| <<<Back 1 day (to 2011/11/01) | 2011/11/02 |
callaghan | Hello, I have prepared some PostScript files for a presentation of mine. They run interactively (gs abc.ps) and change the content of the preview window. Is there a way to change the size of this window so it has scrollbars? I created it on a higher screen resolution, now on my notebook it's small and it is hard to see anything there. When I maximize it just the small content is duplicated.... | 01:48.59 |
| ...Thanks for any help. | 01:49.01 |
arthurf | tor8: smaller font in the toc looks better, like the title update to the nav bar | 02:04.29 |
| I see something about tweaking scale factors in the shortlog... | 02:05.36 |
tor8 | arthurf: sometimes the page dimensions aren't quite right for scaling perfectly | 02:16.12 |
| so that one scale factor will make the page fit a whole number of pixels in say width, but only a fraction of pixels in the height | 02:16.40 |
| usually not visible, since most pages are white :) | 02:16.58 |
| but on pages which draw exactly to the page border with no bleeding outside the edge, you'll see white borders | 02:18.01 |
arthurf | I updated bug 692644 with results of an iOS test. The PDF that was supplied was albeit for an Android problem, but I was curious how it would work with my iPad - it's a tough file from a performance perspective, and also triggers an intermittent failure. | 02:26.48 |
| goes looking for the white lines - maybe none of my cases draw exactly to the page border. | 02:29.49 |
tor8 | arthurf: you'll need a document which draws a non-white background (like a full page image background or similar) | 02:31.55 |
arthurf | tor8: yep - and that's what I thought I had with edge portions of Leverage_Crash.pdf. | 02:34.49 |
| goes looking around the App Store to see what the comments look like for other PDF viewers. | 03:18.13 |
kens | chrisl, you looked into flights yet ? | 08:12.21 |
chrisl | kens: I need to check with my parents about cat-sitting - they were out all yesterday. | 08:13.35 |
kens | Fair enough. | 08:13.47 |
| It'll be just me going, Melanie doesn't want to miss 4 days of school | 08:14.06 |
| So I was thinking of going on Wednesday and joining Robin and Helen on Thursday | 08:14.26 |
chrisl | I should be okay doing the same. I will probably need to come back on the Sunday as well - I think my mother's taking a class on the Monday | 08:15.08 |
kens | Thought you might want to take some more time, I need to get back really.... | 08:15.50 |
| I'll look at flights this morning since Miles is in a rush now. | 08:16.03 |
chrisl | Well, like I say, I *think* I need to be back here on the Monday. I guess I could travel out Tuesday like Robin and Helen, but until I check with Mum and Dad. | 08:17.38 |
kens | Fair enough, up to you. I'll probably fly out Wednesday and back on Sunday. | 08:18.13 |
| Reading the logs, we do have PaintType 2 fonts in the regression suite | 08:18.42 |
chrisl | Yes, Postscript ones. I don't know about PDF ones? | 08:19.19 |
kens | and pdfwrite does the 'swapcolors' thing itself internally, it doesn't expect the stroke/fill to be correct. It does the swapping for PaintType 2 fonts as well. | 08:19.22 |
| I believe we could make a PDF with PaintType 2 fonts using pdfwrite ;-) | 08:19.55 |
| But I think we already have some (not certain) | 08:20.07 |
chrisl | If pdfwrite does the swapcolors itself, doesn't that cause a problem if something else does it first? | 08:20.39 |
kens | Yes it would. Forunately the PDF itnerpreter doesn't do that (now) when the target device handles rendering modes. | 08:21.11 |
| I changed teh PDF interpreter too ;-) | 08:21.20 |
chrisl | Ah, okay (still hate the interpreter behaving differently for different devices......) | 08:22.07 |
kens | It has to in this case, or we would have to move the decision into the rendering library. | 08:22.28 |
| So it would still behave differently, just at a lower level | 08:22.44 |
| At least it no longer behaves differently depending on the device *name* | 08:23.05 |
chrisl | True. At least it seems that PaintType 2 works correctly in PDF, which is what I wasn't sure about | 08:23.55 |
kens | I think it should. | 08:24.12 |
