| <<<Back 1 day (to 2012/10/22) | 2012/10/23 |
mvrhel | henrys you there? | 03:24.53 |
henrys | yes I am here now | 04:25.30 |
mvrhel | henrys: just was going to tell you in email what you already replied to me about | 05:14.31 |
| for yesterday | 05:14.34 |
| i mean tomorrow | 05:14.41 |
| past and future mixup | 05:14.52 |
| trying to plan this cub scout camping thing for this weekend and it is form after form after form and has me all frazzled | 05:15.29 |
| i just hope the weather improves. 45 degrees and raining while camping is not much fun | 05:16.09 |
sebras | morning kens. | 07:54.18 |
kens | Hi sebras, sorry for the delay I'm trying to reply to an email about colour management, and its making my head hurt | 08:02.47 |
| Its too early for this, the coffee hasn't kicked in yet... | 08:03.02 |
sebras | kens: no worries. I have never got this obssession with coffee, luckily. :) | 08:07.32 |
kens | Hmm, I wonder if yesterday's cluster headaches were due to the Amazon data centre not linking Mondays.... | 08:13.38 |
chrisl | Are the cluster problems gone? | 08:16.30 |
kens | chrisl don't know but the register has a sotry about the East Coast Amazon data centre being flaky yesterday | 08:22.06 |
chrisl | Probably 'cause of the power outages in California... we still have several idle nodes. | 08:29.55 |
miha | hello :) | 09:33.43 |
Robin_Watts | Morning | 09:33.54 |
miha | good morning | 09:34.09 |
| (almost noon for me) | 09:34.14 |
Robin_Watts | chrisl_away: The data centre that died was East Coast. | 09:34.21 |
paulgardiner | Robin_Watts, tor8: some commits on paulg/master when you have a spare moment. | 09:53.53 |
| Robin_Watts, tor8: actually, leave the last one for now. I'm not sure it is needed after all. | 09:55.00 |
sebras | paulgardiner: how about a FZ_RENDER_ALL? | 10:24.16 |
paulgardiner | sebras: could do. I had it there for a while and then didn't use it and thought I'll leave it until I can see it will actually be used | 10:25.06 |
sebras | paulgardiner: ok, I was just reading the comments. | 10:25.24 |
| paulgardiner: wouldn't it be useful in fz_run_page()? | 10:26.15 |
paulgardiner | Possibly. I'm not sure which reads more clearly. | 10:28.33 |
sebras | paulgardiner: was just a thought. | 10:37.52 |
paulgardiner | sebras: yeah thanks. I did consider it myself and was undecided. Easy enough to add if we want | 10:40.17 |
sebras | paulgardiner: one question though -- why does one want to render the annotations separately? | 10:41.45 |
| are there toggles somewhere to do this? | 10:41.58 |
paulgardiner | Form fields are represented visually by widget annotations. When a form field is changed, we want to repaing only the annotation for the sake of speed | 10:43.35 |
sebras | paulgardiner: oh, and the representation is in the form of a content stream? | 10:44.12 |
| I thought the annotations were just dicts with information of what to draw, but not _how_ to draw it. | 10:44.32 |
| like the squiggly underlining etc. | 10:45.02 |
paulgardiner | They have appearance streams | 10:45.09 |
sebras | ok. | 10:45.28 |
fommil | I have a 200K PDF which doesn't have any images in it. It's only 4 pages! When I try to shrink it the file actually gets bigger by a few K. I need to get it down below 200K to be able to upload it to a site. Any ideas that are better than this? -> http://www.ubuntugeek.com/ubuntu-tiphowto-reduce-adobe-acrobat-file-size-from-command-line.html | 10:50.43 |
| btw, it only uses the Palatino and Century Gothic fonts and was generated from Mac Word 2012 | 10:51.17 |
kens | I'd have to see the PDF file in ordcer to make any kind of useful comment | 10:56.58 |
Robin_Watts | gah. ok, so I get the same set of differences with and without my latest patch on the cluster. | 12:38.02 |
kens | Well, that seems reasonably positive | 12:38.20 |
Robin_Watts | So, how do I test the patch with interpolation turned on... | 12:38.20 |
kens | hard-code it ? | 12:38.39 |
Robin_Watts | but then it'll compare interpolated with non-interpolated, and every file will be different. | 12:39.21 |
kens | Hmm, yes that's true | 12:39.33 |
Robin_Watts | Ah, could do back to back clusterpushes, I guess and use the 'differences from last clusterpush' to see. | 12:39.53 |
kens | As long as you are comparing against master that's a problem | 12:40.02 |
| That would work too | 12:40.10 |
Robin_Watts | but I can't run a bmpcmp to see the differences. | 12:40.12 |
kens | No | 12:40.17 |
| But you could do that locally, assuming there aren't too many | 12:40.35 |
Robin_Watts | and the bmpcmp is needed, because I'm testing enabling interpolation for rotated things. | 12:40.39 |
kens | are you expecting a lot of differences ? | 12:40.43 |
Robin_Watts | kens: yes, differences in every file with a rotated image. | 12:40.59 |
kens | Well, that's probably quite a few yes | 12:41.13 |
Robin_Watts | (with a rotated by 90 or 270 degrees image) | 12:41.14 |
kens | I can't thiunk of a good alternative | 12:41.53 |
Robin_Watts | Let me hardcode it and force interpolation on, and run just highres ppmraws. | 12:42.22 |
kens | chrisl ping | 13:10.37 |
chrisl | kens: pong | 13:11.25 |
