| <<<Back 1 day (to 2015/05/19) | 20150520 |
mvrhel_laptop | ok that is done. | 00:03.31 |
| s/improper/improper use/ | 00:04.13 |
| Robin_Watts: ok so I got rid of the pdf_clean business | 00:34.20 |
| now to work on this winrt project creation issue | 00:36.57 |
| and I have a bug in gs that I to fix | 00:37.03 |
| done for now though | 00:39.10 |
sebras | tor8: ha! a naïve implementaion of interlaced images was trivial to implement. | 07:16.28 |
| tor8: took all about 15 min, including reading up on the spec. :) | 07:16.53 |
chrisl | I have an appointment at the bank - and being a bank, goodness knows how long it's going to take..... :-( | 07:46.10 |
alexhz | Hello, I am new here. | 11:35.38 |
| I am trying to use mupdf to extract the alt-text that is associated with images in a pdf file. | 11:35.41 |
| So far my idea is to copy and modify the extract tool in mutool. However I cannot find out how to get the alt-text of an image object. | 11:37.00 |
| Any tips? | 11:37.06 |
jogux | hi alexhz, I suspect the people the know the answer to that one aren't at their PCs just now, it may be worth hanging around for a bit (or checking the logs later on) | 11:38.53 |
alexhz | thanks, I'll do that | 11:39.30 |
tor8 | alexhz: PDF doesn't have the concept of alt-text like in HTML; maybe you mean image caption text? | 11:57.29 |
alexhz | then I am using the wrong term. I mean the content that appears in the tooltip when I hover the mouse over an image in Adobe Acrobat. I suppose that is not the caption text either. | 12:00.27 |
tor8 | alexhz: those would probably be annotations | 12:01.40 |
| alexhz: have you got a sample PDF file I could look at? | 12:01.49 |
alexhz | I can make one | 12:02.17 |
| I just created such an pdf. I used Word and then saved as PDF, in Word I added a description "this is my description" to the image: https://www.dropbox.com/s/3lbu8a6hmskdqcx/example.pdf?dl=0 | 12:08.18 |
kens | That's marked content, not annotations | 12:13.15 |
| Specifically its a structure element referring to a marked content object with MICI 1 | 12:14.14 |
| MCID 1 | 12:14.18 |
| You would have to parse the content stream to extract the marked content ID's, then parse the structure tree to find the elements associated with each of the marked content IDs. | 12:15.37 |
alexhz | oh boy, I hoped it would be easy | 12:16.47 |
kens | Its certainly not something easy, no. THs is what the content stream looks like: | 12:17.36 |
| -0.000000495 418.92 595.32 351 re | 12:17.36 |
| W* n | 12:17.36 |
| 451.3 0 0 338.5 72 431.42 cm | 12:17.36 |
| EMC | 12:17.36 |
| Ooops let me try tht again | 12:17.50 |
| "/P <</MCID 1>> BDC q | 12:18.00 |
| "-0.000000495 418.92 595.32 351 re | 12:18.00 |
| "W* n | 12:18.00 |
| "451.3 0 0 338.5 72 431.42 cm | 12:18.00 |
| "/Image9 Do Q | 12:18.00 |
| " EMC | 12:18.00 |
| The " are just to pacify IRC | 12:18.13 |
| The structure tree starts like ths: | 12:18.48 |
| "11 0 obj | 12:18.48 |
| "<< | 12:18.48 |
| " /Type /StructTreeRoot | 12:18.48 |
| " /RoleMap 12 0 R | 12:18.48 |
| " /ParentTree 13 0 R | 12:18.48 |
| " /K [ 14 0 R ] | 12:18.48 |
| " /ParentTreeNextKey 1 | 12:18.49 |
| ">> | 12:18.50 |
| "endobj | 12:18.50 |
| THen: | 12:19.21 |
| "14 0 obj | 12:19.21 |
| "<< | 12:19.21 |
| " /P 11 0 R | 12:19.21 |
| " /S /Part | 12:19.21 |
| " /Type /StructElem | 12:19.21 |
| " /K [ 15 0 R 18 0 R ] | 12:19.22 |
| ">> | 12:19.22 |
| "endobj | 12:19.22 |
| And the bit you want is: | 12:20.00 |
| "18 0 obj | 12:20.00 |
| "<< | 12:20.00 |
| " /P 14 0 R | 12:20.00 |
