| <<<Back 1 day (to 2013/10/01) | 2013/10/02 |
mvrhel_laptop | aha! I see my issue with the knockouts now. | 03:54.19 |
paulgardiner | tor7: there are three commits up on paulg/master - introducing display lists and some small reorganisations. | 12:04.12 |
tor7 | paulgardiner: they look okay, I think. I've come down with a nasty cold so not feeling particularly bright today :( | 12:39.45 |
Robin_Watts | tor7: commiserations. | 12:40.05 |
paulgardiner | tor7: thanks. Hope you get over the cold quickly. | 12:50.45 |
alex__ | hi all | 13:00.49 |
| anybody here? | 13:01.05 |
kens | Just us chikens | 13:01.16 |
Robin_Watts | cluck cluck | 13:01.41 |
alex__ | can somebody help me with Ghostscript? | 13:01.41 |
| Error 1007: An error occured while running Ghostscript | 13:01.46 |
kens | That's not a Ghostscript error | 13:01.56 |
alex__ | I know | 13:02.15 |
| it's bullzip | 13:02.28 |
| Ghostscript cant start in Windows 8 and Windows Server 2012 | 13:02.46 |
kens | Hmm, then I htin kyou will have to takle it up with Bullzip | 13:02.50 |
alex__ | how I understand | 13:02.58 |
kens | Ghostscript certainly starts up in WINdows 8, and I'm working on a customer problem reported on Server 2012, so I thgink it probably works there too | 13:03.15 |
| Possibly the port monitor installed by Bullzip doesn't work on WIndows 8 though | 13:03.40 |
| In fact, it almost certainly doesn't, because Microsoft changed the printing architecture. | 13:03.57 |
| SO the most likely problem isn't Ghostscript , but the Bullzip port monitor | 13:04.14 |
alex__ | They said bullzip works fine with this system... | 13:04.35 |
| sad | 13:04.36 |
| ok, thanks | 13:04.56 |
kens | Well I'm running GS on WINdows 8 here, and I have a vm with server 2012 on it, and that runs GS as well. | 13:04.56 |
| Which version of Gs are you using ? | 13:05.03 |
alex__ | 9.10 | 13:05.09 |
kens | Well, try running GS fomr the command line | 13:05.20 |
| just gswin32 should get you a window | 13:05.38 |
| Or gswin64 if you are using the 64-bit version of Ghostscript | 13:05.50 |
alex__ | command line woks fine | 13:05.56 |
| works | 13:06.01 |
kens | Well then GS works on WIndows 8 :-) | 13:06.10 |
alex__ | ))) | 13:06.14 |
kens | I'm afraid I cannot tell you what the Bullzip problem is, sorry :-( | 13:06.38 |
alex__ | thanks for all, anyway | 13:06.56 |
kens | chrisl I have a handle on the problem with some CET files and Metrics. | 13:07.24 |
chrisl | Error 1007 is apparently "The requested operation cannot be performed in full-screen mode." - not sure how GS could trigger that | 13:07.35 |
| kens: that's good - is it simple? | 13:07.57 |
kens | chrisl I'm not sure.... | 13:08.04 |
| THe 'problem' is that we aren't applying overrides to LSB | 13:08.21 |
| THe 2 and 4 number arrays provide an override for the left side bearing, and we ignore it | 13:08.42 |
| THis is rendering by the way | 13:08.56 |
| We *do* use the width override I think, even in the 4 element case | 13:09.32 |
chrisl | I thought rendering got the correct output? | 13:09.41 |
kens | Not for 16-03.ps | 13:09.48 |
| Note htis is not the problem for the customer | 13:10.04 |
chrisl | Oh right, I see. Well that's a little odd....... | 13:10.41 |
kens | pdfwrite gets the result of 16-03.ps wrong for the same reason, it ignores the LSB override (actually I'm not sure it can even tell its been overridden) | 13:10.45 |
| Well the font is (of course) barking. It sets the LSB = width of the glyph | 13:11.09 |
chrisl | So that's clearly a rarely used feature! | 13:11.54 |
kens | In the real world I don't think anyone actually uses the 2 or 4 element form of the Metrics array | 13:11.57 |