kens | Looking at the email to support regarding EPS+preview | 13:11.45 |
paulgardiner | tor8, Robin_Watts: sanity check... and don't just say "fail" without at least hearing me out. :-) | 13:11.55 |
kens | Do you know how that is stripped in regular GS prcoessing ? | 13:11.57 |
chrisl | I don't, no - I also don't see a mail about that, but gmail isn't talking to me, right now...... | 13:13.17 |
kens | AH, well I guess I'll have to try and figure it out | 13:13.53 |
| Customer is complaining that they used to use 'gsdll API' and are now using gsapi and it doesn't work the same, which is possible | 13:14.23 |
| I'm guessing that the GS API is a lower level API, and hence doesn't execute teh stripping that teh GS DLL does. | 13:15.07 |
chrisl | I wouldn't have thought so...... | 13:15.56 |
kens | Hmm, well gsdll_execute_cont simply calls gsapi_run_string_continue. | 13:16.53 |
chrisl | I can't see anything that ought to influence that behaviour between gspi and gsdll api | 13:20.33 |
kens | Me either, the DLL call is just the API call wrapped up. | 13:22.12 |
| I'm writing that as a reply. | 13:22.20 |
chrisl | Hmm, the file has binary crap at the beginning...... | 13:23.00 |
kens | Yes, its an EPS with preview | 13:23.27 |
| The TIFF is at the end | 13:23.33 |
| Having to rebuild because I just did a rebase | 13:23.56 |
| draws a sort of left square bracket in yuvky purple | 13:25.19 |
chrisl | The only place I can see talking about previews is the DSC parsing code....... | 13:28.19 |
kens | I confess its kind of odd. | 13:30.45 |
| Hmm, our code uses init_with_args. | 13:32.12 |
chrisl | Yes, we effectively do "(filename) run" | 13:32.37 |
kens | I'm not sure what run_string_continue is supposed to do | 13:34.24 |
chrisl | Feed in buffer at a time | 13:35.33 |
kens | Well I can't see how that differs from the DLL version | 13:36.11 |
chrisl | No, me neither. | 13:36.24 |
kens | I'm going to ask them if they ahve a simple application that shows how they are usingGS. | 13:38.31 |
chrisl | kens: the binary preample is stripped in gs_epsf.ps in one of the redefinitions of "run", so I'm struggling to see how it ever worked passing data in bytewise...... | 13:48.47 |
tor8 | paulgardiner: how about separate fz_run_page_contents and fz_run_page_annots functions instead of a bit flag to run_page_partial? | 13:48.57 |
paulgardiner | Yeah. Don't see why not. | 13:49.33 |
tor8 | paulgardiner: or better yet, don't we have an annotation enumerator? then we could just do fz_run_annot in a loop | 13:49.44 |
Robin_Watts | kens: the dll and gsapi are the same thing to all intents and purposes. | 13:49.54 |
kens | chrisl that's what I expected | 13:49.57 |
paulgardiner | ... although that wasn't what I was seeking a sanity check for. | 13:49.59 |
kens | Robin_Watts : agreed | 13:50.03 |
| Robin_Watts : chrisl I suspect that they are sending different data which is one reason to ask for a demo | 13:50.19 |
tor8 | paulgardiner: sanity check for the /P pageref thing? | 13:50.39 |
Robin_Watts | paulgardiner: back. What would you like sanity checked? | 13:50.41 |
| tor8: Personally, I prefer run_page_partial with a bitflags thing. | 13:51.22 |
tor8 | Robin_Watts: that's only because you have a fetish for bit flags ;) | 13:51.42 |
Robin_Watts | We can always offer convenience functions run_page_contents and fz_run_page_annots on top of them. | 13:51.49 |
paulgardiner | tor8: don't think I need the /P pageref thing after all, now I've realised that apps need to keep a pdf_page for each cached bitmap and that has the info necessary to working out which rects need updating | 13:51.49 |
| But still there is a small problem in calculating the changed rects | 13:52.45 |
tor8 | Robin_Watts: I think I'd prefer to add fz_run_annot and move that out from fz_run_page altogether if possible | 13:53.04 |
Robin_Watts | tor8: So the app would have to take responsibility for enumerating the annotations and rendering them? | 13:54.01 |
tor8 | the transparency group stuff may make that impossible though -- did we come to a conclusion about that last week? | 13:54.03 |
| Robin_Watts: yeah. | 13:54.08 |
Robin_Watts | That's contrary to the kind of thing we've done before. | 13:54.11 |
tor8 | I dislike passing around behaviour flags to functions | 13:54.33 |
| eventually you run out of flags or want to do things in a different order, and then you're hosed | 13:55.11 |
Robin_Watts | but if we go down that route, I wonder if we should have something similar to allow us to walk the contents stream by 'marked content' markers. | 13:55.11 |
paulgardiner | It is conceivable that we will need to keep separate display lists for each annot for the sake of speed in the future | 13:55.18 |
tor8 | better to separate functionality into separate functions, IMO | 13:55.18 |
| Robin_Watts: mmm, but the marked content markers can be interleaved with state setting and clip operations :( | 13:56.28 |
Robin_Watts | Yeah, it's a mess, but I suspect it's doable. | 13:56.59 |
tor8 | Robin_Watts: one way to allow splitting marked content would be to add operators to the device interface; then you could (with care to balance clipping operations) slice and dice the display list at the end without worrying about maintaining the interpreter graphics state | 13:58.57 |