| " /S /Figure | 12:20.00 |
| " /Alt (This is my description) | 12:20.00 |
| " /Type /StructElem | 12:20.01 |
| " /K [ 1 ] | 12:20.01 |
| " /Pg 3 0 R | 12:20.02 |
| ">> | 12:20.02 |
| "endobj | 12:20.03 |
alexhz | Thanks, that is definately out of my league. Thank you for your help. | 12:22.40 |
kens | You're welcome | 12:22.47 |
tor8 | yeah, we don't support loading the "structured text" structure at all in mupdf, so you'll have to go digging deep in the code to extract that info | 12:26.55 |
alexhz | thanks again | 12:45.02 |
Robin_Watts | chrisl: Your new lappy takes the same batteries as mine, right? | 13:19.11 |
| I'm considering getting one of these delivered to Marcos to pick up at the staff meeting... | 13:19.32 |
| http://www.ebay.co.uk/itm/Battery-for-Lenovo-0C52861-0C52862-121500143-121500144-5200mah-6Cell-/181627794389?pt=LH_DefaultDomain_0&hash=item2a49dc33d5 | 13:19.37 |
jogux | Robin_Watts: I'd be relatively certain that's a Chinese/HongKongese company using a US based drop shipper. | 13:36.15 |
Robin_Watts | An alternative is: http://www.laptop-battery-shop.com/lenovo-45n1126-45n1132-1122v-6600mah-original-batteries-p-4467.html | 14:28.15 |
| That battery actually says Lenovo on it. | 14:28.26 |
kens | So its a better copy :-) | 14:28.40 |
Robin_Watts | yeah. I suspect it's a phoney too. That site gets crap reviews too. | 14:30.14 |
kens | :-( | 14:30.25 |
| I thought I was kidding..... | 14:30.35 |
Robin_Watts | Original battery is $139.99 in the US. £73.20 quid in the UK. | 14:31.10 |
| That's odd, cos the UK price is lower than the US one. | 14:31.36 |
kens | Hmm, quite a bit. On the other hand, if it isn't going to explode..... | 14:31.36 |
| True, that is strange | 14:31.50 |
jogux | automatically distrusts places in HK unless convincingly proved otherwise | 14:33.01 |
| Robin_Watts: 73 inc VAT for a genuine battery? I'd be going for that, no comparison, wouldn't risk some dodgy HK place over a tenner :) | 14:36.59 |
Robin_Watts | yeah. | 14:41.43 |
| Ok, have bought the battery from there. | 14:47.57 |
| s/there/lenovo/ | 14:48.15 |
mvrhel_laptop | aha. There is an issue when using the Output Intent profile and rendering with multiple threads. | 16:02.47 |
| rayjj: are you available to chat for a second | 16:15.38 |
rayjj | when it comes to lenovo batteries, all of them are chinese (not to mention the rest of the laptop). Why be biased against one that ships from HK ? (other than it takes forever) | 16:41.31 |
Robin_Watts | rayjj: The problem is they may not be genuine lenovo ones. | 16:42.15 |
| lenovo ones probaby have a higher standard of reliability/testing than black market clones. | 16:42.32 |
chrisl | The non-lenovo branded ones I've tried have dreadful charge retention, and decidedly shorter lives - I'd always get a properly branded one | 16:43.05 |
mvrhel_laptop | arg. stupid phone call from the indian guy claiming to be from microsoft to help me with my computer that is sending off messages | 16:43.46 |
| I thought they had shut that down | 16:43.54 |
marcosw | chrisl: are you still at the bank or have the let you go home? | 16:44.01 |
chrisl | marcosw: I got away from the bank about 3pm | 16:44.26 |
marcosw | so then straight to the pub to recover and now you are home... | 16:44.54 |
chrisl | Actually, I went for some squash training.... | 16:45.16 |
mvrhel_laptop | I went along with him for a bit. Until he asked me what I saw on my screen. At which point I described a pornographic scene | 16:45.29 |