| Wednesday to Sunday is £519.93 so that's not too bad | 08:24.58 |
| (Virgin) | 08:25.03 |
| 11 am departure, 4pm arrival on Wed. 7pm departure on Sunday, arrive 9 am London | 08:25.36 |
| 10 hours out, 9 hours back | 08:26.01 |
chrisl | I couldn't find a decent way to upgrade premium econ | 08:27.07 |
kens | Miles+money ? | 08:27.21 |
chrisl | I don't think that was available on the dates I was checking - can't remember what dates it was, though | 08:28.16 |
kens | It is for me. | 08:28.33 |
| Just looking it up now | 08:28.41 |
| Ah, no, it says no fligths for those dates | 08:28.50 |
| Maybe they don't do premium economy on those flights | 08:29.15 |
| No Upper Class either... | 08:29.44 |
| OK this time it gives me a calendar. Would have to come back on the 12th | 08:30.31 |
| So I guess Premium is full. | 08:30.38 |
chrisl | Yeh, probably. I'll call my parents in half an hour, and try to pin them down..... | 08:31.20 |
| I need to go feed the cat................. | 08:31.45 |
kens | Yes, even offering full fare gets me no results | 08:31.46 |
chrisl | Geez, what is it about dying e-mail systems in the last month....... | 08:51.46 |
kens | Your email is dead ? | 08:54.57 |
chrisl | spamcop is partially dead | 08:55.28 |
kens | Oh yes, saw it yesterday | 08:55.38 |
chrisl | Yeh, new mail is back, but old mail is missing - not great for those of using IMAP :-( | 08:56.11 |
kens | I pull all mine off and store locally ;-) | 08:56.28 |
chrisl | Well, I *thought* I had Tbird configured to store mail locally. And it does, sort of, but not ideally | 08:57.25 |
kens | Huh, it now offers me the option to upgrade to premium economy, for £249 each way.... | 08:58.48 |
chrisl | But not with miles? | 08:59.11 |
kens | This is the booking section, so money only. Trying ti with miles earlier didn't get me any results. I suspect money wouldn't work either, it looks like a tandard screen ;-) | 08:59.54 |
| OK boooked in. VS0005 and VS006, depart 11am Wednesday 30/11 return 19:15 04/12. Seat 59G out and nothing yet on the way back (there are 4 seats available currently!) | 09:12.26 |
| OK Settled on 62F, middle seat ;-( | 09:13.46 |
chrisl | Busy flight, then :-( | 09:16.29 |
kens | Very, very busy it seems. Holidays I guess. | 09:16.43 |
chrisl | Well, no reply from my parents again, so I'm going to book and hope for the best...... | 09:18.11 |
kens | Well, there's not a lot of time left... | 09:18.33 |
chrisl | Exacly :-( | 09:18.39 |
| Or even "Precisly"..... | 09:18.54 |
| Phew, parents are confirmed for cat sitting Tuesday 29th -> Monday 5th. | 09:57.05 |
kens | Lucky ;-) | 09:57.13 |
chrisl | Well, I was *fairly* sure. The problem is my mother is subbing for Tai-Chi teachers in the area, so takes classes on fairly random days (as well as the regular one on Mondays). She actually does have one on the Thursday, but it's at the Hampshire side of Wiltshire, so not too far from here. | 10:00.06 |
| Hah, amazon have done it again: of the six sci-fi books their e-mail recommends for me, five of them I already bought *from amazon* - doh! | 10:06.56 |
kens | Yeah, I keep getting that too... | 10:07.09 |
chrisl | I used to quite like it, but I'm now wondering how many people blindly accept the recommendation, and end up with multiple copies - interesting marketing ploy..... | 10:08.14 |
kens | Worthy of Apple ;-) | 10:08.38 |
chrisl | :-) | 10:08.55 |
Robin_Watts | I looked at flights on monday night - 410 quid. By last night it was 530. | 10:15.05 |
kens | £520 this morning not including credit card charge | 10:15.37 |
| Be back in a couple of hours | 10:26.08 |
Robin_Watts | hey tor8: What are your plans for Miami ? | 12:55.47 |
tor8 | Robin_Watts: morning. you asking about flights or the trip to everglade? | 12:58.29 |
Robin_Watts | both. | 12:58.37 |
| Scott and Ellen are up for the everglades if timings work out. | 12:58.58 |
tor8 | haven't looked at flights yet, but I'd love to come with you for the everglades | 12:59.13 |
| anything to get out of the resort! | 12:59.18 |
Robin_Watts | Either we can do an organised tour, or we can hire private boats if there are enough of us. Trying to get the idea of numbers? | 12:59.37 |