Robin_Watts | tor8: Yeah. | 13:59.14 |
| That seems a saner way of working. | 13:59.40 |
| OK, so ignore that random idea. | 13:59.47 |
tor8 | anyway, I think I want to hoist the annotation running out into a separate fz_run_annot call and force the viewers to enumerate and run them in a loop rather than passing flags to run_page. having them run individually rather than as a group would be useful for having to regenerate only the display lists for the annots that actually change, etc | 14:01.32 |
Robin_Watts | tor8: If you are prepared to hold a separate display list per annotaiton, yes. | 14:02.08 |
tor8 | Robin_Watts: and if you're not, run them all into the same display list device sequentially | 14:02.24 |
| or even into the same rendering device as the main page | 14:02.31 |
| separate run calls gives us the most flexibility, and we could add a convenience fz_run_page_annots that just loops and runs each annot if that's a common enough occurrence | 14:03.54 |
| (a convenience function that needn't add to the device/document interface) | 14:04.28 |
| run_annot would need to be added to fz_document though | 14:04.40 |
paulgardiner | Have we settled on a decision? I'm easy, other than a very slight preference for not having to change my current commits, of course. | 14:15.22 |
| The sanity check I was after was whether it seemed reasonable to add an integer field to xobjects that increments each time the contents is altered, so that I can tell its changed elsewhere in the code. | 14:19.17 |
Robin_Watts | paulgardiner: I'm happy to go with tor8s preferences. | 14:19.30 |
| A generation counter. | 14:19.44 |
| Yes, I think we may even have used that somewhere else... | 14:19.56 |
paulgardiner | Yes | 14:19.56 |
| I think I've seen you use it somewhere | 14:20.10 |
tor8 | sounds familiar... | 14:20.11 |
| I think robin's used something like it elsewhere, or at least talked about it | 14:20.24 |
Robin_Watts | probably in the pixmap caching? | 14:20.29 |
paulgardiner | It does seem to get me out of a hole and makes everything else otherwise work out neatly | 14:20.51 |
| I guess there's the 2 billion wrap around, but I can't see that ever happening. | 14:21.17 |
Robin_Watts | If we ever differ by more than 1 billion, regenerate it regardless. | 14:21.42 |
| That way we never get to the 2 billion mark where we start to not regenerate where we should. | 14:22.29 |
paulgardiner | Still you might wrap right the way around without noticing the halfway point, but this would have to happen between altering a field and updating the screen, so pretty much impossible | 14:23.06 |
sebras | Robin_Watts: ehm... you know I'm testing mupdf right? >;-) | 14:41.00 |
Robin_Watts | sebras: mmm? | 14:41.15 |
sebras | Robin_Watts: i.e. don't rely on things "never" happening. :) | 14:41.32 |
Robin_Watts | ah. | 14:41.56 |
tor8 | a generation counter overflowing... I doubt even sebras has the patience to wait for that ;) | 14:58.44 |
Robin_Watts | tor8: careful. That could be read as a geek-challenge. | 14:59.44 |
Vantir | Hi everyone, I need help. | 15:00.09 |
henrys | oh yikes time for a meeting | 15:00.09 |
Robin_Watts | so it is. | 15:00.16 |
Vantir | How do I remove ghostscript? | 15:00.33 |
| I used it for CutePDF | 15:00.42 |
| now I don't need it anymore | 15:00.50 |
Robin_Watts | Vantir: Hold with both hands and pull really hard. | 15:00.52 |
| Vantir: It depends on how it was installed. | 15:01.08 |
Vantir | Robin_Watts: Ha, wish that works! | 15:01.10 |
kens | Vantir I'd suggest you ask the CutePDF people | 15:01.10 |
Vantir | alright | 15:01.17 |
| i installed using the converter | 15:01.30 |
Robin_Watts | There *should* be an uninstaller - if there isn't, then as kens says, it's time to speak to CutePDF. | 15:01.34 |
kens | Presumably their application installed it, so their uninstaller should remove it | 15:01.37 |
henrys | paulgardiner:apologies I'm reading the report now. | 15:01.42 |
paulgardiner | henrys: np | 15:01.53 |
Vantir | but I installed separately from cute pdf | 15:02.01 |
kens | Then use the Ghostscript uninstaller | 15:02.18 |
Vantir | I didnt find it on ghostscript web page | 15:02.46 |
kens | Why woudl you expect to ? THIs is Windows right ? | 15:02.59 |
Vantir | huh? | 15:03.28 |
henrys | tor8:any word from raph beyond what I was copied in on? | 15:03.32 |
Robin_Watts | Vantir: Control Panel -> Add/Remove programs. | 15:03.33 |
| henrys: I didn't see anything from Raph. | 15:03.47 |
Vantir | Robin_Watts: I wish that it was that easy | 15:03.54 |
kens | Vantir what is the problem ? | 15:04.02 |
| If you sued the Ghostscript installer then you should be able to use teh Control Panel | 15:04.15 |
tor8 | henrys: nope | 15:04.20 |
Vantir | kens: it is not on the Add/Remove list, also cannot find in program files | 15:04.40 |
kens | WHat version of Windows is this ? | 15:04.50 |
Vantir | Win7, 64 bit | 15:04.59 |
kens | Right, so go to control panel | 15:05.04 |
| Then Program | 15:05.15 |
Vantir | ok | 15:05.16 |
kens | Programs* | 15:05.18 |
henrys | tor8:we should forward stuff to Robin_Watts and Paul - the technical pieces. yes? | 15:05.23 |
