| Ah. File manager | 15:59.20 |
Robin_Watts | Opening via the file manager always works. | 15:59.40 |
paulgardiner | It didn't orginally on the transformer prime because of this content:// thing | 16:00.08 |
Robin_Watts | right. | 16:00.30 |
sebras | Robin_Watts: yes, IRT3000.pdf uses CMYK a lot. | 16:00.39 |
henrys | meeting time. How's performance for beta? | 16:01.10 |
Robin_Watts | henrys: I've committed some stuff that improves it significantly on the slower files that I found. | 16:01.35 |
| specifically, we now have image subsampling implemented. | 16:01.53 |
| and a faster CMYK -> RGB conversion. | 16:02.02 |
| I'd like to ARM code the subsampler. | 16:02.18 |
| I looked through the first 100-200 files in the PDFIA subset and only found maybe 10 that were sluggish/slow. | 16:02.50 |
| and the ones I have checked since are better. | 16:03.03 |
| Now I have the profiler hooked up, it's a much easier job. | 16:03.14 |
henrys | that's very good news. | 16:03.14 |
| the plrm and technical manuals I have seem good. | 16:03.29 |
Robin_Watts | yeah, text stuff is fine. | 16:03.40 |
| Oh, and I optimised a bit of the fax decoder too. | 16:03.57 |
henrys | have we tested scanned material. A lot of pdf material is just a scan + ocr | 16:04.32 |
| ? | 16:04.35 |
Robin_Watts | henrys: Yes. | 16:04.40 |
tor8 | yup. lots of good optimizations from robin this week! | 16:04.43 |
henrys | Robin_Watts:so do you feel ready for the show? | 16:05.54 |
| do you think anything other than what is on paul's list should be done for the show? | 16:06.18 |
Robin_Watts | henrys: Yeah, I think we can have a decent stab at anything people throw at us. | 16:06.24 |
| Urm... I can't immediately think of any glaring ommisions. | 16:06.48 |
| except in my spelling. | 16:06.57 |
| I'd like to have a look at the shading stuff at some point. | 16:07.30 |
henrys | paulgardiner:alerts look pretty challenging. | 16:07.50 |
tor8 | I'd like to revamp the android UI. the icons we use seem to be a bit hit-or-miss with the standard button look (since they change it every release, and I'm sure different vendors have different looking buttons too). I think we might be better off if we use a completely custom style for the buttons. | 16:07.54 |
| I did some experiments a while ago with setting the button backgrounds and colors with the XML. | 16:08.20 |
henrys | tor8:before the show you mean? | 16:08.23 |
paulgardiner | henrys: Yeah, it's all the threading. I ran into another bad case since writing the report, but that;s fixed and it seems hardy enough now. | 16:08.37 |
tor8 | henrys: hopefully, since it might look terrible on some devices :/ | 16:08.49 |
Robin_Watts | We used to break shadings down to 32x32 meshes in the parse phase, but that used a vast amount of memory for the display list. So we changed to breaking them down to a 32x32 mesh at render time. We should be able to break them down to less than 32x32 now based onto the zoom level (so we may only have to go to 8x8 or something) | 16:09.14 |
henrys | tor8:you don't have a lot of time | 16:09.14 |
tor8 | henrys: how much time remains? | 16:09.35 |
Robin_Watts | show is in 14 days or so? | 16:09.42 |
| Dec 4-7 | 16:10.08 |
| so we need to have scott and miles up and running with the final version by the end of next week I guess. | 16:10.34 |
henrys | what are they using for devices? | 16:10.55 |
Robin_Watts | henrys: No idea. | 16:11.19 |
henrys | I thought you were going to be there for the whole show | 16:11.48 |
Robin_Watts | If they don't have devices yet, then I guess they'll need to be buying them. | 16:11.49 |
| Not me. Michael is doing this one. | 16:11.57 |
| I'm arriving on the afternoon of the 6th along with kens and chrisl. | 16:12.29 |
henrys | well so we'll talk to michael next meeting. | 16:12.30 |
Robin_Watts | michael has a kindle fire. | 16:12.43 |
| dunno if he can load our stuff onto that. | 16:13.07 |
henrys | I could mail him my nexus but they are relatively cheap. | 16:13.22 |
