| <<<Back 1 day (to 2012/09/11) | 2012/09/12 |
mvrhel_laptop | goodnight all | 07:05.07 |
Robin_Watts | Mornint tor8. Good journey home? | 10:35.40 |
tor8 | Robin_Watts: just woke up... | 10:35.58 |
| no delays, bomb threats, missing fuel lids, cancellations, strikes or any of the usual sort so all's good I reckon :) | 10:36.48 |
| most dreadful plane food in a while though... | 10:37.12 |
odinho | I'm not able to RGBify this one: http://static.universitas.no/docs/pdf/universitas_2012-23.pdf | 10:38.17 |
| Doing >> gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/screen -dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf universitas_2012-23.pdf << | 10:38.42 |
Robin_Watts | does the VMware Player update shuffle. How many reboots today? Place your bets now... | 11:18.39 |
sebras | tor8: welcome back. I have a lipovitan-d somewhere... | 11:24.00 |
tor8 | sebras: I fear I shall need more than one these coming days... | 11:28.21 |
Robin_Watts | Morning kens. | 11:52.40 |
| How you feeling? | 11:52.44 |
kens | Hi Robin_Watts | 11:52.46 |
| Feeling OK, stayed up till 10:30 last night | 11:52.57 |
Robin_Watts | is considering updating to Ubuntu 12.04LTS for my local VMware today. | 11:54.05 |
kens | Shouldn't be a problm | 11:54.31 |
Robin_Watts | except it'll change to unity from gnome. | 11:54.49 |
kens | You cna always install gome over it I think ? | 11:55.03 |
Robin_Watts | yeah, but gnome 3 rather than 2. | 11:55.20 |
| I suspect that actually it's worth taking the pain and moving to Unity though. | 11:55.39 |
kens | Next you'll be sayign Windows 8 is OK really.... | 11:55.56 |
Robin_Watts | no :) | 11:56.10 |
| I think I'm going to copy the VM and then upgrade it :) | 11:56.25 |
kens | Someone on the Register was saying that (more or less) today | 11:56.26 |
Robin_Watts | "The data is the UI" thing? | 11:56.52 |
kens | Yes, that's it | 11:57.00 |
| I remain unconvinced | 11:57.09 |
| But I have to admit I haven't actually looked at Windows 8 | 11:57.25 |
Robin_Watts | The problem I have with Windows 8 is not the addition of the new UI. | 11:57.54 |
| I have no objection to that. | 11:57.59 |
| The problem is that they've removed the start menu so that the desktop sandbox has to keep kicking you back to the new UI to launch anything. | 11:58.27 |
| The two metaphors should be kept separate, IMO. | 11:58.43 |
kens | According to that article, they haven't removed the start button | 11:58.51 |
| You just have to mouse down to the bottom left corner | 11:59.06 |
| Which is no problem for me as my taskbar is set ot hide | 11:59.16 |
| But possibly I misread the article | 11:59.26 |
Robin_Watts | Well they've hidden it bloody well, and are actively seeking to stop people putting it back. | 11:59.32 |
kens | Not having used it, I don't feel qulaified to comment really. | 12:00.04 |
Robin_Watts | TIFKAM effectively becomes the start menu, which is a huge fullscreen wrench. | 12:00.05 |
kens | Yeah, some of the aspects he was toutitn I thought "Oooh, just liek Windows 3.1...." | 12:01.22 |
Robin_Watts | Morning chrisl | 12:05.35 |
chrisl | Afternoon, Robin_Watts..... | 12:05.45 |
kens | <pedant> actually its afternoon </pedant> | 12:05.57 |
Robin_Watts | My soul is still midatlantic, so, it's morning there. | 12:06.36 |
chrisl | I assumed we wouldn't have an IRC meeting yesterday, did I miss anything? | 12:11.01 |
kens | Nope, no meeting | 12:11.10 |
chrisl | Good | 12:11.38 |
kens | Hmm, compile fail for no obvious reason, doesn't seem to be my code | 12:33.43 |
| chrisl: | 12:34.26 |
| creating links to configuration sources | 12:34.26 |
| running autoconf | 12:34.26 |
| Modification of a read-only value attempted at /usr/share/perl/5.10/Getopt/Long.pm line 865. | 12:34.26 |
chrisl | Huh? autoconf and co don't use perl at all | 12:35.22 |
kens | its a commit mail, so feel free to look at the wjhole mail | 12:35.41 |
chrisl | Beats me - let me pull the change and try it here | 12:37.03 |
kens | OK | 12:37.10 |