kens | Then Programs and Features | 15:05.25 |
| Locate GPL Ghostscript | 15:05.38 |
Vantir | it is not on the list | 15:05.49 |
henrys | tor8:can you do that? | 15:05.49 |
kens | Vantir did you use our installer ? | 15:06.00 |
| Because that adds it to the list ofr me on WIndows 7 64-bit | 15:06.13 |
| I currently have 3 versions installed, 9.05 (32 bit) 9.05 (64bit) and 9.06 (64-bit) | 15:06.43 |
Vantir | kens: i used GPLGS8.15 | 15:06.44 |
kens | Vantir well that is absolutely a=ncient | 15:06.58 |
henrys | paulgardiner:screen updating seemed reasonable fast on the nexus 7, just FYI | 15:07.02 |
kens | preadtes 64-bit on WIndows fort starters so it will be 32-bit | 15:07.11 |
Robin_Watts | Vantir: And may well not actually come from us. | 15:07.23 |
kens | I was coming to that :-) | 15:07.34 |
paulgardiner | henrys: even with calc.pdf? There's quite a noticable delay on my Galaxy S2 | 15:07.46 |
Vantir | heh, so now what do i do? | 15:07.54 |
kens | *If* it cmae from us (some years ago) then it should be in C:\Program FIles (x86)\gs | 15:08.08 |
henrys | paulgardiner:I don't remember it being a problem. Like I said one time I was going really fast and it hung. | 15:08.31 |
paulgardiner | henrys: also the forms we've been testing may be particularly easy | 15:08.32 |
kens | Vanti you can delete that directory, there *may* be some small pieces left around but that should be the vast majority of it. | 15:09.02 |
Vantir | kens: Robin_Watts thanks guys, guess I will have to contact cutePDF, yea I am searching the registry files for it, can't find it | 15:09.40 |
paulgardiner | Also I realised that I was updating only the current page of the android app. If I updated all three, it would get 3 times slower | 15:09.41 |
Vantir | oh and no directory also | 15:10.12 |
| wierd | 15:10.14 |
| thanks anyways | 15:10.18 |
kens | Vantir 8.15 is 8 years old, I can't tell you where it may have written to in the Registry | 15:10.36 |
henrys | tor8:I think there is useful tech stuff in the last email for paulgardiner and Robin_Watts | 15:10.48 |
kens | If you can run the executable it must be on the path, so you could start by looking in all the places mentioned on the path | 15:11.07 |
tor8 | henrys: forwarded. Robin_Watts, paulgardiner: raph replied a bit about hardware acceleration, I forwarded the mail. nothing earth shaking, just another confirmation that opengl looks to be the way forward if we want accelerated rendering. | 15:12.27 |
Robin_Watts | tor8: So a way forward would be to write a draw device that used OpenGL ES. | 15:14.37 |
paulgardiner | on hardware acceleration: I wonder if there is a format that's a bit like a display list which already commonly has support for hardware rendering? | 15:14.59 |
Robin_Watts | The canvas class? | 15:15.33 |
| I suspect that's a) not rich enough for us, and b) is the wrong side of the jni interface. | 15:15.48 |
paulgardiner | The C could generate a list of canvas commands | 15:16.14 |
tor8 | Robin_Watts: paulgardiner: yeah, the canvas api is on all the wrong side of things for us. the overhead of talking to it will probably outweigh any benefits the acceleration would yield. | 15:16.27 |
paulgardiner | Some way to export a display list into a java object, plus display2canvas writtn in java? | 15:17.19 |
Robin_Watts | OpenGL ES sounds like the right thing for us to use - as tor8 says in the email it's the defacto standard for everything these days (except TIFKAM). | 15:17.23 |
henrys | I didn't see a performance problem with forms nor general pdf rendering on nexus, so is this OpenGL suggestion premature? | 15:18.39 |
Robin_Watts | henrys: I think it's the typical "we want our next versions to be faster/smoother" etc thing. | 15:19.41 |
tor8 | rendering is fast enough for the common case. we could potentially get significant speed ups of shadings and images and text blitting though. vector art, not so much. | 15:20.05 |
henrys | tor8:yes I didn't look at complex graphics. | 15:21.31 |
Robin_Watts | Gah. 1 rendering problem in the rotated interpolate tests :( | 15:22.24 |
henrys | okay so I assume raph's mail has been digested we can move on to paulgardiner's report | 15:22.56 |
Robin_Watts | I have nothing to add to the report. All seems fine to me (and we've discussed stuff here). | 15:25.58 |
henrys | tor8? | 15:26.17 |
tor8 | nothing to add | 15:26.53 |
henrys | wow first one under 1/2 hour | 15:27.09 |
Robin_Watts | We can talk about progressive display if you'd like :) | 15:28.05 |
henrys | if we talk about anything we should discuss how to sell this stuff that's becoming the hard part in all this. | 15:30.25 |
| Robin_Watts:thanks for helping with support | 15:32.35 |
Robin_Watts | np. | 15:32.42 |
| it seemed sensible for me to answer Gemma given the time zone. | 15:32.56 |
henrys | I'm really haven't gotten up to speed on those changes anyway, so you could have explained it to her or to me. | 15:33.44 |
Robin_Watts | Well, hopefully my explanation to her makes it clear to you? | 15:34.11 |
chrisl | Robin_Watts: the PDF interpreter sets up the device for the spots in the file, so there should be little/no performance impact setting MaxSpots high for a PDF workflow | 15:34.16 |