| we'll ask him to get one. | 16:13.46 |
| or something similar | 16:13.56 |
| paulgardiner:are we still on for beta in February? | 16:14.28 |
| it seems like alerts were more of an investment than you thought. | 16:14.59 |
paulgardiner | I think so. The list is in any case a bit arbitrary. We'll know better what to focus on when we get some more complaints. | 16:15.49 |
henrys | tor8:so if you are going to do something we need it soon. It probably is a good idea to have multiple apk releases available there is a glitch at the show. | 16:16.08 |
| s/there/in case there/ | 16:16.28 |
paulgardiner | Item 6 may well not happen by Feb. That's a load of work, and our update may still not be fast enough for that. | 16:16.30 |
| Also Item 3 covers only the parts of the DOM that our test files use. If a customer wanted a complete DOM it would be a huge task. | 16:18.35 |
henrys | okay, I didn't really have anything else for the meeting. Just from looking I think saving and submission are more important than other items on the list so maybe that should be next, but whatever you think. | 16:18.49 |
Robin_Watts | So we have 1) Submission, 3) Some DOM functions, 5) Better discrimination for saving, 6) Inline editing. | 16:18.50 |
paulgardiner | henrys: ah good. I was thinking that too. | 16:19.06 |
Robin_Watts | I'd like to see saving moved up the priority list. | 16:19.07 |
| Because if it shows problems, then it's something that tor8/I can fix. | 16:19.24 |
paulgardiner | Robin_Watts: I was thinking saving next | 16:19.35 |
Robin_Watts | I'd hate to hit problems in that 2 weeks before we have to ship :) | 16:19.43 |
paulgardiner | Otherwise we annoy people by having them fill out forms and throw the info away | 16:19.57 |
| We already have saving on the windows working so it shouldn't be hard. | 16:22.36 |
Robin_Watts | paulgardiner: Right, but that's not be hugely tested. | 16:24.32 |
henrys | okay so we'll talk to michael next meeting about the device, anything else we need to talk about? | 16:24.44 |
paulgardiner | henrys: I'm done, I think. | 16:25.57 |
henrys | oh I should say I won't be here next Tuesday, maybe we should request michael attend next tuesday to give him a run down before the show. | 16:29.25 |
Robin_Watts | henrys: Would it make more sense to move to next wednesday or something then? | 16:29.54 |
| (subject to tor8's availability obviously) | 16:30.15 |
henrys | I'm fine with wednesday. | 16:30.51 |
| tor8, paulgardiner? | 16:32.10 |
paulgardiner | Fine with me | 16:32.43 |
henrys | we can convince tor8 later. | 16:36.35 |
ray_laptop | I saw about mupdf performance. I don't know what the license situation wr.t. mupdf is, but we should find out in case we can hook in the luratech decoders for JBIG2 and JPX. | 16:47.06 |
Robin_Watts | ray_laptop: I've not hit a case where the decompression speed has been an issue. | 16:47.47 |
| It's the color conversion/scaling in all the cases so far. | 16:48.09 |
ray_laptop | it was full page scans that I've seen with JBIG2 that might be helped. Also I have some files from cust 532 that had big JPX images. I'll dig a couple out and forward them. | 16:49.49 |
| BTW, this tablet looks pretty nice. I'll probably get one. http://www.jr.com/ematic/pe/EI_EGLIDEXLPRO2/ | 16:51.45 |
Robin_Watts | ray_laptop: I wouldn't touch that with a bargepole :( | 16:53.03 |
ray_laptop | Robin_Watts: why not ? | 16:53.11 |
chrisl | A bargepole wouldn't work on a capacitance touch screen? | 16:53.35 |
Robin_Watts | The key thing with a tablet is down to whether you can get OS fixes/upgrades. | 16:53.59 |
| Noname tabs are a bad bet in that regard. | 16:54.12 |
henrys | FWIS my nexus upgrade was painless | 16:54.44 |
mvrhel_laptop | henrys: did you want be to get a nexus for the show. Just reading the logs | 16:55.03 |
Robin_Watts | henrys: Right. That's cos Nexus == Google. | 16:55.06 |
henrys | Robin_Watts:of course | 16:55.20 |