chrisl | Works fine here - has someone been cluster tamperin'? | 12:39.51 |
kens | It might be a glitch | 12:40.15 |
chrisl | Can we re-queue it? | 12:40.34 |
kens | I have no idea | 12:40.48 |
chrisl | paging Robin_Watts....... | 12:41.03 |
Robin_Watts | back, sorry, had gone to lunch | 13:12.57 |
kens | Robin_Watts : can I requeue a cluster test from a commit ? | 13:13.11 |
Robin_Watts | No, but I can. | 13:13.20 |
kens | OK, would yuo mind please? | 13:13.32 |
Robin_Watts | hopefully it should be running now. | 13:15.26 |
kens | 'something' is :-) | 13:15.46 |
| Assuming its the current head it should be the right one | 13:15.59 |
Robin_Watts | It's a010600d3d99b16251fbfa5fede57d4583978632 | 13:21.38 |
kens | That's the one | 13:21.54 |
Robin_Watts | ok, I've hit the magic upgrade button in Ubuntu. Let's see what happens. | 13:22.11 |
kens | No problems building this time, it must have been a glitch on kilometers | 13:22.18 |
justme_ | hi, guys! I'm trying to modify your muPDF android sample code so that, instead of a single PDF file, it opens several one-page PDF files. I have moved the core object from the adapter to MuPDFPageView objects, and I create the core everytime I instantiate a page from the adapter. The problem is that I get some native errors, with the DEBUG tag and a crash, preceded by this: 09-12 18:08:10.220: E/libmupdf(5442): warning: assert: o | 15:09.32 |
Robin_Watts | justme_: Right. | 15:09.58 |
| The C core of mupdf can support multiple files being opened at once. | 15:10.35 |
| The android wrapper stuff... quite possibly not. | 15:10.46 |
| I'm guessing that you're hitting an assert statement in the C where we're expecting o to be non NULL (o is probably a pdf_obj *) | 15:11.36 |
justme_ | so the "overwrite hash slot" is referring to a Java problem, and not a native one, and I must find the error there | 15:13.16 |
Robin_Watts | What "overwrite hash slot" error ? | 15:14.04 |
justme_ | warning: assert: overwrite hash slot | warning: ... repeated 3 times ... | warning: ... repeated 3 times ... | warning: freetype load glyph (gid 42): invalid argument | warning: freetype render glyph (gid 51): cannot open resource | warning: assert: overwrite hash slot | 15:14.48 |
Robin_Watts | Ah, your report got truncated. | 15:15.05 |
justme_ | these are the "libmupdf" errors that I get before the crash | 15:15.05 |
| yup | 15:15.10 |
Robin_Watts | So ignore what I said before about a pdf_obj *o being unexpectedly NULL. | 15:15.40 |
| That's almost certainly an error being given by the C code. | 15:16.32 |
| probably because the C code is being called in an unexpected way. | 15:16.42 |
| The java code is not written to cope with multiple files open at once. | 15:16.58 |
| You'd need to look at it, and figure out what it's doing wrong. | 15:17.15 |
justme_ | I am thinking that the error is given by the fact that the asyc doesn't return in time and it causes erratic behavior | 15:17.30 |
| I realy like the way your sample is handling the rendering, making it so smooth, and instead of starting an app from scratch, I figured I'd start modifying the sample, until I get what I need | 15:19.46 |
Robin_Watts | justme_: If it was me, I'd be tempted to actually hold several pdf files open at once rather than closing and reopening them as you move between pages. | 15:20.32 |
tor8 | Robin_Watts: justme_: another option is to add a new fz_document type at the C level that "concatenates" several other documents | 15:21.18 |
justme_ | that's what I was going to ask next, if it's possible to concatenate several documents | 15:21.53 |
tor8 | the java wrapper to mupdf could also be rewritten to more closely mimic the fz_document api | 15:21.58 |
| (an API which didn't exist at the time the android java code was written) | 15:22.21 |
| justme_: not out of the box, but writing a multiplexing fz_document wrapper should be a fairly trivial exercise | 15:23.19 |