| Just you? or you + 1 ? | 12:59.44 |
tor8 | still trying to convince the +1 to come, he's not too keen on sunshine and resorts | 13:00.14 |
Robin_Watts | tor8: flights were looking very full for us. I'd recommend you look soon. | 13:02.37 |
tor8 | Robin_Watts: interesting. it's coming up with virgin atlantic on my flight search! | 13:05.18 |
| Robin_Watts: it still looks like I have a lot of options -- but then again, I can fly from basically any airport in europe and it doesn't make much difference | 13:06.49 |
Robin_Watts | tor8: Did you see kammerers comments yesterday? | 13:09.29 |
| he said that the new ARM code stuff doubled the speed of the rendering, but 80%+ of the CPU time was still in the scaling functions. | 13:09.56 |
| At the moment, I use the same code for scaling up and scaling down, even though we only actually use it for scaling down. | 13:10.44 |
tor8 | upscaling is rare enough, I'd hope | 13:11.12 |
Robin_Watts | And the tables I build are designed to be capable of more complex filters (such as the mitchell filter) which have a wider "support" than the simple interpolation filter that we actually use does. | 13:11.36 |
tor8 | but our rendering code can do linear upscaling, so we shouldn't need to even scale the image for magnification | 13:11.37 |
| bilinear sampling, I mean | 13:11.45 |
Robin_Watts | I'm tempted to try a cut down scaler that uses bytes as the intermediate storage rather than ints. | 13:12.26 |
| We'll hit a lot less memory that way, and we don't run into problems with over/underflow because the interpolated filter doesn't ever produce negative values (unlike the mitchell scaler) | 13:13.16 |
| The only downside to it is that we won't be able to cope with scaling down by more than a factor of 256 - but I don't think that's a real issue :) | 13:14.20 |
| It should mean I can do SWAR (or maybe even SIMD) inner loops too. | 13:15.45 |
tor8 | sounds good to me | 13:18.02 |
| unless you feel like messing more with the architecture and allowing non-alpha and 1-bpp monochrome images in fz_pixmap | 13:18.45 |
| but that may be worth splitting fz_pixmap into two data types, one for input images and one for rendering targets | 13:19.10 |
Robin_Watts | I believe that kammerer is using 8bpp greyscaling images. | 13:20.15 |
| Splitting fz_pixmap into 2 versions presents new problems; we sometimes plot from unscaled images (like when we'd be upscaling) | 13:21.00 |
| and sometimes from scaled ones. | 13:21.08 |
| So if we split it, we'd need twice the rendering cores. | 13:21.24 |
tor8 | well, I'd envision all image plotting to come from the fz_image struct which does support more variations | 13:23.07 |
| but all plotting has a the current fz_pixmap as a target | 13:23.34 |
| the scaling could go from a 1bpp grayscale image to 8bpp grayscale | 13:24.04 |
| forget about it if you feel the combinatorial explosion would be too great | 13:24.18 |
Robin_Watts | I'll stick with what I've got for now. | 13:28.40 |
| but it should be possible to take 1bpp into the scalers without a huge problem. | 13:28.57 |
tor8 | allowing non-alpha pixmaps as source images would save a bit of memory, and hopefully not be too invasive a change | 13:29.25 |
Robin_Watts | The scalers cope with that already. | 13:29.48 |
tor8 | yeah, the image painting functions don't though :) | 13:30.02 |
Robin_Watts | Today, I shall just be messing with scalers. | 13:30.13 |
| Hmm. | 13:30.19 |
| draw_simple_scale ? | 13:30.24 |
tor8 | that sounds fine, this has been on the todo list for years now it feels like :) | 13:30.27 |
Robin_Watts | draw_scale2 ? | 13:30.28 |
tor8 | I prefer the first one, I never was fond of tacking on numbers at the end | 13:31.36 |
Robin_Watts | ok. | 13:32.36 |
| Morning Michael. | 13:50.17 |
mvrhel2 | good morning. | 14:05.37 |
henrys | Robin_Watts:I'd like to go but I booked to stay after the meeting. | 14:53.37 |
Robin_Watts | henrys: Ah, Helen had to get back. | 14:54.21 |
mvrhel2 | Robin_Watts: I will probably be interested | 14:54.51 |
Robin_Watts | henrys: Key West looks like too far to go (there and back) in a day, really. | 14:54.54 |
| mvrhel2: You, or you+1 ? | 14:55.00 |