| mvrhel_laptop:yes that was the first meeting topic. | 16:55.33 |
Robin_Watts | Similarly the Transformer has upgraded to ICS and Jellybean flawlessly. | 16:55.34 |
henrys | you don't have to get a nexus. just something to show mupdf | 16:56.04 |
mvrhel_laptop | what size would be good | 16:56.39 |
henrys | I could also mail you mine, but they're pretty cheap I don't know if it is worth that. | 16:56.44 |
| I think a size larger than what I have would be better mine is kindle fire sized | 16:57.38 |
mvrhel_laptop | ok they come in 10, 7 and 4 inch versions | 16:58.08 |
| http://www.google.com/nexus/ | 16:58.23 |
henrys | so I think 10 is best for a show - but maybe others feel differently. | 16:58.35 |
mvrhel_laptop | I would think for a demo the 10 would be good | 16:58.38 |
| yes | 16:58.40 |
Robin_Watts | ray_laptop: No mention of what actual res that device is. | 16:58.49 |
| If henrys has a 7, getting a 10 would be good just for variety. | 16:59.11 |
henrys | now if I could just install linux on this thing I could get rid of all these computers ;-) | 17:00.13 |
Robin_Watts | henrys: There is an Ubuntu on Android thing somewhere... | 17:00.32 |
mvrhel_laptop | 16GB or 32GB? | 17:00.42 |
henrys | Robin_Watts:yes I saw that. | 17:01.00 |
Robin_Watts | Unless there is a really good reason why not, I'd get the 32Gig one. | 17:01.14 |
mvrhel_laptop | right | 17:01.18 |
| well any demo device we want the best | 17:01.32 |
| crap sold out of the 32GB device | 17:02.00 |
Robin_Watts | Where are you looking? | 17:02.10 |
mvrhel_laptop | https://play.google.com/store/devices/details?id=nexus_10_32gb | 17:02.20 |
henrys | mvrhel_laptop:we also wanted to invite you to a meeting next wednesday at 8:00 Pacific and we'll talk more about the show, with paulgardiner et al. | 17:02.39 |
mvrhel_laptop | that would be great | 17:02.48 |
| I have been playing around with mupdf | 17:02.53 |
| but in the context of a windows app | 17:03.04 |
| http://androidcommunity.com/nexus-10-32gb-sold-out-on-google-play-store-20121113/ | 17:04.00 |
| this is interseting | 17:04.03 |
Robin_Watts | eek. No microSD slot. | 17:04.45 |
| That's why you want the 32Gig one. | 17:04.52 |
mvrhel_laptop | yes. that stinks | 17:04.53 |
henrys | I just went to my local office store to get mine. | 17:04.54 |
mvrhel_laptop | ah ok | 17:04.59 |
Robin_Watts | Hmm. The 10s only launched last tuesday. | 17:05.22 |
henrys | network problems kens? | 17:05.37 |
kens | Seems to be | 17:05.42 |
| I have a new router on order to see if that fixes it | 17:05.50 |
| If not I will complain to my ISP | 17:06.02 |
Robin_Watts | kens: rainy day there ? | 17:06.16 |
kens | Robin_Watts : not especially | 17:06.23 |
tor8 | henrys: wednesday is fine | 17:06.25 |
henrys | I'll be out next tuesday, so let's just make it optional next week. | 17:06.55 |
ray_laptop | mvrhel_laptop: did you see the OpenICC thread discussion on "Wrong color output of Xerox printer - Color Management problem?" | 17:06.56 |
mvrhel_laptop | ray_laptop: yes. I have half of a reply typed | 17:07.10 |
henrys | tor8:okay good | 17:07.21 |
mvrhel_laptop | then had to help get kids out the door and had not finished what I started | 17:07.29 |
kens | mvrhel_laptop : I could use a little help with a transparency problem if you haev some time later | 17:07.33 |
mvrhel_laptop | kens: ok | 17:07.44 |
kens | after meeting :_) | 17:07.50 |
henrys | mvrhel_laptop:hopefully you'll have the new device by next wednesday and we'll field any questions you might have. | 17:08.15 |
ray_laptop | OK. I was concerned because Gerard was saying "seem to be caused by the fact that the new ICC-color-managed gs 9.x does obviously no longer support CRDs (CRDs are however a well-defined PostScript functionality - i.e. the new gs is no longer fully PostScript L2/L3 compliant wrt. this functionaily)." | 17:08.16 |
mvrhel_laptop | ray_laptop: yes. that is what this bug is about http://bugs.ghostscript.com/show_bug.cgi?id=691749 | 17:09.57 |