chrisl | mvrhel_laptop: you should tell him you'd just take the computer along to MS, get them to look at it directly...... | 16:46.13 |
mvrhel_laptop | I will do that next week when they call | 16:46.31 |
chrisl | Luckily, they seem to have stopped calling me | 16:46.59 |
marcosw | there were many regressions in the weekly UFST build with the change to FreeType 2.5.5. Is this reasonable? The comment in the Makefile says freetype will be disabled if UFST_BRIDGE is set. | 16:47.53 |
chrisl | marcosw: I don't think that's true any more - since I couldn't get UFST to work reliably with embedded fonts, we still *have* to use Freetype | 16:48.45 |
| Erm, and in which Makefile does it say that? | 16:49.41 |
marcosw | chrisl: that's what I thought, but the Makefile comment convinced me I was wrong. | 16:49.45 |
| gs/Makefile | 16:49.53 |
| # Define whether to compile in UFST. Note that freetype will/must be disabled. | 16:50.01 |
chrisl | Oh, I never noticed that before - I'll change it | 16:50.19 |
| marcosw: yeh, that comment is nonsense...... | 16:52.26 |
marcosw | thx. btw, this would be a good time for someone (henrys?) to introduce another commit that changes most of the bitmaps since I've just about caught up with regression testing from the FreeType update (and mvrhel_laptop's commit that subtly changed output colors and chrisl's square glyph scaling commit). | 16:54.36 |
mvrhel_laptop | ;) | 16:55.05 |
chrisl | marcosw: so, ideal for me to commit the update to libjpeg 9a, then...... | 16:55.11 |
marcosw | has to stop believing comments and documentation when it disagrees with my understanding. | 16:55.23 |
henrys | marcosw: I'll work on that | 16:55.28 |
Robin_Watts | suggested that the last "MS helpdesk" guy that called should go and become a male prostitute, because it would be a more honest job. | 16:55.54 |
chrisl | marcosw: and there will probably be another Freetype update within the next month! | 16:55.55 |
Robin_Watts | I told him he should follow his mother into the family business. | 16:56.08 |
| He was so mad, he kept phoning me back repeatedly over a period of days to shout at me, | 16:56.28 |
chrisl | I suspect you've just insulted many prostitutes..... | 16:56.50 |
mvrhel_laptop | ha | 16:57.12 |
Robin_Watts | I now have a home phone that screens calls automatically. No more cold calls :) | 16:57.33 |
chrisl | Which one did you go for? | 16:57.49 |
marcosw | chrisl: I was just commenting to Jill that we'd already entered the summer period where customer bugs drop off and I didn't have much to do, so it's nice that you are keeping me busy. | 16:58.14 |
Robin_Watts | http://www.shop.bt.com/products/bt8500-advanced-call-blocker-twin-9M2R.html?utm_source=google&utm_medium=ppc%20product%20search&utm_campaign=PLA%20|%20Shopping%20|%20Top%20Performers&gclid=Cj0KEQjw4fCqBRDM1ZKhk5jfo6IBEiQAZQ97OGudFtVccg8cYee6r6MMKEep_O6AbgGTIP6GMygiPt0aAmBI8P8HAQ | 16:58.21 |
chrisl | marcosw: you should *never* say anything like that - it is a clear invitation for all hell to break loose! | 16:58.53 |
Robin_Watts | chrisl: It's best to replace all the phones in your house with these. I kept my old phone in my office, so I have to learn not to pick it up on the first ring :) | 16:59.20 |
marcosw | Robin_Watts: people in the UK still have land lines? Do you still use fax machines and the royal mail as well? :-) | 16:59.50 |
chrisl | Robin_Watts: many of the ones I get now are from withheld numbers, which makes it a pain | 17:00.20 |
Robin_Watts | chrisl: So, what happens is this... | 17:01.30 |