| The only reason for changin the LSB would be to introduce white space to the left of a glyph, if you change LSB and width you can introduce white space on each side of a glyph | 13:12.34 |
| But its the sort of thing which will not give a pleasing result. | 13:13.06 |
chrisl | Sure, but there are much easier ways to achieve the same result, if required. | 13:13.15 |
kens | Yes, simply using ashow for one, or xshow etc. | 13:13.38 |
chrisl | So, unless it actually helps with the customer problem, you can just open a bug and punt it to me | 13:13.52 |
kens | I'm guessing nobody uses it in the real world. | 13:14.00 |
| Yeah I'm going to move on to a different CET file wihch uses WMode 1 | 13:14.17 |
| And which I htink renders correctly | 13:14.27 |
chrisl | We'd probably have seen complaints if anyone used it - that's one reason I never got around to looking at it | 13:14.50 |
kens | Fixing 16-03.ps in pdfwite looks like a problem and I don't want to tackle it at the moment. And yes, since nobody really uses it, I'm not inclined to do much about it | 13:15.07 |
| So on 10 19-03.ps | 13:15.17 |
| I'll open a report and assign it, just in case you ever want to look at it. | 13:15.38 |
chrisl | My first thought is that it's probably not hard to fix - but I enter the real world, and think it'll probably be a 'mare! | 13:16.28 |
kens | Yeah, I know right now that fixing it in pdfwrite is not going to be easy. I'd need to know the original side-bearing first and I don't think I do | 13:17.14 |
| OK bug opened, I've chosen Font API hich maybe isn't strictly correct but it means it gets assigned to you | 13:19.26 |
| Now for a coffee | 13:19.31 |
henrys | tor7:are the light flickering? http://www.nytimes.com/2013/10/02/world/europe/jellyfish-invasion-paralyzes-swedish-reactor.html?partner=rss&emc=rss&_r=1& | 13:43.09 |
tor7 | henrys: hehe :) | 13:45.08 |
| henrys: now that paulgardiner looks like he has the ios app under control, I'm considering getting a chromebook pixel as a new laptop... only problem is, can't buy it here :( | 13:46.09 |
henrys | tor7:really why not a conventional laptop, the chrome books are pretty low powered no? | 13:48.14 |
tor7 | the chromebook pixel has a 13" 2560x1700 panel, 3:2 aspect ratio... | 13:48.37 |
| all the conventional laptops are 16:9 | 13:48.47 |
| but yes, the pixel is still pretty low powered compared to a conventional laptop. only an i5. | 13:49.24 |
henrys | well one of us can order it an bring it to maui but that is far off | 13:51.21 |
| resolution is impressive | 13:51.42 |
Robin_Watts | tor7: Where is it sold? | 13:53.03 |
tor7 | Robin_Watts: I can find one (somewhat overpriced) on amazon.co.uk via a thirdparty seller | 13:53.23 |
| Robin_Watts: it's only sold on google play marketplace, IIRC | 13:53.34 |
Robin_Watts | tor7: 1049 UKP from google play for me. | 13:53.45 |
tor7 | Robin_Watts: interesting. how much would shipping that to sweden be for you? | 13:54.46 |
Robin_Watts | tor7: Not sure. Not a huge amount, I'd imagine. | 13:55.02 |
| You're in the EU, right, so no stupid import taxes to pay ? | 13:55.16 |
tor7 | I can find one for 1100 UKP on amazon.co.uk, but I guess there'll be shipping costs added on that one too | 13:55.17 |
| Robin_Watts: yeah. unlike if I got it from the US. | 13:55.27 |
| US gets slapped with a $15 customs fee and 25% added VAT | 13:55.47 |
| Robin_Watts: if I send you the money to your paypal, could you order one and mail it? | 13:56.41 |
Robin_Watts | tor7: I wonder if I can order it from here to be sent to you, on your credit card... | 13:57.08 |
| tor7: Let's swap to skype. | 13:57.55 |