henrys | chrisl:I am seeing an error return from fapi each job when running with -Z#, I've tried to keep pcl -Z# clean or at least minimal. | 17:10.09 |
mvrhel_laptop | and yes ray_laptop, he is essentially correct | 17:10.17 |
henrys | tor8:anything new with the viewer? | 17:10.36 |
chrisl | henrys: I'll look into that.... | 17:10.45 |
ray_laptop | mvrhel_laptop: I thought you did handle CRD's -- or was that something that never got done? | 17:10.56 |
mvrhel_laptop | ray_laptop: that is something that did not get done | 17:11.10 |
| hence the open bug | 17:11.22 |
ray_laptop | mvrhel_laptop: oops. | 17:11.24 |
mvrhel_laptop | it should be moved out of the enhancement level probably. but the fact that it has been out since 9.0 and never an issue shows how few people do PS color management these days | 17:12.19 |
kens | very few people ever got CRDs to work well | 17:12.37 |
| Too complicated for the average user | 17:12.49 |
ray_laptop | mvrhel_laptop: true. | 17:12.53 |
| kens: CRD's used to be quite popular and there were several 'color management' companies that generated them. Most of those map easily to ICC profiles. | 17:14.06 |
kens | ray_laptop : yes, but few people ver used them succesfully. BEST were an exceptionand there were others | 17:14.29 |
oy | even lcms supports them | 17:14.42 |
kens | But on the whole, it was an idea that didn't really work | 17:14.46 |
mvrhel_laptop | oy: no it does not | 17:14.51 |
| not directly | 17:15.01 |
oy | it generates them as far as I know | 17:15.05 |
ray_laptop | how could lcms support them -- it requires (in general) a PS parser | 17:15.11 |
mvrhel_laptop | it can generate a CRD from an ICC profile | 17:15.26 |
| that is all | 17:15.28 |
| we need to do the opposite | 17:15.32 |
ray_laptop | oy: yes, icc2ps is a MUCH easier step | 17:15.36 |
oy | agreed to both of you :-) | 17:15.48 |
henrys | I can't imagine a crd is going to require that much of a postscript interpreter. | 17:16.51 |
ray_laptop | but for most CRD's (excluding screwy ones that may be in the QL CET) they should convert easily. Usually just a few matrices and a RenderTable | 17:16.54 |
mvrhel_laptop | yes. I don't for see any issues. | 17:17.09 |
| I already have the code I wrote for doing the CIEDEF(G) conversions to ICC profiles | 17:17.29 |
ray_laptop | henrys: the one that gets screwy is the TransformPQR procs (iirc) | 17:17.31 |
mvrhel_laptop | the code for the CRD is not going to be that much different | 17:17.48 |
| I will try to get this in before the next release | 17:18.06 |
alexcher | Why CRD problems were not detected by regression testing? | 17:19.04 |
mvrhel_laptop | I also need to ping customer #330 to see how he is doing with the stuff I did | 17:19.06 |
henrys | alexcher:anything for the meeting? Your machine is still down. | 17:19.07 |
alexcher | henrys: I've successfully updated the repositories and would like to put alex_x6 back to the cluster. | 17:20.14 |
Robin_Watts | I can do that now... | 17:20.45 |
| marcosw: speak now if I shouldn't... | 17:20.57 |
chrisl | henrys: the -Z# output should be a bit quieter now..... | 17:21.20 |
Robin_Watts | he's not here, so done. | 17:21.21 |
alexcher | GS reportedly fails on Sparc. Should we get a Sparc box on ebay to test? | 17:21.32 |
henrys | chrisl:thanks | 17:21.32 |
chrisl | alexcher: do you want me to take a look at that Solaris/SPARC issue? | 17:21.44 |
alexcher | yes | 17:21.50 |
chrisl | Okay, I'll assign it to me | 17:22.01 |
alexcher | chrisl: we need to run regression testing on minority platforns to prevent such problems in the future. | 17:23.22 |
chrisl | alexcher: I know, the problem I have is that my SPARC machines are too noisy to leave on overnight, and too slow to get much testing done in one day :-( | 17:24.09 |
henrys | I didn't have anything else for the meeting. I'll try to get the in person agenda out early so everyone can have time to make contributions. | 17:24.34 |
