| | <<<Back 1 day (to 2012/10/17) | 2012/10/18 |
| sebras | the hitbox invertation works less well though. if I search for part of a word and then the entire word I get a nice blinking effect. | 00:00.06 |
| mvrhel_laptop | night all | 06:32.27 |
| sebras | OMG! tor8 beat kens today! .... ;) | 07:21.45 |
| kens | maybe he didn't log out.... | 07:22.09 |
| tor8 | yes, but now I'm not so sure it was a good idea... *yawn* | 07:22.42 |
| sebras | he did last night, but he might have snoozed after ligging in.. | 07:22.52 |
| kens | Seems like a high price to pay... | 07:23.24 |
| sebras | tor8: you only have to do it once to prove it's possible. ;) | 07:23.39 |
| Robin_Watts | Bah. No meta operations on mupdf streams. | 12:17.49 |
| | I had a nice idea about how to do this progressive stuff as I was out running, but it relied on a meta call for streams. | 12:18.30 |
| paulgardiner | If we had a lock guarding the dictionary structures and the xref, then we could read pdf in one thread, while rendering whatever we have at any moment in time in another... maybe? | 12:32.24 |
| tor8 | Robin_Watts: meta call? | 12:39.46 |
| Robin_Watts | Yeah, akin to dev_spec_op. | 12:40.07 |
| | so I could have a call to say "are you a stream that's still downloading?" | 12:40.49 |
| | and to tell me how much data it had got so far, and how much it was expecting, kinda thing. | 12:41.10 |
| | I can probably live without it? | 12:41.41 |
| tor8 | I vaguely recall talking about loading streams into fz_buffers in a worker thread and blocking stream open operations until the buffer was complete | 12:42.25 |
| ray_laptop | kens: are you working on supporting PDF-A/3 ;-) ??? | 14:39.27 |
| | I got a new release (spam) emial from Luratech saying their PDF compressor now supports PDF-A/3. Just what the world needs -- embedding word and excel docs in a PDF :-( | 14:42.26 |
| | so you can archive the viruses :-) | 14:42.47 |
| Robin_Watts | We have a potential muPDF customer that wants us to load a PDF/A file, alter it (add a signature image as an annotation), and save it out again, also as PDF/A. | 14:51.15 |
| | kens: What are the restrictions on PDF/A again? Is it likely that just adding an image is going to upset it? | 14:51.39 |
| kens | Robin_Watts : no | 14:59.07 |
| | colour spaces mthough | 14:59.14 |
| | rayjj : no | 14:59.27 |
| | Robin_Watts : need to make sure image colour space is consistent with the PDF/A file | 15:01.44 |
| | ICCBased is always OK | 15:01.57 |
| Robin_Watts | but DeviceRGB or DeviceGray might not be? | 15:02.13 |
| kens | DeviceN is not permitted IIRC. For device spaces it must be *one* of DeviceRGB or DeviceCMYK | 15:03.45 |
| | and *all* the colours in the document must be in the smae device space | 15:04.15 |
| | So whatever the PDF/A file is using, that's what you need to put the image in | 15:04.31 |
| | Or ICCbased | 15:04.36 |
| Robin_Watts | Ah, so I can't unconditionally insert the image as DeviceRGB, rats. | 15:04.45 |
| kens | Am I making sense ? | 15:04.47 |
| Robin_Watts | You are. | 15:04.50 |
| | Thanks. I'll worry about it if the deal actually comes off though :) | 15:05.05 |
| kens | I beleive Gray is OK as well | 15:05.06 |
| | Ler me go quickly review the spec | 15:05.20 |
| | Yes, section 6.2.3 | 15:06.16 |
| | All spaces shall be ICCbased. | 15:06.39 |
| | But.... | 15:06.42 |
| | You can use *either* DeviceRGB or DeviceCMYK, provided you supply an OutputIntent profile | 15:07.09 |
| | If you do, tehn DeviceGray is also OK (treated as RGB or CMYK for colour purposes) | 15:07.31 |
| | SO if you image is color, hyour best bet is to use an ICCBased space. | 15:07.51 |
| | If you can't do that then you need to look at the PDF/A file ands see if it has an OutputIntent | 15:08.06 |