henrys | tor7:are you sure you can do development on these things - I've always thought of them as thin clients but I guess with an i5 and a decent amount of memory it should work. Does the GNU tool chain work like a regular linux box? | 13:59.00 |
tor7 | henrys: I plan on installing debian on it. | 13:59.12 |
| Linus uses one as his main machine, so I expect everything Linux to Just Work ^TM | 13:59.27 |
henrys | wow I'm surprised byzantine X11 can handle that hi res monitor | 14:00.06 |
| tor7:I'm thinking about linux too I'll be curious to see how you do with the chromebook. | 14:05.08 |
tor7 | henrys: I'm arranging for Robin to ship me one, I'll bring it next meeting unless it's absolutely useless :) | 14:05.54 |
mvrhel_laptop | rats, I see a weird case that uses a mask that is used to change the knockout effect that we still get wrong | 14:10.00 |
Robin_Watts | tor7: What onboard storage does the pixel have ? | 14:40.06 |
| eek, just 32Gig ? | 14:40.30 |
tor7 | 32gb SSD | 14:40.32 |
| more than enough for my needs | 14:40.37 |
| but YMMV :) | 14:40.47 |
| 4G RAM is the more worrying limit | 14:40.57 |
| but nothing I expect to pose any problems | 14:41.08 |
| it might be easy to open it up and install a bigger SSD, but I haven't looked into that | 14:41.38 |
Robin_Watts | RAM and SSD are both soldered in | 14:42.29 |
kens | Soldering gun works :-) | 14:42.53 |
tor7 | boot from SD-card ;) | 14:43.09 |
Robin_Watts | tor7: I can imagine it would be very tempting to stick a 128Gig SD in there, yes. | 14:59.17 |
henrys | I can't imagine living with 32G that's like and iPad or something. | 15:00.21 |
| s/and/an | 15:00.50 |
Robin_Watts | How big is a mupdf git clone these days? :) | 15:07.33 |
kens | I'm sure 32G is enough to do GS development | 15:07.51 |
Robin_Watts | as long as you don't want any test files :) | 15:08.11 |
mvrhel_laptop | aha. finally one that mupdf also gets wrong.... | 15:21.35 |
| I will open a bug about an interesting knockout case for mupdf today Robin_Watts | 15:22.53 |
Robin_Watts | mvrhel_laptop: If we both disagree then the universe is at fault :) | 15:22.58 |
mvrhel_laptop | oh I have to send out an email still, but I am going to be out tomorrow henrys | 15:24.02 |
tor7 | Robin_Watts: a mupdf checkout is about 200M, 65M in the .git | 15:24.08 |
mvrhel_laptop | and friday morning | 15:24.14 |
| off to pioneer camp with my daughter | 15:24.30 |
henrys | mvrhel_laptop: have fun | 15:34.34 |
| ccitt fuzzing bug woohoo! | 15:36.04 |
kens2 | goodnight all | 16:10.25 |
Robin_Watts | Ha. | 16:33.22 |
| America was not shutdown properly. Would you like to start America in safe mode, with free healthcare & without the guns? [Recommended] | 16:33.57 |
ray_laptop | Robin_Watts: oops. I was already responding to the URGENT guy and hit send before I saw your reply. Thanks. | 16:35.58 |
Robin_Watts | ray_laptop: no worries. | 16:36.13 |
ray_laptop | Robin_Watts: THAT'S FUNNY! I have to send that to my nephews that work in IT at the local University Hospital | 16:36.57 |
| Robin_Watts: was that original ? | 16:37.09 |
Robin_Watts | Not original to me, no :) | 16:37.26 |
| Someone should do a 'screengrab' of a windows boot screen edited with that text. | 16:37.53 |
ray_laptop | Robin_Watts: I should be able to grab one from a VM -- give me a minute... | 16:39.55 |
| then photoshop in the text | 16:40.26 |
| Oh, darn. The VM I have is XP -- it doesn't do that message :-( | 16:46.38 |
| or at least I can't get it to give me that message, even if I force "power off" to the VM. | 16:47.07 |
| I did power off during the startup and got the "Windows did not start correctly" screen. | 16:47.33 |
| I guess I really shouldn't waste time on it. | 16:48.01 |