kens | mvrhel_laptop : can you spare me a few minutes at the moment ? | 17:25.08 |
chrisl | henrys: are you aware of any HP PCL printers that *don't* do Truetype hinting? | 17:25.14 |
ray_laptop | have a good thanksgiving, everyone (at least those that observe it). | 17:25.18 |
alexcher | chrisl: Is it worth to buy an old 500MHz Sparc on ebay for about $100? | 17:25.24 |
kens | Oh yeah, enjoy all of you | 17:25.30 |
mvrhel_laptop | kens: yes | 17:25.31 |
kens | mvh can you look at bug #693456 please, theres a simplified file | 17:25.51 |
| mvrhel_laptop: * | 17:26.00 |
| THe problem is when rendering transparency for lower levels of PDF output, ie making an image | 17:26.17 |
ray_laptop | I have to go now. I probaby won't be looking at IRC, so someone call me if I am needed. (working on cust 532 issues). | 17:26.27 |
kens | I noticed a few days back that this is not working for images with SMasks | 17:26.33 |
chrisl | alexcher: it's up to you. My intention is to revise my own testing scripts so that they record where they get to in the test files, and then can pick up where they left off. So I can do testing in bits. | 17:26.47 |
Robin_Watts | alexcher: if it's in working condition (with OS etc), and that includes shipping it sounds worth it to me. | 17:26.51 |
marcosw | sorry I'm late. Traffic over the hill was bad due to rain and fog. | 17:27.20 |
chrisl | alexcher: but we'll never get a cheap SPARC fast enough to put on the cluster :-( | 17:27.33 |
henrys | chrisl:I think we should stick with hinting yes. | 17:27.45 |
mvrhel_laptop | kens: does it have these issues when just going out to a raster device too? | 17:27.49 |
kens | mvrhel_laptop : no, raster rendering is fine | 17:28.02 |
Robin_Watts | we can get a cheap sparc online all the time for anyone of us to ssh into though. | 17:28.09 |
| and we can run smoke tests on it. | 17:28.14 |
kens | checked with tiff32nc and display device | 17:28.15 |
mvrhel_laptop | kens: just when rendering out to your pdfwrite raster? | 17:28.27 |
chrisl | henrys: No, what I was wondering was should I add a flag to allow the interpreter to disable the hinting? | 17:28.27 |
kens | mvrhel_laptop : THe problem only seems to arise when rendering a bitmap for pdfwrite | 17:28.45 |
| THe output looks like the SMask is being ignored totally | 17:28.55 |
mvrhel_laptop | kens: ok. If you want me to poke at it today I can | 17:29.22 |
henrys | chrisl:I could have sworn I added code to make that resolution dependent an now I can't find it. | 17:29.28 |
kens | mvrhel_laptop : If you could take a quick look it would be very helpful, I'm at a loss with it. | 17:29.49 |
mvrhel_laptop | kens: ok. there have been quite a few changes with respect to the softmask handling in the last year. It is possible something was missed for pdfwrite | 17:30.24 |
henrys | chrisl:let me search for what I did before. | 17:30.32 |
chrisl | henrys: the reason I ask is that we have a couple of files that look worse with full hinting - but I believe the "worse" output is actually correct | 17:30.41 |
henrys | chrisl:can you tell me which files? | 17:31.26 |
kens | mvrhel_laptop : in gdevpdft.c, gdev_pdf_create_compositor I see that we call psdf_creater_compositor which ends up calling gx_default_create_compositor. Later I see an image arrive at pdfwrite (which I expect), and that's effectively all I see, the image seems to be incorrect and I have no idea why | 17:31.34 |
mvrhel_laptop | kens: is banding used in this case? | 17:32.00 |
| just curious | 17:32.06 |
kens | mvrhel_laptop : Its entirely possible that there is something screwy with the special case handling by the PDF itnerpreter which causes it not to do something it should when the device is pdfwrite. | 17:32.07 |
| mvrhel_laptop : pdfwrit (currenlty) never uses the clist | 17:32.22 |
mvrhel_laptop | ok good | 17:32.25 |
chrisl | henrys: bug690105.pcl (the "P" glyph), and formulas.hp8100pcl5e.pcl (the beta glyph) | 17:32.35 |