henrys | She completely ignored my build suggestion as expected. | 15:34.31 |
Robin_Watts | chrisl: Ah, true. | 15:34.38 |
chrisl | Also I *thought* the intention was the PDF interpreter would override MaxSpots, but maybe I picked that up wrong..... | 15:35.03 |
henrys | Robin_Watts:clear | 15:35.21 |
Robin_Watts | Yeah, maybe that's been messed up. | 15:35.24 |
chrisl | henrys: I guess I could delay the build next time they have a bug-fix pending...... | 15:35.59 |
Robin_Watts | The problem is (I suspect) not the time taken to do the builds so much as the time taken to make a customer release with the appropriate patches. | 15:36.50 |
| We might be able to train monkey her into building, but applying patches is a different kettle of ballparks. | 15:37.26 |
chrisl | We've just been giving them the latest code | 15:37.41 |
Robin_Watts | chrisl: Right, but a customer release thereof. | 15:37.55 |
| To remove ourselves from the workflow we'd want to give them a patch and have them apply it to the last customer release they had. | 15:38.28 |
chrisl | Yes, I suppose. I suspect kens's assessment was correct - all their development is done "elsewhere" | 15:38.51 |
henrys | theorists | 15:41.00 |
Robin_Watts | imagines henrys changing his password. | 15:41.50 |
henrys | chrisl, kens:reminds me we found out a few shows back larry spevack has never written a line of code. | 15:42.10 |
chrisl | henrys: that comes as no surprise, and something of a relief! | 15:42.33 |
henrys | chrisl:he does have very deep color ideas knowledge that do get translated into real software products, so I was somewhat astounded. | 15:44.47 |
chrisl | henrys: I don't think I ever got that far with him - he tended to get an idea into his head, and wouldn't let go, however wrong it was..... | 15:47.36 |
kens2 | Larry is a salesman really | 15:49.11 |
henrys | chrisl:he's absolutely maddening to communicate with but I over the years I've learned to admire his persistence. He never gives up. | 15:49.23 |
Robin_Watts | paulgardiner: This remind you of anyone? :) | 15:49.25 |
paulgardiner | Very much so. | 15:49.44 |
chrisl | henrys: "persistence" and "never gives up" are not always positive attributes | 15:50.19 |
Robin_Watts | Urgh. PDF14, clipping and rotated interpolation all required to reproduce this bug. | 15:56.13 |
| Thankfully no clist as of yet :) | 15:56.20 |
henrys | refreshes coffee before meeting II | 15:57.24 |
| michael is doing school stuff and won't be here. | 15:59.59 |
kens2 | 5 minutes | 16:00.13 |
henrys | kens:are my clocks wrong? | 16:00.35 |
kens2 | No sorry, I meant I need 5 minutes, back now | 16:02.15 |
| Had to put dinner in the oven | 16:02.52 |
henrys | Robin_Watts:how is the transition stuff coming, is that done? | 16:03.15 |
Robin_Watts | sorry, back. | 16:03.48 |
| The mupdf transition stuff? | 16:03.56 |
henrys | it would be nice to copy the emails out to tech we like to keep the projects transparent. | 16:04.12 |
Robin_Watts | Everything so far has been committed. | 16:04.23 |
henrys | Robin_Watts:I'll talk to Miles about that also. | 16:04.53 |
Robin_Watts | If you build a mupdf binary from the current source you can hit 'p' to go into 'presentation' mode, and then it autoadvances, and does a simple transition. | 16:05.04 |
| There's no documentation etc at the moment. | 16:05.18 |
| I was waiting to hear back from the customer. | 16:05.26 |
henrys | we also have another private project going on ray_laptop and ets. | 16:05.41 |
Robin_Watts | I'm not sure there have been any emails to copy to tech about this. I haven't touched transitions in a week or so. | 16:06.22 |
| I've got a first version of progressive download/display working in mupdf. | 16:06.41 |
| Needs tidying, and that's not committed yet. | 16:07.06 |
| but it's on my private repo. | 16:07.14 |
henrys | ray_laptop:I'm a little nervous depending on outside help, I'd much rather see you, michael, maybe robin and I sit down and figure out ETS if we are going to sell it. Thoughts? | 16:07.23 |
Robin_Watts | ETS = Even Toned Screening ? | 16:07.51 |
henrys | yes | 16:08.01 |
| chrisl:wanted to check in on freetype - it has seemed quiet are you good with the schedule? | 16:08.54 |
kens2 | There was a clal from Werner the other day to run testsuites against the latest version chrisl, I mean to ask you if you'd tried that ? | 16:09.24 |
henrys | and finally tor8 the viewer, we absolutely need to have a GSView - what are we going to do? | 16:09.49 |
chrisl | henrys: I've got PCL and XPS now working, and getting decent results from the cluster. I need to tidy up the APIs a bit, and sort out the Windows build(s) | 16:10.06 |
henrys | chrisl:wow great | 16:13.17 |
chrisl | I also need to add in a "disablefapi" option for the non-PS languages, then I'll be ready to move it onto master | 16:14.03 |
ray_laptop | henrys: (sorry, was on the phone) Miles pointed out that the performance tuning for cust 532 is higher priority than digging into ETS. There are some things that may be quite a booger to change/fix, particularly the simplistic handling of multiple dot sizes (levels). | 16:14.43 |