| but Key Largo looks doable. | 14:55.16 |
mvrhel2 | well me plus 1 or plus 3. trying to figure out what to do with the kids | 14:55.18 |
Robin_Watts | mvrhel2: Alligator food! | 14:55.38 |
| :) | 14:55.40 |
henrys | I used to canoe in a place in the south where alligators frequented. They keeps you alert. | 14:56.33 |
mvrhel2 | ha | 14:56.50 |
henrys | oh you might be right about key west.. | 14:57.31 |
| it's all highway though - a bit over 2 hours | 14:58.37 |
| alexcher? | 15:23.43 |
mvrhel_laptop | Robin_Watts: question for you | 16:07.05 |
Robin_Watts | go on. | 16:07.20 |
mvrhel_laptop | I am getting visual studio set up on a new machine. Way Way back when you helped me to get a nice light red line in the editor at around a column number of 80. I thought I had to do something in the registry but I can't remember what it was | 16:08.14 |
| do you recall this? | 16:08.17 |
Robin_Watts | I do. I had to redo it just the other day :) | 16:08.31 |
mvrhel_laptop | awesome | 16:08.37 |
Robin_Watts | http://saraford.net/2004/11/15/guidelinesa-hidden-feature-for-visual-studio-editor/ | 16:09.23 |
mvrhel_laptop | cool. thanks | 16:10.06 |
henrys | mvrhel_laptop:we have quite a few wts segfaults when will your stuff be a suitable replacement? | 16:12.27 |
mvrhel_laptop | henrys: the screens are ready to go now I think | 16:12.47 |
| I did a bunch of fixes last month | 16:12.55 |
henrys | it might be best if you pulled out the wts code. | 16:13.36 |
alexcher | henrys: I'm back | 16:14.01 |
mvrhel_laptop | ok. one thing | 16:14.02 |
henrys | alexcher:since the type 1 project has been take off your plate can you field Tim Waugh's coverity issues and applying the patches or assign them to owners if there is any uncertainty? | 16:14.28 |
| alexcher:just notify Tim to assign all coverity stuff to you. | 16:15.04 |
mvrhel_laptop | the wts code generated screens on the fly. I am not set up to do that. They are generated off line right now. Do we want to go down the route that wts did, which is to have a PS like option to enable and create screens or simply have someone generate the screens off line and then load them as needed | 16:15.29 |
alexcher | henrys: Yes, of course. What to do with the contributed drivers. I cannot test them. | 16:15.41 |
henrys | buffer up the contributed drivers and we'll discuss them as a group. | 16:16.31 |
| I don't think we want to generate screens. | 16:16.45 |
| on the fly | 16:16.53 |
mvrhel_laptop | ok. good | 16:17.14 |
| I will write up a document on the screen generation process then and how to use it | 16:17.40 |
henrys | okay can I assign bugs to you that will be fixed by pulling out the wts code? | 16:18.32 |
mvrhel_laptop | one other question. do we want to work on a device for PCL that does chunky contone until the end and then does color conversion (converting to planar) and then fast threshold? | 16:18.47 |
| henrys: yes that would be fine | 16:18.55 |
alexcher | henrys: What to do if an obviously correct patch creates a few differences in an unrelated rendering at 300 dpi ? | 16:19.23 |
mvrhel_laptop | robin_watts: may be interested in the above too | 16:19.36 |
Robin_Watts | mvrhel_laptop: That seems SOOO much the right thing to do. | 16:19.59 |
mvrhel_laptop | yes. I think we just need to do it to get it out of our systems | 16:20.15 |
henrys | mvrhel_laptop:I think we've folded on that route. Note Robin_Watts' tiffscaled could be repurposed for that kind of thing. | 16:20.25 |
Robin_Watts | If we're doing transparency, then we're doing contone internally anyway. | 16:20.39 |
| The hoops we're having to jump through to do planar stuff with rops etc, are just unbelievable. | 16:21.15 |
mvrhel_laptop | I believe the effort to do the above task is fairly small and it will give us a nice comparison for performance | 16:21.54 |
Robin_Watts | mvrhel_laptop: As henrys says, that's basically what tiffscaled does. | 16:22.11 |
henrys | right we may not want error diffusion though. | 16:22.33 |
mvrhel_laptop | so it does the fast threshold? | 16:22.33 |
| oh. hehe that will be a bit slower | 16:22.50 |
Robin_Watts | henrys: right, so we rip out the error diffusion and replace it with fast threshold. | 16:22.56 |