ray_laptop | kens: your 'infinite loop' fix -- are you sure ? if "Length" is 1, it will write 2 characters -- is that correct ? | 15:25.55 |
| kens: (also it isn't clear how an infinite loop is averted -- the original just writes "Length" characters before exiting the loop. The new version writes Length+1 chars | 15:27.21 |
kens | ray_laptop : No, I'm just fixing it now, commit in 2 secs | 15:27.37 |
ray_laptop | kens: OK. Thanks | 15:27.49 |
kens | ray_laptop : because it can start negative | 15:28.07 |
| and then tries to countdown to 0 | 15:28.36 |
ray_laptop | kens: Ahh -- that makes sense. Well it wasn't infinite -- just a LOT of chars written ;-) | 15:28.42 |
kens | close enough ;-) | 15:28.53 |
ray_laptop | kens: particularly on a linux 64-bit build :-) | 15:29.30 |
kens | It was close enough to infinite on my PC.... | 15:29.52 |
justme_ | about the concatenation thing, I'm guessing that documents with over 100 pages or so will cause an outofmemory exception, being loaded entirely in the RAM, am I right ? | 15:29.53 |
ray_laptop | kens: just saw the commit message. Thanks. Good thing the cluster doesn't test linearization | 15:30.32 |
kens | ray_laptop : yes, I haven't tested it properly yet, need the timeout change from marcosw | 15:30.58 |
ray_laptop | just got an ad for a product... real marketing hype: $400 value Reg. $299.99 ... Now $199.99 ... Save 50% (guess they didn't go to school in Arkansas) | 15:33.32 |
| kens: not sure that the timeout change would have helped before your fix ;-) | 15:36.56 |
| mvrhel: I have a question about my pdf14_clist_optimization regarding the changing of color_spaces. Can you give me a call ? | 15:58.42 |
henrys | ray_laptop:Trying to decide about buying the PDF ATS. What version exactly does your customer have? | 15:59.46 |
Robin_Watts | henrys: Buying the PDF ATS would not help with rays customer, AIUI. | 16:00.23 |
| Because we'd end up with lots of word files etc, and we'd still never know what they'd used to make the pdf files etc they were testing. | 16:00.54 |
| i.e. without exactly the same drivers/software versions etc, we can't reproduce their test files. | 16:01.12 |
| (Unless by us buying it, rays customer felt able to just give us copies of their files) | 16:01.38 |
chrisl | I suspect if we own the suite, the customer would feel able to share their driver's output from the suite? | 16:02.53 |
Robin_Watts | Right, but we should check that first. | 16:03.05 |
chrisl | OR they could give us their driver | 16:03.14 |
Robin_Watts | driver/apps/installed OS etc... | 16:03.28 |
| whatever happened to be installed on their PC the day the youth experience guy got told to "go print this lot to files" | 16:04.13 |
chrisl | Well, once we have a "license" for the tests, there should be no impediment to them giving us their driver's output | 16:04.37 |
Robin_Watts | Yes, but we should check that with them. You never know how insane companies can be. | 16:05.01 |
| It would be silly to spend a lot of cash only to find that they won't. | 16:05.15 |
henrys | I would guess Quality Logic gave them app files and a printing to distiller, and the customer uses the distiller files which we would get as well, but we should check. | 16:05.51 |
chrisl | Hmm, QL only lists "10-20+ test files for each application", nothing about the "derived" files | 16:08.41 |
kens | rats, my clever fix for a customer problem causes a number of segmentation faults. I expect they are easy to fix, I'll do it tomorrow. | 16:08.50 |
| Goodnight all | 16:08.54 |
chrisl | 'nite kens | 16:09.00 |
Robin_Watts | Well, unity still has a box I can click to get a terminal, so I guess I'm OK :) | 16:10.10 |
chrisl | I wonder if they added that after the version I tried.... | 16:10.48 |
henrys | chrisl:well if that is the case we'll need to simply explain the test file situaution to QL and ask them the best way we can have the same files as the customer. I don't think this is a serious obstacle, once we know the actual test file history from ray_laptop we can proceed. | 16:11.31 |