Robin_Watts | ray_laptop: I wouldn't stress on it. Someone with less to do will do it, I'm sure :) | 16:52.37 |
ray_laptop | wonder why peeved is so slow. It's supposed to be fast :-( | 17:11.48 |
mvrhel_laptop | ok. I see my issue. if I have other non-isolated groups nested within non-isolated groups all within a knockout group, the backdrop that I have to blend with is the parent knockout. The issue I ran into was that one of the nested non-isolated groups was also a knockout group and getting its special backdrop buffer initialized with the current state as opposed to the parent knockout. so I... | 17:12.05 |
| ...need to traverse the stack through non-isolated groups making sure that I don't have a parent that was a knockout when I create the background buffer for the knockout group | 17:12.07 |
Robin_Watts | ray_laptop: Have you pushed the "turbo" button? | 17:12.12 |
| mvrhel_laptop: That sounds horribly familiar. | 17:12.32 |
mvrhel_laptop | hopefully this is the finally issue | 17:12.33 |
Robin_Watts | does mupdf get that right? | 17:12.41 |
mvrhel_laptop | Robin_Watts: this is the one that mupdf does not get right | 17:12.45 |
ray_laptop | maybe I'm running a debug version of the kernel ;-) | 17:12.47 |
Robin_Watts | ha. | 17:12.51 |
| mvrhel_laptop: So you'll fix it in mupdf too while you're at it then? :) | 17:13.17 |
mvrhel_laptop | I will open a bug :) | 17:13.26 |
Robin_Watts | thanks. | 17:13.41 |
mvrhel_laptop | this particular case also has a soft mask | 17:14.05 |
| so it may be that that is the issue for mupdf. not sure | 17:14.16 |
Robin_Watts | You are lost in a maze of twisty blending groups all alike. | 17:15.52 |
| You have been eaten by a non-isolated knockout grue. | 17:16.11 |
ray_laptop | I wonder why "i7" is so much faster. At least peeved caught up to peeves an passed it | 17:19.44 |
mvrhel_laptop | ugh. of course I have to worry about color space differences with a group that may be nested far below since I now allow non-isolated groups with different color spaces | 17:21.39 |
| this is getting painful | 17:22.00 |
| whew. luckily the groups do retain their ICC profile | 17:26.56 |
| so this should be not too bad | 17:27.06 |
ray_laptop | Robin_Watts: cust 532's most recent timing seems to be missing a lot of functions. In particular we don't see the text processing time --it's just buried in the CLIST creation time, but we know there's a lot of text because of the gx_default_copy_mono | 17:55.50 |
| not sure if we can do anything with the text, however. Maybe make sure the glyph cache is optimal and getting a good hit rate. | 17:56.41 |
| bbiab. | 17:57.33 |
Robin_Watts | I have managed to get a (mostly) optimised build out now, so my profiles make sense | 17:57.34 |
mvrhel_laptop | grrr. that did not solve my problem | 18:11.26 |
| time to simplify a bit more | 18:21.01 |
| looks like it is a problem with the softmask and knockout interaction | 18:24.35 |
ray_laptop | Robin_Watts: are you looking at anything in particular on cust 532's stuff (or have you given up on it for the night) ? | 19:27.30 |
Robin_Watts | ray_laptop: I'm looking at my latest profile now. | 19:27.41 |
| The fonts are type42's. | 19:27.47 |
ray_laptop | Robin_Watts: I just don't want to be poking around in the same area | 19:27.48 |
Robin_Watts | hence I don't think ufst is involved. | 19:28.03 |
ray_laptop | Robin_Watts: do we seem to be getting a good glyph cache hit rate ? | 19:28.22 |
| isn't UFST used for TTF's ? | 19:28.48 |
Robin_Watts | You ask questions like that, forgetting that I know nothing about fonts :) | 19:29.23 |
ray_laptop | Robin_Watts: I'll have a look with the debugger... | 19:29.37 |