| or make it an option. | 16:23.13 |
mvrhel_laptop | oh that sounds good. | 16:23.21 |
| does tiffscaled have planar buffers? | 16:23.41 |
Robin_Watts | I knocked tiffscaled up in just a day or so; it's a refreshingly simple thing to do, because you're completely free of the legacy internal gs limitations. | 16:23.56 |
| mvrhel_laptop: Currently, no. | 16:24.26 |
henrys | alexcher:if patches cause problems post to tech for review with a pointer to diffs. We've regressed with coverity fixes in the past lets try to avoid it with review. Further it will give us more awareness about the cluster problems (inderterms) | 16:24.42 |
mvrhel_laptop | so we could just add in the color management of chunky to planar and the thresholding? | 16:24.59 |
Robin_Watts | It just does calls getbits to get buffers of standard chunky RGB (or grayscale), then error diffuses that. | 16:25.45 |
mvrhel_laptop | what does it do for color management now on the RGB? | 16:26.12 |
Robin_Watts | It doesn't :) | 16:26.33 |
mvrhel_laptop | ah. ok. | 16:26.37 |
henrys | note for company 'R' this doesn't fly you have to have the "compression" of halftoning in the pipeline to meet their memory requirements. | 16:26.46 |
mvrhel_laptop | yes | 16:26.51 |
| ok. we will fool with this in the background. I will focus on getting Robin_Watts patch for the early halftone finished | 16:27.30 |
| ok. off to work on this regression that I introduced first | 16:28.03 |
| a fun one that includes clist shading and transparency | 16:28.21 |
henrys | woopee! | 16:28.36 |
| alexcher it might be best to create a branch for simple coverity fixes and commit them all at once instead of having a cluster run for each change | 16:34.05 |
mvrhel_laptop | brb. restarting | 16:37.51 |
henrys | off to shovel snow bbia few minutes | 16:43.27 |
kens | Goodnight everyone. | 17:04.37 |
henrys | bye kens | 17:05.19 |
Robin_Watts | night kens | 17:05.30 |
| tor8: Did the fast bitmap scaling code work for you on iOS ? | 17:41.24 |
| mvrhel_laptop: Did you pass me an xps bug ? | 17:44.11 |
mvrhel_laptop | Hmm. I sent you the file. | 17:44.26 |
| I may have forgotten to assign you the bug | 17:44.33 |
| hold on | 17:44.35 |
| ok yes. it is part of bug 692513 | 17:45.21 |
Robin_Watts | Ah, found the email, thanks. | 17:45.42 |
mvrhel_laptop | did you get the shading simplifed file? | 17:45.45 |
| ok | 17:45.46 |
| it has one triangle that is wacky | 17:46.03 |
Robin_Watts | It's not a triangle that's not being plotted. | 18:08.52 |
| The shading is plotted as 4 'quadrant arcs'. | 18:09.20 |
| And the second one is overlaying the first one in a strange way that means you get this wierd area that looks wrong. | 18:09.48 |
mvrhel2 | ok | 18:48.49 |
Robin_Watts | The code renders each quadrant in turn, in anticlockwise order. | 18:51.04 |
| It picks the starting quadrant so that the last quadrant drawn is the 'major' one (i.e. the one on top) | 18:51.38 |
| A better order would be to do the 'minor' (underneath) one, then the 'left' then the 'right' then the 'major' (on top) one. | 18:52.36 |
| I have something that renders right. | 19:14.23 |
| I'm just scribbling some notes into the file and then I'll clusterpush it. | 19:14.36 |
mvrhel_laptop | great! | 19:19.48 |
| so it was in the xps interpreter which was screwing up the chopping of the shading? | 19:20.15 |
Robin_Watts | No. I believe it was this routine at fault. | 19:20.29 |
| I've changed it to render the 'underneath' tensor patch first, then the two sides, then the top. | 19:20.54 |
| It seems logical to me, but we'll need to wait for a clusterpush to see. | 19:21.08 |
mvrhel_laptop | ok. I tracked down the source of my regression. need to grab some lunch then I will try to fix it and test | 19:22.18 |
henrys | Robin_Watts:you made 2 color spaces but it looks like we can only support 1 color space and 2 colors. Is that right? | 23:10.42 |
| oh no it looks like it should work - nvm | 23:13.33 |
mvrhel_laptop | wow. the new ubuntu is weird. Even I am lost in gui hell | 23:41.46 |
| oh all kinds ok strange keyboard shortcuts that I was not aware of | 23:48.37 |
| s/ok/of/ | 23:48.45 |
| Forward 1 day (to 2011/11/03)>>> | |