chrisl | henrys: I can't see there would be a problem with getting the files from the customer, but as Robin_Watts says, probably best to check | 16:14.01 |
Robin_Watts | IME trying to predict what companies will do is like trying to predict what your car will fail it's MOT on. | 16:14.50 |
| It's never what you expect. | 16:14.58 |
chrisl | I have to say, it's things like this that make me question what value the QL suite actually represents...... | 16:16.00 |
henrys | not really looking at the value of the QL suite - I think we should regression test it for this customer and eliminate a predictable source of problems. | 16:22.22 |
chrisl | henrys: yeh, I just mean having an "industry standard" test suite that you can't guarantee being to test like-for-like seems insane to me | 16:26.11 |
henrys | chrisl:years back I thought it would be fun to do a test suite. With decent code coverage tools and our PDF expertise I'm sure we could put together something much better than QL | 16:30.43 |
Robin_Watts | Because testing is fun. | 16:33.02 |
| '/name henrys glados | 16:33.12 |
henrys | no money is fun you know how much these test suites go for? | 16:34.29 |
Robin_Watts | I hate perl. I hate shell. How hard can it be to grep some lines away before diffing? Very hard if you have to quote things 8 ways to sunday. | 16:34.56 |
chrisl | henrys: I reckon you need the reputation to get the big bucks - I doubt a new entry in the test suite market would command the dollars that QL do..... | 16:36.44 |
henrys | chrisl:probably true. | 16:39.12 |
Robin_Watts | chrisl: What do the numbers mean in the following: | 16:39.30 |
| Loading NimbusSanL-Regu font from %romesource/Font/NimbusSanL-Regu... 4058952 2689057 4297704 2995424 3 done. | 16:39.39 |
| Are they timings? or memory stats? or what? | 16:40.04 |
chrisl | Robin_Watts: it's something about the resources used by the font - I forget the exact meanings | 16:40.12 |
Robin_Watts | Is it reasonable for them to change between runs ? | 16:40.12 |
| The 2689057 is varying between different threads. | 16:40.41 |
| I shall just grep away "Loading" lines. | 16:41.08 |
chrisl | Hmm, not sure, it's a long time since I looked at them - I can look at it properly tomorrow if you like | 16:41.16 |
Robin_Watts | Let me have a quick grep. I'll send an email if I think it's a real problem. | 16:41.46 |
mvrhel_laptop | henrys: so I don't know if you saw but I did the commit for the preband clist stuff. I am going to go ahead and close 693015 | 16:42.38 |
chrisl | Robin_Watts: if it's causing you testing issues, you can use -dQUIET | 16:48.34 |
henrys | mvrhel_laptop:right the 28 differences, those were checked? | 16:48.47 |
mvrhel_laptop | yes. I checked them | 16:48.59 |
Robin_Watts | chrisl: Ah. Well, I've got it grepping away now, but -sQUIET is probably nicer... | 16:49.08 |
| -d eve. | 16:49.15 |
| I give up. | 16:49.20 |
henrys | mvrhel_laptop:and marcosw can now check the other fast threshold stuff now - it looks like you got the indeterms. right? | 16:49.38 |
mvrhel_laptop | henrys: yes. I believe I got the indeterminisms | 16:50.00 |
chrisl | Robin_Watts: the numbers come from vmstatus, so minor variations might be reasonable | 16:50.19 |
Robin_Watts | chrisl: Yeah, I got that far in my grep and decided that I'd be happy to ignore them. | 16:50.51 |
| OK, I'm running the apitest stuff on peeves with revised logging, so hopefully the log file won't hit 90+Gig this time. | 16:55.48 |
ray_work | henrys: sorry. I was getting coffee. I'll check with "my" customer to see what rev they have so we can decide. But even if our version is different, being able to tell them that we have purchased the ATS will let them give us THEIR test file (however it was produced) | 17:07.14 |
henrys | well we'll want to check that with both parties, but I'm sure you are right | 17:08.26 |
Robin_Watts | ray_work: That certainly sounds like how it should work, in a sane world. | 17:09.11 |