henrys | ray_laptop:that doesn't effect michael or robin looking at i.t | 16:15.24 |
ray_laptop | henrys: on ETS, finding/fixing the 'overlay' of black dots on color dots is probably not too hard. | 16:16.09 |
| henrys: of course, anyone can look at it. It's GPL | 16:16.32 |
tor8 | henrys: I don't know. I must admit I'm a bit fuzzy on why it has to be gsview and should it include ghostscript and why the name change? only because miles wants to feel he got something out of the deal with russell? | 16:16.38 |
| henrys: anyway, gtk+ is not going to fly on windows, distributing with over 20 dll:s isn't something I feel is acceptable. | 16:17.08 |
Robin_Watts | tor8: Can we not distribute with 20 DLLs as a stopgap? | 16:17.37 |
tor8 | I had hoped we could get away with one dll or static linking, but from googling it looks like static linking is impossible to do as well | 16:17.45 |
| Robin_Watts: we can distribute a linux version as a stopgap ;) I think if we nail down the UI specifics of a desktop app we can make something that works with minimal amount of reworking on win, gtk and macosx. | 16:18.33 |
henrys | tor8:it's more me than miles I want artifex to have a descent viewer. The name is a separate issue really. | 16:18.37 |
tor8 | henrys: have you played with the gtk viewer on my branch? | 16:18.59 |
| henrys: you use linux on one of your machines right? | 16:19.09 |
henrys | ray_laptop:pointer to the code please? | 16:19.19 |
tor8 | henrys: it compiles and runs on osx, but installing gtk+ is a bit of a hassle | 16:19.26 |
| henrys: for one, gtk install requires that you *not* have fink or macports or any of those installed... | 16:19.50 |
| it's all a big fat mess | 16:19.59 |
chrisl | I'm surprised there isn't a MacPorts version of gtk+ | 16:20.22 |
ray_laptop | henrys: ETS is at http://www.artifex.com/ets_138/ | 16:20.28 |
henrys | tor8:windows is the important target. | 16:20.32 |
alexcher | What's wrong with the native GUI application that interacts with a portable rendering engine? | 16:20.52 |
henrys | ray_laptop:can you forward max's issues to tech please? | 16:20.55 |
ray_laptop | henrys: sure. just a sec. I'll also forward the questions as I posed them to Raph. | 16:21.25 |
tor8 | henrys: I think I know enough win32 programming to get something that works like the current gtk version and reuses much of the code working. but it'll be very limited in terms of how much gui widgetry we can add, so if you expect something that looks like sumatra or gsview with a ton of toolbars and dialogs it won't happen | 16:21.45 |
henrys | tor8:we'll take it. Maybe we can get some contributions also. | 16:23.01 |
tor8 | my idea is to make a viewer that has a search text input field, a togglable bookmark outline view, and a right click context menu for layout modes (facing, single page, fit width, fit page, that sort of thing) and access other features like print etc | 16:23.05 |
henrys | Russell is going to stop working on all his stuff, he's sold everything's name to miles, we have to take over that niche. | 16:23.44 |
tor8 | no visible scroll bars and toolbars, maximize the area for page display. basically look like our current bare bones viewer but with continuous scroll and a proper text input for search. | 16:23.54 |
Robin_Watts | ooh, we do want to show scrollbars, IMHO. | 16:24.16 |
tor8 | Robin_Watts: a scrollbar on a 100+ page document is less than worthless, IMAO. | 16:24.35 |
Robin_Watts | (unless the page is small enough to fit without them) | 16:24.35 |
henrys | and distiller built in, right? | 16:24.48 |
Robin_Watts | tor8: A scrollbar that's properly proportional is NEVER useless. | 16:24.57 |
tor8 | henrys: call external programs to convert document types we can't handle natively | 16:25.07 |
| Robin_Watts: when 1 pixel of movement exceeds one page of movement, it's pretty useless | 16:25.33 |
Robin_Watts | No, cos you use it as a draggable page selector. | 16:25.48 |
tor8 | Robin_Watts: and fisheye zoom thingy scrollbars aren't exactly standard fare | 16:25.50 |
henrys | thanks ray_laptop for the emails | 16:25.57 |
Robin_Watts | fisheye? | 16:26.04 |
alexcher | I need to leave at 12:30. Does anybody have any questions to me? | 16:26.19 |
ray_laptop | henrys: the ones from Max are still on the way. | 16:26.25 |
Robin_Watts | alexcher: Sorry to ask again, but any progress on the softmask stuff? | 16:26.42 |
ray_laptop | (they had big attachments) | 16:26.42 |
henrys | alexcher:yes what is your current project? | 16:26.57 |
tor8 | Robin_Watts: like those hyperbolic zoom things. think the mac os x dock in default configuration where it zooms near your mouse | 16:27.00 |
| to give you more local precision but still show the whole range | 16:27.08 |
Robin_Watts | tor8: Right. I wasn't expecting that. | 16:27.16 |
alexcher | henrys: softmask is the current project. | 16:27.28 |
tor8 | Robin_Watts: that's the only way a scrollbar thumb is going to be useful for a long document | 16:27.40 |
Robin_Watts | but if a page is too big to fit in the window, we need to have scrollbars. | 16:27.48 |
henrys | alexcher:okay right, that was my obligatory dilbert cartoon question. | 16:28.08 |