henrys | chrisl:so pcl does gs_setgridfittt() off >= 300 | 17:32.38 |
mvrhel_laptop | kens: I don't understand how the image could be right for the raster device and wrong for what you need but I will see if I can figure out | 17:33.25 |
kens | mvrhel_laptop : Its got me baffled too.... | 17:33.36 |
henrys | what pcl does under 300 dpi really doesn't make a great deal of difference real world wise. Is that what you are looking at? | 17:33.52 |
chrisl | henrys: okay, so fapi ignores that - I don't think we want to keep that resolution dependence, personally. Too many fonts actually use the hinting to really change the outlines | 17:33.56 |
marcosw | henrys: I need assistance with http://bugs.ghostscript.com/show_bug.cgi?id=693387 Basically I'm having trouble dealing with reversing the memory increase between 8.54 and 8.64. I've identified the commits that caused the increase. | 17:33.57 |
chrisl | henrys: no, this is 300dpi | 17:34.13 |
kens | mvrhel_laptop: My only thought is that 'maybe' the PDF itnerpreter is doing some special handling because the device claims to be able to handle transparency natively, but I can't even find such a case | 17:34.31 |
henrys | chrisl:I do recall that was for performance and it "mattered" to a customer. | 17:34.59 |
| norbert if I remember right. Is it possible there has been some improvement with the new ft? | 17:35.45 |
chrisl | henrys: compared with FT1? Yes, the hinting is considerably quicker in FT2 | 17:36.16 |
Robin_Watts | marcosw: Looking at those commits some of them are just including new devices. | 17:36.29 |
kens | I need to dash off, certificate evening at Melanie's old school, does nayone need any thing from me ? I won't be back until after 9pm tonight | 17:36.46 |
marcosw | you mean http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=ab3e109142b9cec86c993b4bfbabfe32ad3bc3ef | 17:37.19 |
Robin_Watts | which makes me think they shouldn't have any such effect. Which makes me further think that it must be down to random placement of memory affecting gc in some way. | 17:37.20 |
henrys | okay so I'll go ahead and remove that code. | 17:37.32 |
kens | Oh, mvrhel_laptop to reduce the processing time, add -r72 to the command line, it really doesn't need to be rendered at 720 dpi | 17:37.42 |
henrys | bye kens | 17:37.48 |
Robin_Watts | marcosw: and http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=62b66eb67bda44d5becebc709b565d45bf70c585 | 17:37.50 |
mvrhel_laptop | kens: ok thanks | 17:37.53 |
kens | Goodnight all | 17:37.55 |
henrys | marcosw:should we do bugs it has been a while but I haven't seen anything much to process. | 17:38.19 |
| ? | 17:38.21 |
marcosw | Robin_Watts: the previous commit didn't compile: http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=34f0b7f064072fabdbe47ab1a15d4f1341a0417d | 17:38.50 |
chrisl | henrys: if it really is important to norbert, I'm willing to throw in a command line option, and FAPI support for disabling hinting, but I think it's a bad idea, to be honest. | 17:39.01 |
Robin_Watts | marcosw: Ah. | 17:39.28 |
marcosw | henrys: I'm presenting in lab meeting today so under pressure to work my school stuff. can we meet tomorrow? | 17:39.33 |
henrys | marcosw:yeah that's fine or you can process it when you have time and I'll review it if that's easier. | 17:40.14 |
marcosw | Pretty much every one of these issues is like that (you can tell from the first column, which is the svn revision number, another reason that I suggest we drop git: it's really hard to tell how many commits there were between to md5 hashes, you can ask rayjj for another reason to abandon git) :-) | 17:41.45 |
| btw, I've started checking for memory usage increases in the nightlly/weekly regression tests. | 17:43.22 |
Robin_Watts | marcosw: git name-rev SHA | 17:46.15 |
Gigs- | can mutool list all the bounding boxes, not just the mediabox? | 17:46.35 |
| like cut trim etc | 17:46.58 |