ray_work | henrys: when it comes to 'legal' opinions, don't be sure that I am right :-) Two different lawyers would have to agree | 17:09.24 |
| henrys: If you ask QL (along with asking them "how much") if another of their customers can send their version of the file, that would be enough for Dennis, I think | 17:10.30 |
henrys | okay | 17:11.02 |
ray_work | I'm sure Steve Kang understands the versioning issue with ATS files | 17:11.19 |
| the A/C in the office is not working today, so if peeves dies, don't be surprised. I've called it in, but last time it was 1.5 days before someone came out. | 17:12.23 |
| I may have to buy a new "aux" AC (I had an old Sears window mount exhausting to the huge plenum above my office, but it doesn't work too well anymore). I've turned my aux AC on tho', just in case | 17:14.04 |
| I'm going to head home (or someplace more cool)... bbiab. | 17:14.30 |
henrys | you can just turn peeves off | 17:19.17 |
| why risk overheating it. | 17:19.24 |
ray_laptop | darn. XnView (one of the best free Windows image file viewers I've found) can't handle a 'P7' (generated by plank) | 17:39.26 |
Robin_Watts | ray_laptop: convert out.plank out.png :) | 17:39.47 |
ray_laptop | trying Adobe PS 3 (without much hope) | 17:40.01 |
| Robin_Watts: I don't have Imagemagick for windows installed. | 17:40.39 |
| PS CS3 is SUCH a pig. but it sort of displays it. | 17:41.59 |
| Robin_Watts: what I'd like to convert it to is P6 (RGB). I'll just write one. Faster than installing convert :-) | 17:43.23 |
Robin_Watts | Oh, gawd. | 17:43.35 |
| In order for gsapi to work properly with multiple instances, we need to remove all uses of errprintf_nomem | 17:44.05 |
| and outprintf. and errprintf. | 17:44.14 |
| and dprintf. | 17:44.25 |
| and eprintf. | 17:44.38 |
| There are quite a few :( | 17:44.51 |
henrys | we just though debugging would be broken | 17:44.51 |
| s/though/thought | 17:45.12 |
Robin_Watts | Right. One of the SEGVs I'm getting is that we're hitting a bad jbig2, and jbig2dec is trying to print a message saying it's broken. | 17:45.34 |
ray_laptop | Robin_Watts: can't we have the errprint_nomem just go directly to stderr (dprintf and eprintf that work from a 'mem' pointer should be OK, I would think) | 17:48.05 |
| Robin_Watts: it's not good, but errprintf_nomem is not supposed to be widely used, right ? | 17:48.50 |
| Robin_Watts: that or bite the bullet and do what every other app (and the CRTL) does -- use TLS | 17:49.32 |
Robin_Watts | ray_laptop: We should move stuff to use emprintf rather than eprintf. | 17:51.07 |
henrys | but there are many test file that product errors from the postscript interpreter why is only this one failing? | 17:51.08 |
| s/product/produce | 17:51.20 |
Robin_Watts | henrys: It's not 'only' this one. That's just one of the first ones I test. | 17:51.29 |
| ray_laptop: We cannot just make errprint_nomem go to stderr, because gsapi has code in to allow you to redirect stderr etc. | 17:52.36 |
ray_laptop | henrys: the PS interpreter errorhander goes to stdout (and has the i_ctx_p to hook the I/O anyway) | 17:53.59 |
| Robin_Watts: but for the 'nomem' cases, can't we just not do the redirection. | 17:54.35 |
| Robin_Watts: if we have a libctx (or i_ctx) then we can get to the 'instance' hooks | 17:55.04 |
Robin_Watts | All we need is a mem. | 17:55.15 |
| (cos the mem gets us to libctx) | 17:55.28 |
ray_laptop | Robin_Watts: right -- that's where we pick up the lib_ctx | 17:55.39 |
henrys | we've done this exercise before passing mem all over the place. | 17:55.52 |
| I think tor8 came out loudly against this and it was all pulled out. | 17:56.07 |
Robin_Watts | henrys: Right. That was need for the globals stuff. | 17:56.09 |
ray_laptop | henrys: right -- so the 'nomem' cases will be things like libs printing | 17:56.24 |
Robin_Watts | Most places have a dev (and hence have a dev->memory) | 17:56.31 |