Robin_Watts | Unless you plan to use a 'push' model for moving around on a page ? | 16:28.09 |
tor8 | Robin_Watts: we need to have scrolling. visible scrollbars are a different issue :) | 16:28.18 |
Robin_Watts | Moving away from a consistent UI for the sake of 'coolness' is bad. IMHO. | 16:28.58 |
chrisl | henrys: oh, I've also got the stream code (I think) handling >4Gb offsets, and Ghostscript (64 and 32 bit) reading >4Gb PDF files.... again, a little tidying up needed (in progress) | 16:29.11 |
tor8 | Robin_Watts: osx 10.7+ does away with visible scrollbars by default. | 16:29.16 |
alexcher | Robin_Watts: There's some progress but I'm not ready to commit yet. | 16:29.24 |
Robin_Watts | Right, but that's consistent across OS X. | 16:29.33 |
kens | chrisl, please let me know when you havethat done so I can work on pdfwrite with it | 16:29.40 |
henrys | thanks alexcher, sorry to hold you up | 16:30.13 |
tor8 | anyway, scrollbars are there for gtk+, and I've written scrollbar code on win32 for the old mozilla plugin so that's certainly possible (but icky) whether I like them or not. and a cocoa port would be using the normal scroll view so come for free there too. | 16:30.21 |
chrisl | kens: nearly there - some 32 bit issues came up today, which I fixed, and then I need to check it on windows | 16:30.23 |
Robin_Watts | The amount of crappy apps for windows that feel free to do their own set of UI conventions because "they are better than windows standard ones" and just end up being unusable is incredible. | 16:30.30 |
kens | Norush... | 16:30.32 |
henrys | it is 9:30 so attendance optional | 16:30.33 |
tor8 | Robin_Watts: consistency on win32 is a moving target ;) you want a ribbon ui to go with those toolbars? ;) | 16:30.45 |
chrisl | Am I the only one having gmail issues today? | 16:31.06 |
kens | Mine seems OK | 16:31.11 |
chrisl | Hmm, it's taking minutes to load an e-mail for me - must be the IMAP end :-( | 16:31.53 |
tor8 | Robin_Watts: no argument there. skinned ui apps are atrocious. just look at the ati control panel. | 16:32.03 |
Robin_Watts | defected to Nvidia. | 16:32.20 |
tor8 | I regretfully jumped ship to ATI when nvidia stopped making passively cooled graphics cards | 16:32.54 |
| worst decision ever. crappier opengl drivers you have to look for... | 16:33.14 |
Robin_Watts | groans whenever he sees another vendor specific wireless configuration or sound card control panel app. | 16:33.33 |
sebras | tor8: so gtk+ is targeted only for linux? | 16:35.32 |
| tor8: or does it still include osc? | 16:35.46 |
| osx. | 16:35.48 |
ray_laptop | Robin_Watts: the thing that irritates me on Windows is all of the apps that improperly handle their dialog boxes when the system scaling is non-standard | 16:36.01 |
tor8 | sebras: osx only. if I have to make separate apps for each platform, a native cocoa version for osx isn't a lot more work. | 16:36.19 |
| sebras: s/osx/linux/1 | 16:36.28 |
sebras | tor8: right. | 16:36.42 |
ray_laptop | Robin_Watts: buttons get placed outside the dialog boxes sometimes and text sometimes overflows the area it's supposed to fit | 16:37.14 |
Robin_Watts | ray_laptop: Yeah. Well, that's windows for you :( | 16:37.47 |
| The whole world should use RISC OS! :) | 16:37.56 |
ray_laptop | and windows programmers | 16:38.02 |
| henrys: the thing I hoped Raph could do fairly easily is to address the description of the ETS parameters. I can look at the code and see the permissible range of values, but things like the "useful range" is harder. | 16:48.34 |
alexcher | I'm back. | 16:50.16 |
Robin_Watts | oh, great. I spoke to soon. clist is involved :( | 16:50.19 |
ray_laptop | henrys: also I can puzzle out the "purpose" of some of the parameters, but really wanted that info from Raph, because my guess might be wrong and I didn't want to give out guesses | 16:50.24 |
Robin_Watts | s/to/too/ | 16:50.25 |
henrys | ray_laptop:I hope he can, but our history with him ... he could fall off the earth at any minute and months can go by... | 16:50.28 |
ray_laptop | Robin_Watts: so, you have all components involved. | 16:50.52 |
| Robin_Watts: what problem is this ? | 16:51.02 |
Robin_Watts | ray_laptop: I've updated my patch to do interpolation on landscape bitmaps. | 16:51.27 |
ray_laptop | Robin_Watts: I see. and it breaks ONLY if going through the clist ? | 16:51.53 |
Robin_Watts | and if I run a cluster test (gs highres -filter=ppmraw) where I've nobbled the code to always turn interpolation on for landscape bitmaps, then I get differences, as you'd expect. | 16:52.28 |
| bmpcmping them, I see that all the differences are fine - except one. | 16:52.43 |
ray_laptop | Robin_Watts: based on the performance issues, maybe just doing pre-clist interpolation is better | 16:52.44 |
Robin_Watts | When I tested this patch before it was failing with the clist. | 16:53.36 |
| but that was before my recent changes to the code to support clipping. | 16:53.49 |
kens | OK time to make dinner, night all | 17:01.14 |
| I may come back and try to catch Michael if he is on later | 17:01.27 |
ray_laptop | does XPS have control for interpolation on images ? | 17:11.22 |