| phone detects incoming ring, looks at the number. If it's one that's on the authorised list, it goes straight through. | 17:02.02 |
marcosw | in the US we can configure the phone system to refuse all withheld numbers, apparently many people have done this because most of the calls I receive from telemarketers show a phone number (presumably not the one they are actually calling from). | 17:02.34 |
Robin_Watts | If it's not (or it's withheld) it talks to the caller and says "Calls to 'Robin and Helen' are being screened by BT call guardian. Please say your name and press the # key." | 17:02.55 |
| Most cold callers hang up at this point. | 17:03.05 |
| Real callers say their name and press #. | 17:03.15 |
mvrhel_laptop | thats nice | 17:03.28 |
| I would think I should be able to set such a thing up with my phone service which is through the cable company. | 17:04.00 |
marcosw | henrys: i've figured out how to test ghostscript x11 output via Xvfb, did you want me to start regression testing this? | 17:04.03 |
Robin_Watts | Then the phone rings. I pick it up and it says "'mary' is calling. Press 1 to accept the call, 2 to always accept, 3 to always reject" | 17:04.04 |
chrisl | Robin_Watts: Yeh, that's good, it just means it can't track repeated calls. But it's better than what I have, so I think I'll order..... | 17:04.23 |
Robin_Watts | You can also tell the phone to be silent between certain hours, so it will never ring at all for people calling during the night (unless it's one of your VIP numbers). | 17:05.04 |
chrisl | I don't have that problem on my land line these days | 17:05.35 |
Robin_Watts | I did. But only after suggesting the career change to the cold caller | 17:05.55 |
chrisl | I've been wanting something like that for a while, but the other call blocking systems all have a subscription and very variable reviews | 17:06.59 |
Robin_Watts | chrisl: Yeah, me too. This one has no subscription, and works well. | 17:07.23 |
chrisl | Well, thanks for the tip - it's in my amazon basket, will go in with my next order | 17:08.05 |
henrys | any desktop recommendations for a henrysx6 replacement? | 17:10.37 |
Robin_Watts | chrisl: When I looked, the bt shop was cheaper. | 17:11.15 |
| ooh, not now though. | 17:11.44 |
henrys | mvrhel: didn't we want to close 695441? | 17:12.02 |
chrisl | Robin_Watts: yeh, not a massive difference, but less hassle and a slight saving, so..... | 17:12.28 |
Robin_Watts | henrys: looks closed to me. | 17:12.36 |
chrisl | And to me..... | 17:13.02 |
henrys | weird ... I'm reloading in the browser ... | 17:13.51 |
| strange when I killed the tab and revisited it was fine. | 17:14.42 |
mvrhel_laptop | chrisl: Can I give you a new beta for gsview on windows to replace the current one? | 17:15.34 |
| there have a been a few fixes over the last couple weeks | 17:15.48 |
chrisl | mvrhel_laptop: sure, just stick it in your directory on casper, and let me know | 17:16.35 |
Robin_Watts | mvrhel_laptop: Hold on... | 17:16.38 |
mvrhel_laptop | ok. cool | 17:16.42 |
| uhoh | 17:16.44 |
Robin_Watts | Give it to me first, and I'll try signing it. | 17:16.50 |
mvrhel_laptop | oh cool | 17:16.57 |
| don't we need to do that in nsis? | 17:17.06 |
| or can you just sign the installer exe? | 17:17.19 |
chrisl | *can* you do that in nsis?? | 17:17.20 |
mvrhel_laptop | I don't know | 17:17.24 |
| let me give it to Robin first | 17:17.38 |
Robin_Watts | I would imagine that we need to sign the exe, then build it into nsis, then sign the installer ? | 17:17.43 |