sebras | is it only me that is having problems installing the latest mupdf/android without first uninstalling the previous version? | 17:57.07 |
Robin_Watts | I'll see how much can be changed without too much major upsets. | 17:57.13 |
ray_laptop | Robin_Watts: filters have a 'mem' as well, but if a lib prints directly we can't do much | 17:57.16 |
sebras | I seem to recall that paulgardiner was talking about some signing..? | 17:57.34 |
Robin_Watts | ray_laptop: All our libs offer error handling callbacks etc, I think, so it's a question of setting them up right | 17:57.46 |
ray_laptop | Someone wants to buy 'ghostscript.com' -- I wonder if Miles wants to sell it ;-) | 17:58.05 |
Robin_Watts | As I say, I'll see what I can do with simple tweaks. | 17:58.13 |
| If we can get it to the stage where it's just devices that get it wrong, that's probably acceptable. | 17:58.28 |
| but the core operation should get it right, IMAO. | 17:58.39 |
sebras | ray_laptop: I guess that depends on what he's being offered... ;) | 17:58.42 |
ray_laptop | if he wants it bad enough, he can buy the whole company ;-) (maybe that's Miles' exit plan) | 17:58.48 |
henrys | why does it crash though - I would expect just to get interleaved output? | 17:59.02 |
Robin_Watts | henrys: At the moment errprintf_nomem just uses a static to hold a mem pointer. | 17:59.26 |
| If that mempointer is from a thread that's finished and gone away, then any other thread using it will access into hyperspace and we'll crash. | 17:59.55 |
henrys | got it | 18:00.00 |
Robin_Watts | And anyway, tor8 isn't allowed to have an opinion on this (until he's finished the viewer ;) ) | 18:01.39 |
mvrhel_laptop | bbiaw | 18:02.33 |
sebras | tor8: how about changing android/res/values/strings.xml::picker_title from "MuPDF" to "MuPDF 1.1"? (or possibly 1.2?) | 18:05.18 |
| that way the version number would be instantly visible when opening the app. | 18:05.33 |
| (as I read the code) | 18:05.40 |
Robin_Watts | Sounds reasonable to me. (paulgardiner is away this week, IIRC) | 18:06.01 |
sebras | Robin_Watts: good. we just need to remember to update the version number in one more place. | 18:08.04 |
| well, tor8 does, as he's making all the releases. | 18:08.19 |
Robin_Watts | Also, the gs debugging flags (-Z etc) are shared between all instances. | 18:26.56 |
| But I am henceforth rebranding that as a "feature". | 18:28.09 |
| sjbig2_luratech.c looks to be using malloc/free etc rather than gs allocation functions. | 18:34.23 |
sebras | I/DEBUG (24265): pid: 24298, tid: 24361 >>> com.artifex.mupdf <<< | 18:51.12 |
| I/DEBUG (24265): signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr deadbaad | 18:51.13 |
| this is using 1.1.... :-( | 18:51.23 |
| tor8: you were worried about the screen size changing on apples products... I guess you got a new one with iphone 5 -- 1136 x 640. | 19:03.13 |
| tor8: apps that have not been updated will be running letterboxed... | 19:03.46 |
tor8 | sebras: oh. that kind of day today? | 19:16.45 |
sebras | tor8: yeah, I don't know what happened. so I gave up testing and started learning chinese. | 19:52.40 |
| I guess I need to setup the android environment here soon if I'm continuing to test stuff... | 19:53.08 |
| tor8: I'm thinking it might be an out of memory error, where mupdf is accessing an uninitialized pointer. | 19:54.15 |
| tor8: i.e. something that ought to have been malloced but never was. | 19:54.27 |
| luckily the crash happens consistently at least. | 19:55.35 |
| after further testing this can not possibly be an out of memory error. the two documents is one with a lot of paths on it, so slow rendering, and the second just has text. | 20:09.26 |
Robin_Watts | kens: (For the logs) | 23:23.36 |
| ./base/gdevpdf.c:1925: warning: format '%d' expects type 'int', but argument 3 has type 'int64_t' | 23:23.40 |
| ./base/gdevpdf.c:1934: warning: format '%d' expects type 'int', but argument 3 has type 'int64_t' | 23:23.41 |
| Forward 1 day (to 2012/09/13)>>> | |