marcosw | Robin_Watts: I don't think so, e.g. "git name-rev 34f0b7f064072fabdbe47ab1a15d4f1341a0417d ab3e109142b9cec86c993b4bfbabfe32ad3bc3ef" | 17:47.54 |
henrys | chrisl:the log message for the change didn't reference anything specific so if there are no regression slowdowns we should be okay. | 17:47.59 |
chrisl | henrys: okay, thanks. | 17:48.39 |
alexcher | Gigs-: Submit an enhancement request. Otherwise it will be forgotten. | 17:49.44 |
Gigs- | ok | 17:49.59 |
chrisl | henrys: the other thing, if norbert really wants to use UFST for truetype downloaded fonts, we have *much* less control over how it handles hinting than we do with freetype, anyway. | 17:50.31 |
Robin_Watts | marcosw: $ git name-rev 34f0b7f | 17:50.37 |
| 34f0b7f tags/ghostscript-8.64~399 | 17:50.37 |
| So that was 399 commits before 8.64 | 17:50.48 |
| $ git name-rev ab3e109 | 17:51.10 |
| ab3e109 tags/ghostscript-8.60~129 | 17:51.12 |
marcosw | How does that tell you how many commits there were between those two md5 hashes? | 17:53.23 |
| with svn it's a bit more obvious: 8954-8027 | 17:53.42 |
henrys | that reminds me it would be nice if the regression message listed one line summaries next to the sha of each instead of a range of sha's but no big deal. | 17:54.17 |
marcosw | which works out to 927 (or maybe 926, depending on the definition of between) | 17:54.40 |
henrys | the performance regression that is. | 17:55.15 |
marcosw | henrys: I think we do that for the commit regression emails, I could do it for the nightly/weekly/performance regressions as well. | 17:55.17 |
Robin_Watts | marcosw: I bet it can be done, but I'm not gittish enough. | 17:56.23 |
chrisl | Of course it can be done, otherwise bisect wouldn't be able to give you the summaries about commits to test, and steps remaining..... | 17:57.34 |
Robin_Watts | git rev-list SHA1...SHA2 | wc -l | 17:58.04 |
marcosw | Robin_Watts: I think that's the point I'm trying to make, git is the best version control system ever, if you know how to use it. But it's clear that no one knows how to use it (except perhaps Linus Torvalds). | 17:58.31 |
Robin_Watts | marcosw: Except with SVN that's not true. | 17:58.34 |
| If I make commit 1 on the trunk, then make 2,3,4,5,6 on a branch then commit again to the trunk I get 7. | 17:59.18 |
| 7-1 = 6 is NOT the number of intermediate commits. | 17:59.30 |
| Hence git gets it right, and svn doesn't. | 17:59.53 |
| mvrhel_laptop, henrys: Amazon are selling the Samsung Galaxy Tab 2 for $299 this week. | 18:07.29 |
| 16GB + a microSD slot. ICS. | 18:08.03 |
| having a Samsung tab in the company might not be a bad idea given scotts mail the other day :) | 18:08.32 |
tor8 | Robin_Watts: I have a samsung 10" tab. too bad I hate it :( | 18:09.45 |
Robin_Watts | tor8: how so? | 18:09.54 |
tor8 | it's got something on the order of 24 hours standby before the battery runs dry | 18:10.09 |
| and a stupid custom connector that I always have to dig out the bloody charger for | 18:10.28 |
mvrhel_laptop | the connector thing is really lame | 18:10.53 |
tor8 | so everytime I want to use it, I find it's dead because the battery needs charging, then I have to go find a spare socket for the damn custom connector | 18:11.09 |
| kind of beats the purpose. at least if it had a micro usb I'd just plug it into my regular phone charger that's always lying around | 18:11.43 |
| the ipad also has an annoying custom connector, but at least it has battery enough to last a few weeks powered off | 18:12.26 |
Robin_Watts | The transformer has a custom connector (which it needs because of the keyboard doc etc). | 18:12.48 |
| but it's got good standby. | 18:12.55 |
mvrhel_laptop | So the apparently the 16GB nexus 10s are expected to sell out too | 18:14.56 |
chrisl | marcosw: as a slightly brute force approach: "git log <sha 1>...<sha 2> --pretty=oneline | wc -l" gives you the number of commits between revisions | 18:15.28 |