chrisl | It's just the installer, I think.... | 17:18.08 |
mvrhel_laptop | yes just the installer | 17:18.13 |
| let me push it up to casper | 17:18.20 |
| ok it is in mvrhel/gsview | 17:19.14 |
marcosw | henrys: You probably don't need to replace henrysx6, we have lots of machines on the cluster, unless you wanted a desktop linux machine for other purposes. If so a 4 core i7 is still the sweet spot in terms of price/performance. e.g.: http://dealnews.com/1354693.html | 17:19.36 |
mvrhel_laptop | so bug 695400 refers to a file on picas.ghostscript.com . I don't seem to have access to that machine | 17:23.26 |
| marcosw: are you in charge of this machine? | 17:23.54 |
Robin_Watts | Unless you play games, you might question whether you need a desktop rather than a laptop these days... | 17:24.42 |
mvrhel_laptop | Robin_Watts: that reminds me | 17:24.53 |
Robin_Watts | mvrhel_laptop: I can get to picas I think. What file do you want? | 17:25.11 |
mvrhel_laptop | My son keeps wanting me to ask you about what is the best graphics card to get for our desktop mahchine | 17:25.16 |
| machine | 17:25.20 |
| 3001Page.pdf | 17:25.32 |
Robin_Watts | mvrhel_laptop: Budget? :) | 17:25.36 |
mvrhel_laptop | in Perf 1.8Gb directory | 17:25.43 |
Robin_Watts | mvrhel_laptop: Ah. I have that here. | 17:25.49 |
mvrhel_laptop | well that is my thought, but no in terms of performance | 17:26.03 |
| he plays a lot of stuff on steam | 17:26.11 |
| and wants to be able to record his stuff | 17:26.31 |
Robin_Watts | OK, presumably it's a PCI-e slot? | 17:26.52 |
mvrhel_laptop | its a small slot on the back if thats what you mean. I am not that literate in hardware as you can tell | 17:27.20 |
Robin_Watts | mvrhel_laptop: No, I meant, "what is your budget?" | 17:27.22 |
mvrhel_laptop | ha | 17:27.26 |
| thats a good question | 17:27.48 |
Robin_Watts | mvrhel_laptop: The motherboard has a slot in it into which the graphics card fits. | 17:27.51 |
mvrhel_laptop | right | 17:28.14 |
Robin_Watts | old motherboards were PCI, then AGP. Yours will almost certainly be PCI-e | 17:28.19 |
| But there are different size PCI-e slots. | 17:28.26 |
mvrhel_laptop | this thing is about 3 years old | 17:28.35 |
Robin_Watts | Do you happen to know the motherboard? | 17:28.43 |
mvrhel_laptop | hold on | 17:28.47 |
| I can find the original invoice | 17:29.02 |
Robin_Watts | mvrhel_laptop: Run cpu-id ? | 17:29.34 |
| http://www.cpuid.com/ <- cpu-z will tell us I think. | 17:30.00 |
mvrhel_laptop | ha. I had typed cupid.com | 17:30.55 |
Robin_Watts | In fact, that's ideal. On the mainboard tab, it'll tell you what the Graphic interface is. | 17:31.13 |
| I have PCI-e x16. | 17:31.23 |
| I have an NVidia 560Ti 2GB thing which was pretty hot when I got it, but that may be 6 years ago now. | 17:32.08 |
| It still runs pretty much every game at 1920x1200 with the graphics settings on high. Many games even with AA too. | 17:32.36 |
mvrhel_laptop | this has a AMD Radeon HD 6450 | 17:32.44 |
| PCI Express 2.1 x16 | 17:33.30 |
Robin_Watts | http://www.tomshardware.co.uk/gaming-graphics-card-review,review-32899-7.html | 17:34.46 |
| That's a useful chart. | 17:34.50 |
| http://gpuboss.com/gpus/Radeon-HD-6450-vs-GeForce-GTX-560-Ti | 17:36.47 |
| You can look at GPU differences on that site too. | 17:36.59 |
mvrhel_laptop | ok. so I can get something with some power for under $250 | 17:37.28 |
Robin_Watts | mvrhel_laptop: Yeah. | 17:37.38 |
mvrhel_laptop | considering that what I have costs $36 | 17:37.47 |
Robin_Watts | ISTR that mine was $500 ish. | 17:37.59 |