Robin_Watts | ray_laptop: Are you going to push your image changes? | 17:41.05 |
| (The code to only send images through the clist for the appropriate bands) | 17:41.31 |
ray_laptop | Robin_Watts: running a clusterpush now | 17:41.39 |
Robin_Watts | ok | 17:41.46 |
ray_laptop | lots of differences. :-( | 17:42.17 |
| I have to go to cust 532 for a meeting, so I'll have a look later | 17:42.39 |
| started the bmpcmp | 17:43.02 |
Robin_Watts | ok. | 17:43.04 |
ray_laptop | later ... | 17:43.12 |
miha | Robin_Watts: i manage to sometimes crash current version of android mupdf.. but i see no error message, just "clean" exit | 17:44.17 |
Robin_Watts | miha: Try running with the device plugged in, and "adb logcat" running on the PC. | 17:44.49 |
| That way you'll see any error messages. | 17:44.55 |
miha | Robin_Watts: i do. there's no crash. just mupdf activity closes. | 17:45.10 |
Robin_Watts | oh. how odd. | 17:45.23 |
sebras | miha: can you reproduce it consistently, or is it random? | 17:46.26 |
miha | sebras: random, when i scroll a lot fast | 17:47.04 |
Robin_Watts | miha: And you've built this from a recent git checkout ? | 17:47.28 |
sebras | miha: and have you made any changes yourself? | 17:47.53 |
miha | Robin_Watts: yes. only modified it so readerview fetches more pages at once (left and right of current page).. | 17:48.00 |
| in a queue | 17:48.11 |
Robin_Watts | can you make it crash with vanilla mupdf? | 17:48.20 |
miha | sorted from current page :) | 17:48.20 |
| Robin_Watts: let me see | 17:48.28 |
| yes | 17:51.53 |
sebras | miha: do you get any logcat error in this case? | 17:52.20 |
| miha: any information would be interesting since no one here appears to be able to reproduce this but you. :) | 17:52.49 |
miha | sebras: even i have problems. i better be quiet :( | 17:58.09 |
sebras | miha: I just tried it with a single-page pdf with a big image image, zoomed in full, flicked around and let it render after it flick, and then as soon as the render was completed I flicked again.\ | 17:58.46 |
| on my HTC sensation I see a steady memory usage, without too much fluctuation. | 17:59.36 |
miha | sebras: how can i see memory use? | 18:00.48 |
| i have android sdk and eclipse | 18:01.13 |
sebras | miha: I just observed the GC_EXPLICIT lines in logcat which on my device consistently stated 62% unused. | 18:02.53 |
| miha: I have also been told that ddms which is part of android sdk can show heap usage. | 18:03.26 |
| miha: are you using windows or linux? or perhaps osx? | 18:04.17 |
| in linux there is a small green cylinder in the upper left corner which needs to be checked after you have clicked on the process in the tree beneather you devices in ddms. | 18:04.46 |
miha | sebras: linux | 18:09.55 |
sebras | miha: great, then I'm in better position to help you with ddms. | 18:10.18 |
| paulgardiner: adb shell monkey -p com.artifex.mupdf -v 5000 | 18:12.26 |
| paulgardiner: wow... it still survives... :) | 18:12.39 |
| paulgardiner: this is like an automated version of me. I though I had copyright on me! | 18:13.08 |
Robin_Watts | what's that do? | 18:13.18 |
sebras | Robin_Watts: sends random gestures to the app. | 18:13.50 |
| Robin_Watts: 5000 of them. | 18:13.55 |
Robin_Watts | ah. flings poo at it :) | 18:14.05 |
sebras | Robin_Watts: gesture based fuzz testing. | 18:14.18 |
Robin_Watts | I see why they call it monkey. | 18:14.24 |
sebras | Robin_Watts: but now I killed mupdf! :) | 18:15.00 |
| java.lang.IllegalArgumentException: width and height must be > 0 | 18:15.28 |
| looks like there's more thinking to do... | 18:15.40 |
Robin_Watts | henrys: DRY violation? | 18:16.26 |
miha | sebras: thank you. so this heap is all i need to monitor? :) | 18:18.22 |
| sebras: non-java is libmupdf? | 18:18.51 |
sebras | miha: I think so. I only learned of this tool yesterday. :) | 18:19.12 |
miha | sebras: http://www.wolfey.si/tmp/mupdf.txt this i think is crash.... so creating too many bitmaps? :D | 18:32.05 |
sebras | miha: what is above this part in logcat? | 18:37.27 |
miha | sebras: refresh | 18:42.01 |
| sebras: my dirty hacks? :) | 18:42.59 |
sebras | miha: I'm just wondering since the part you cite says "Found by:" and I'm wondering "what is it that is found?" | 18:43.29 |
miha | btw i think in PageView.java whenever is line mEntire.setImageBitmap(null); that bitmap should be recycled too http://stackoverflow.com/questions/3823799/android-bitmap-recycle-how-does-it-work | 18:44.50 |
henrys` | likes public libraries | 18:45.20 |
| dry - do not repeat yourself | 18:45.52 |
Robin_Watts | Ah, right. I guessed the meaning, but not the TLA :) | 18:47.19 |
| enum { DEFAULT_MITER_LENGTH=5 }; | 18:47.46 |
kens | Does IRC have a buil-in compiler now ? :-) | 18:49.56 |
Robin_Watts | syntax error in kens at line 1. | 18:50.20 |
miha | sebras: http://www.wolfey.si/tmp/recycle.txt i'll use this, waiting for GC to clean bitmaps is bad | 18:57.58 |
sebras | miha: ok. I'm by no means the android expert around here. | 18:59.18 |
miha | sebras: i just pretend a lot :) | 19:02.55 |
sebras | kens: harder, better, faster, stronger. | 20:33.42 |
| Forward 1 day (to 2012/10/24)>>> | |