| | If it doesn't, you can add one that is the space yo ulike and use the colour space for the image. | 15:08.28 |
| | If it does, then you need to examine it ands see if its RGB or CMYK, and convert your image approriately | 15:08.48 |
| | Kind of pants really. | 15:09.22 |
| Robin_Watts | Sounds like a massive pain in the ass frankly. | 15:12.20 |
| kens | Yes indeed. Your best solution might be to run it through GS :-) | 15:13.05 |
| | Hmm, my code has suddenly started working, I wonder why.... | 15:13.39 |
| | Ah, GPF :-) | 15:14.11 |
| paulgardiner | What's GPF? I think my code could do with some of that. | 15:46.41 |
| Robin_Watts | General Protection Failure (or Fault). A SEGV, Bus Error, Type 5 :) | 15:47.17 |
| paulgardiner | Ah, no. My code already has plenty of that. | 15:47.48 |
| | I read it as "GPF made my code work" | 15:48.12 |
| kens | Nah, it just got further before falling in a more total variety of heap | 15:50.18 |
| kens | hates garbage collection | 15:50.43 |
| Robin_Watts | kens: I find that a memento build with --debug=gc-disable can be very useful. | 15:52.28 |
| kens | Robin_Watts : I kind of know what's wrong, sort of, its fixing it that;s the problem. | 15:52.53 |
| | If it weasn't GC objects it would be easy | 15:53.02 |
| Robin_Watts | It lets you debug memory leaks etc without gc randomly moving the floor under you. | 15:53.13 |
| | ok. | 15:53.23 |
| kens | My head hurtrs with all this stuff, I'm going to wait and see what MIchael has to say about this... | 16:02.12 |
| Robin_Watts | mvrhel_laptop: Morning. | 16:15.08 |
| mvrhel_laptop | good morning | 16:15.17 |
| Robin_Watts | Kens was talking about having a gc memory problem that he wanted to discuss with you. | 16:15.30 |
| mvrhel_laptop | ok. hold on a sec | 16:15.42 |
| | oh is kens gone | 16:16.43 |
| Robin_Watts | indeed. | 16:16.59 |
| mvrhel_laptop | I know he sent me an email and also gave me a bug | 16:17.05 |
| | so I will look over all of that | 16:17.13 |
| Robin_Watts | Ah, that's probably it then. | 16:17.15 |
| mvrhel_laptop | need to also do some customer 330 follow up | 16:17.29 |
| Robin_Watts | Customer 330 is to michael as customer 532 is to ray. | 16:17.53 |
| henrys | 190's my curse. | 16:33.13 |
| | but I did win a trip to france! | 16:33.28 |
| Robin_Watts | henrys: really? | 16:33.42 |
| henrys | yes I think you'll be coming to california - for reasons you don't know about yet, and I'll be going to france - we should pass each other somewhere over the Atlantic | 16:34.36 |
| Robin_Watts | coo. | 16:36.43 |
| henrys | sabrina is happy | 16:37.01 |
| | Robin_Watts:meaning to ask you if you've been to Majorca - sounds a bit like Malta to me. Is it very touristy? | 16:38.30 |
| Robin_Watts | I went to Majorca when I was 7 :) | 16:38.46 |
| | Yes, it's very touristy. | 16:38.53 |
| | Helens aunt lives in Minorca (less touristy) | 16:39.08 |
| | I suspect Majorca is Malta without the history :) | 16:39.33 |
| mvrhel_laptop | henrys: how did you win a trip to france | 16:40.20 |
| henrys | I was being a smartass - I'm going to visit the customer. | 16:40.50 |
| mvrhel_laptop | ha | 16:40.57 |
| | ok. I was wondering because our local NPR station is giving a trip to France as a raffle prize | 16:41.32 |
| | and the kids wanted me to enter | 16:41.41 |
| | if they heard someone actually one such a thing the pressure would be on me even more to enter | 16:42.20 |
| Robin_Watts | mvrhel_laptop: First Prize an all inclusive 1 week trip to France. | 16:42.27 |
| | Second price, an all inclusive 2 week trip to france... | 16:42.39 |
| mvrhel_laptop | hehe | 16:42.46 |
| | spoken like a true brit | 16:43.01 |
| henrys | yeah you know I'm fine with going but Sabrina is a bit more excited than I. | 16:43.20 |
| mvrhel_laptop | that area is pretty neat | 16:44.04 |