| but that was a long time ago. | 17:38.05 |
mvrhel_laptop | ok | 17:38.12 |
Robin_Watts | I had AMD cards for ages, but I defected to NVidia. | 17:38.39 |
mvrhel_laptop | I had a lot of problems with an NVidia driver on a laptop once and that pissed me off enough to avoid them | 17:39.17 |
| of course that was also in the days of vista | 17:39.35 |
Robin_Watts | http://www.amazon.com/MSI-GTX-960-GAMING-4G/dp/B00V4HY522/ref=sr_1_fkmr0_1?ie=UTF8&qid=1432143567&sr=8-1-fkmr0&keywords=MSI+GeForce+GTX+960+Gaming+2048MB+GDDR5+PCI-Express+Graphics+Card | 17:39.44 |
| mvrhel_laptop: Oh, power supply.... | 17:40.04 |
| Modern GPUs are Monsters in terms of the power they need. | 17:40.20 |
mvrhel_laptop | let me find out what this machine has | 17:40.34 |
Robin_Watts | You'll need to check the PSU, yeah. | 17:40.39 |
| Your existing graphics card consumes 18W. At full pelt this one will consume 120W. | 17:42.50 |
| mvrhel_laptop: OK, so i don't have access to picas, it seems. Where did I get this file from them? | 17:48.43 |
| Aha. | 17:48.59 |
| /home/marcos/HCL_TestFiles/Perf 1.8GB/3001Pages.pdf on casper. | 17:49.15 |
mvrhel_laptop | Robin_Watts: ok thanks | 17:51.05 |
| looks like it is a 525W supply | 17:52.32 |
Robin_Watts | mvrhel_laptop: OK, with what CPU? | 17:52.46 |
| cpu-z will tell you again :) | 17:53.10 |
mvrhel_laptop | quad core i7 | 17:53.11 |
| i7 960 | 17:53.23 |
Robin_Watts | yeah, but that's a massively movable feast :) | 17:53.27 |
| ok, so that's 130W for the CPU. | 17:54.33 |
| RAM ? | 17:55.47 |
mvrhel_laptop | 12G | 17:56.22 |
| DDR3 | 17:56.34 |
Robin_Watts | http://www.extreme.outervision.com/psucalculatorlite.jsp | 17:56.54 |
| Might be easier if I just let you fill it in :) | 17:57.45 |
mvrhel_laptop | yes | 17:58.43 |
| thanks | 17:58.45 |
Robin_Watts | My guesswork has you at about 400W or so, so you should be OK. | 17:58.55 |
mvrhel_laptop | yes. it came up with 428 | 18:04.29 |
Robin_Watts | Ok, I'm uploading a signed gsview_setup_6.0.exe to casper. | 18:04.30 |
mvrhel_laptop | Robin_Watts: cool. | 18:04.40 |
Robin_Watts | The Certificate (and lots of exported stuff from it) is in 'Certificates' in my home dir on casper. | 18:05.20 |
| So you should be able to download it. | 18:05.59 |
| and you can sign yourself in future. | 18:06.09 |
mvrhel_laptop | ok. I will check it out. thanks | 18:06.18 |
| Are you pushing it up to the website or will chrisl do that? | 18:06.32 |
Robin_Watts | I would NOT include it in the solution though, unless you can find a way to do it without embedding the password :) | 18:06.34 |
| I think someone should test it first :) | 18:06.42 |
mvrhel_laptop | ok let me uninstall and try | 18:06.53 |
Robin_Watts | My machine may be polluted by knowing about the key. | 18:06.59 |
chrisl | Test? Nah - you wimp! | 18:07.09 |
mvrhel_laptop | Robin_Watts: so did you push the signed gsview installer where the unsigned one was? | 18:09.27 |
Robin_Watts | mvrhel_laptop: No. it's in my homedir. | 18:09.46 |
mvrhel_laptop | ok | 18:09.51 |
Robin_Watts | I can move it if you want. | 18:10.01 |
mvrhel_laptop | no its fine | 18:10.07 |
| got it | 18:10.27 |
| nice! | 18:10.39 |
| very clean | 18:10.49 |
| no warning | 18:10.54 |
Robin_Watts | excellent. | 18:10.59 |
mvrhel_laptop | so chrisl, the one in Robin's home directory is ready to go | 18:11.54 |
chrisl | OKay, will sort it out in a moment..... | 18:12.38 |
mvrhel_laptop | no hurry if you want to wait until tomorrow, its late there | 18:12.51 |