Robin_Watts | 53% of the time according to this profile is in gs. | 19:30.03 |
ray_laptop | Robin_Watts: if you get some performance numbers you can share, then I'd appreciate it | 19:30.16 |
Robin_Watts | (which means we should mentally double everything). | 19:30.16 |
| ztype42execchar accounts for 3.29% | 19:30.46 |
| and it looks to me that that ends up calling gs_fill (1.96%) | 19:31.25 |
ray_laptop | the performance analyzer (and very sleepy) don't work. VS 2008's analyzer crashes my machine to start rebooting :-( | 19:31.33 |
Robin_Watts | ray_laptop: Ah. Can I mail you a profile somehow? | 19:32.20 |
ray_laptop | very sleepy just refuses to work. (although I haven't tried it in a while) | 19:32.24 |
| Robin_Watts: please do. email is fine | 19:32.34 |
Robin_Watts | Sending.. | 19:33.17 |
| As far as I can see, we handle type42's entirely within gs. | 19:33.39 |
ray_laptop | Robin_Watts: are you starting with qp600.ini ? | 19:34.27 |
Robin_Watts | See Lens email to me about how to start it. I'm in 600dpi 2bit mode. | 19:34.54 |
| Your jpeg stuff seems to be working fine. No color conversions are showing up at least. | 19:35.24 |
| ktd2bit_fast_fill_rectangle gets hit a lot, but I can't immediately see a way to speed that up. | 19:37.51 |
| and we spend a lot of time in gx_remap_DeviceGray. | 19:38.20 |
ray_laptop | Robin_Watts: yes, I tried starting with qp600_2bpp.ini and that didn't work, so I start with qp600.ini and just use the PJL | 19:42.14 |
Robin_Watts | ray_laptop: Oh, it works for me. | 19:42.37 |
| Can you read the profile alright ? | 19:44.14 |
ray_laptop | Robin_Watts: I used the debugger, and ForceGrayOut (in sdctd.c) is set, so that it's using my changes | 19:44.19 |
Robin_Watts | I'm going to be called for dinner in a mo. | 19:44.47 |
ray_laptop | Robin_Watts: what do I open a .vsps with ? | 19:46.18 |
Robin_Watts | ray_laptop: vs2008. | 19:46.30 |
| Sent again as .csv's | 19:48.35 |
| gotta go. will try to check in after dinner. | 19:48.44 |
ray_laptop | Robin_Watts: I got it to open. Thanks. Enjoy your dinner. | 19:49.17 |
| Robin_Watts: for the logs. We *might* be able to squeeze a little out of ktd2bit_fast_fill_rectangle by optimizing color == 0 as well as color == 255 (we wouldn't have to track the tile) | 20:14.40 |
| Robin_Watts: that's what gets used for "copy_mono" painting the runs of pixels | 20:15.13 |
Robin_Watts | yeah. | 20:18.22 |
| Currently ktd2bit_fast_copy_mono goes back into gx_default_copy_mono and thence to kts2bit_fast_fill_rectangle. | 20:19.40 |
| I suspect we might do better to just code that. | 20:20.00 |
| ray_laptop: For the logs: You're right color == 0 would make sense too. | 20:22.46 |
| using memset rather than a while loop could be a good idea too. | 20:23.18 |
| Let me know by email or on here if you look at either of those, otherwise I'll have a fiddle tomorrow. | 20:24.09 |
SpNg | I'm running a postscript program against an EPS file and Ghostscript is only using one of my cores. Is there a way to have GS use multiple cores for this type of task? I found -dNumRenderingThreads but it does not appear to be working for this situation. | 21:20.59 |
Robin_Watts | SpNg: Nope. | 21:27.13 |
| Interpretation is intrinsically single threaded. | 21:27.22 |
SpNg | Robin_Watts: hmmm, so the only option is a faster CPU. | 21:28.35 |
Robin_Watts | indeed. | 21:28.47 |
SpNg | Robin_Watts: ok thanks. | 21:31.53 |
Robin_Watts | ray_laptop: For the logs: I recoded ktd2bit_fast_fill_rectangle to handle both the 255 and 0 cases - and it got slower in my profile. | 22:58.11 |
| Will worry about why in the morning. | 22:58.18 |
| Forward 1 day (to 2013/10/03)>>> | |