mvrhel_laptop | Robin_Watts: henrys: should we go with the nexus 10 16GB or should we go with a different brand, or the nexus 7 32GB? | 18:16.48 |
Robin_Watts | Given the lack of expansion cards, I'd be tempted to avoid the 16gb one. | 18:17.19 |
marcosw | chrisl and Robin_Watts: thanks for your suggestions, both seem to produce the correct answers | 18:17.22 |
mvrhel_laptop | Robin_Watts: right | 18:17.37 |
Robin_Watts | I knew of one computer company that refused to ever buy 2 PCs of the same model. | 18:18.27 |
| On the grounds that that provided a better test for their software (this was back in the day with issues with graphics cards etc). | 18:18.53 |
| so there is an argument to say we should buy a completely new brand of tab. | 18:19.15 |
| but then do we want to be remote debugging it the week before the show? :) | 18:19.31 |
henrys | the show is good enough reason to get the nexus. | 18:19.56 |
mvrhel_laptop | Robin_Watts: right. I would steer toward having something that is as bullet proof (and idiot proof for me) | 18:19.58 |
Robin_Watts | Then the nexus 7 32Gig sounds like a good bet. | 18:20.19 |
mvrhel_laptop | ok. I will get that ordered now. they may actually have that at the store here | 18:20.37 |
| hey I can get 15% off at staples today | 18:27.31 |
| oh but does not include tablets | 18:29.35 |
| oh. well they do have it in stock so I will pick up later today | 18:29.56 |
Robin_Watts | mvrhel_laptop: Excellent. | 18:30.28 |
mvrhel_laptop | I need to get one of these for my office at home http://www.thinkgeek.com/product/e9b8/ | 18:41.09 |
| I have been watching the old star treks with my son who asked me a few weeks back what star trek was | 18:42.31 |
Robin_Watts | Excellent. If I had a clamshell phone, I'd want one that made the tweedlesquirge noise when you flipped it open. | 18:44.59 |
mvrhel_laptop | yes | 18:45.05 |
| the log message below this is funny | 18:45.24 |
| http://www.thinkgeek.com/product/8506/ | 18:45.26 |
| if you remember the salt sucking vampire episode | 18:45.33 |
| Robin_Watts: you could get one of these | 18:46.00 |
| http://www.thinkgeek.com/product/9c69/ | 18:46.02 |
| suggestsions to make this work with bluetooth | 18:46.48 |
| which would be way cool | 18:46.54 |
Robin_Watts | mvrhel_laptop: http://www.youtube.com/watch?v=D_0xaOQWg6I&list=PLB42D4A12C6D49C07&index=3&feature=plpp_video | 18:53.21 |
mvrhel_laptop | funny | 18:56.00 |
| bbiaw | 21:18.30 |
Robin_Watts | tor8: ping | 21:29.13 |
tor8 | Robin_Watts: pong | 21:29.38 |
Robin_Watts | The current file I'm looking at has a big fax image on one page. | 21:29.41 |
| but it's not 1 image, it's 160 of them. | 21:29.48 |
| they are full width images, but only a few pixels high for each one. | 21:30.17 |
| So when we scale each one in turn, we recalculate the weights for each image. | 21:30.38 |
| When actually we might as well just have used the weights we calculated before. | 21:30.51 |
| I'm tempted to add a 1 place cache to the system for such weights. | 21:31.09 |
| You have any strong opinions on that? | 21:31.42 |
| 2 ARM code commits on my repo awaiting review (or at least a nod) | 21:38.14 |
tor8 | Robin_Watts: where do you stick the cache? | 21:48.15 |
| 1-place caches are useful and I like them in general | 21:48.34 |
| Robin_Watts: _ARM caps? | 21:51.36 |
Robin_Watts | tor8: I was thinking either in the context or in the draw device. | 22:03.59 |
| I think I prefer the draw device because it's better for multithreading. | 22:04.14 |
| but then that falls down if we want to try to use the cache over draw device invocations. | 22:05.12 |
| dinner. I'll ponder it and try it tomorrow. | 22:06.18 |
mvrhel_laptop | henrys: got the nexus 10 32GB ordered. | 23:37.47 |
| google suddenly made more available | 23:37.59 |
henrys | ah outstanding | 23:40.06 |
| next week we'll make sure you have form filling pdfs and interesting examples ready to go. | 23:41.05 |
mvrhel_laptop | henrys: great | 23:55.07 |
| Forward 1 day (to 2012/11/21)>>> | |