| Robin_Watts | hey marcosw. Do you have multiple drives on the mac that's corrupting the symlinks ? | 17:20.47 |
| marcosw | Just the main drive and a time machine volume. | 17:21.24 |
| Robin_Watts | Time machine volume is a local thing, or network thing though? | 17:21.57 |
| marcosw | local, via usb. | 17:22.19 |
| Robin_Watts | Ok, so multiple drives then. | 17:22.32 |
| | That sounds like the common denominator to all the problem in that thread. | 17:22.50 |
| | Is it worth disconnecting that drive and seeing if the corruption goes away? | 17:23.14 |
| marcosw | yeah, but I've add multiple drives for years and years. Seems wrong to disconnect the backup drive when I'm having problems with file corruption on the main drive. | 17:24.24 |
| Robin_Watts | marcosw: True. | 17:24.46 |
| | Could it actually be time machine causing the problem? | 17:25.12 |
| marcosw | might be. Since I'm now monitoring the gcc symlink I will hopefully find the exact time it's corrupted. Should be easy to see if it matches up during a backup. | 17:28.06 |
| | The other thing that I'm wondering about is system updates. I ran a system update last night; I think it was a Java update. | 17:28.53 |
| | Thinking back on it I've run three system updates in the past two weeks and have had symlink issues three times. However I don't recall if the date/times work out. | 17:29.54 |
| ray_laptop | OMG! The amount of times we do all of the 'setup' in pdf4_fill_rectangle for tiny rectangles when doing 'copy_color' with data that came from interpolated images is horrendous. | 17:32.05 |
| | mvrhel_laptop: I think it is time for me to code up a pdf14_copy_color for the simple cases (i.e. most of them) | 17:32.53 |
| mvrhel_laptop | sounds good | 17:33.31 |
| | bbiab | 17:34.13 |
| ray_laptop | hmm... amazingly, on the simulator, totally skipping the call to 'fill_rectangle' from the default copy_color only saves us 30% (wall clock time when writing the output file). Stepping through the code, I would have thought it would be more dramatic. | 17:49.11 |
| | and since writing the page to disk is in a separate thread and the pages without transprency run in less than 1/2 the time of the pages with transparency, writing the file is probably not a factor | 17:50.37 |
| Robin_Watts | How good is the target with floating point? | 17:51.33 |
| ray_laptop | still, if I can get 30% then that's better than nothing. But recent experiences have been disappointing -- what speeds up the simulator seems to have NO effect on the target printer | 17:51.36 |
| Robin_Watts | ray_laptop: Get them to remove the fill_rectangle call and test the time? | 17:52.43 |
| ray_laptop | Robin_Watts: might be worth a shot. Turnaround on changes and target runs is frustratingly slow, but I will wait for data from them before doing premature optimization (at least that involves a lot of risk or code changes) | 18:01.31 |
| malc_ | Robin_Watts: getting a segfault when using transitions | 18:05.39 |
| Robin_Watts | in your own code? | 18:06.15 |
| malc_ | no, vanilla mupdf | 18:06.34 |
| Robin_Watts | ok, off the phone, sorry. | 18:27.52 |
| | malc_: Can you see where it's SEGVing? | 18:28.16 |
| malc_ | Robin_Watts: sure, want a backtrace? | 18:28.25 |
| Robin_Watts | please. | 18:28.29 |
| malc_ | Robin_Watts: http://pastebin.ca/raw/2247278 | 18:29.27 |
| | scenario, mupdf <file> p,space,q -> segfault | 18:30.08 |
| Robin_Watts | brilliant, thanks. | 18:30.35 |
| | I can't look into it now, but I will do so asap. I've got everything I should need to recreate it. | 18:31.06 |
| malc_ | glad to be of service | 18:31.23 |
| sebras | Robin_Watts: sebras/master | 22:22.22 |
| Robin_Watts | sebras: Thanks. | 23:50.35 |
| | Forward 1 day (to 2012/10/19)>>> | |
IRC Logs