| poor piano tuner has been here 2+ hours. I think the piano was way out of tune | 18:13.22 |
| kids had been complaining about it | 18:13.30 |
chrisl | No, I'm going to be around for a while longer.... after my bank excursion today. | 18:13.31 |
| That should be it live on the website now | 18:15.21 |
mvrhel_laptop | thanks chrisl | 18:15.35 |
chrisl | NP | 18:15.40 |
Robin_Watts | I really wonder if we should be signing the exe too. | 18:20.02 |
| Cos if (say) a virus attacks the exe, it would be nice for people to be able to detect that. | 18:20.26 |
mvrhel_laptop | I had not thought of that | 18:21.46 |
chrisl | I don't think there is any *requirement* to sign the exe | 18:22.28 |
Robin_Watts | chrisl: No. For the purposes of downloading and running the installer (which is what matters most), just signing the installer is fine. | 18:25.25 |
| But I suspect people like ms sign their exes too. | 18:25.57 |
| I wonder if we can sign DLLs. | 18:26.10 |
| yeah. | 18:26.51 |
rayjj | mvrhel: grrr. I was going around and around on the fast threshold stuff not matching the non-fast only to find out that it is due to the fast thresholding not using the transfer function (at least properly) so the data being thresholded was different | 20:23.43 |
| mvrhel: or Robin_Watts: I have a patch that I'd like to commit that fixes bug 695929 (definite) and has some progressions, such as ridt91.eps and 12-07B.PS. It still doesn't match the non-fast thresolding HT tiling code (see the log), but I understand why that is | 23:29.31 |
| mvrhel: Robin_Watts can one (or both) review http://git.ghostscript.com/?p=user/ray/ghostpdl.git;a=commitdiff;h=f59026fe240d317fdd5c36a8705b40f1a252e766 | 23:29.49 |
Robin_Watts | rayjj: Speaking personally, some comments about what num_levels, num_shades etc actually mean would be really useful :) | 23:32.22 |
| I appreciate that it's probably really obvious to you, having spent hours looking at this code recently. | 23:32.49 |
| but every time I have to come back to this code it takes me time to figure out what they all plausibly might be. | 23:33.17 |
rayjj | Robin_Watts: ok, well, maybe we'll just rely on the favorable regression testing :-) | 23:34.35 |
| Robin_Watts: I don't want the make you dig down into it | 23:34.54 |
| (again) | 23:34.58 |
Robin_Watts | no, I'd really like to understand it. | 23:35.04 |
| and that was my first review comment. This code really (IMHO) needs comments. | 23:35.27 |
| Your code looks like a step up from the old stuff though, in that there are some comments, and less voodoo. | 23:35.56 |
rayjj | num_levels and num_shades aren't really changed (just used) by this. They are part of the gx_ht_order (nshades is just calculated from the num_levels * max_value (max_value takes into account 2-bit or greater pixels) | 23:38.36 |
Robin_Watts | All that sentence? Whoosh! straight over my head. | 23:39.12 |
rayjj | Robin_Watts: the voodoo for the t_level_out based on num_levels is still a hack that will go away once I fix the transfer function. Maybe I should comment that 'if' | 23:40.11 |
| Robin_Watts: that also makes the 'off' not needed since we always can use t_level without the hack | 23:41.35 |
Robin_Watts | So max_value is 1 for mono dithers? and 3 for 2 bit dithers? etc? | 23:41.47 |
rayjj | AFAICT the 'off' compensates for the default transfer function that is used when the default 'sethireshalftone' is used | 23:42.36 |
| Robin_Watts: yes, 1-bit max_value=1 2-bit, max_value=3 4-bit, max_value=15 | 23:43.10 |
Robin_Watts | That's the kind of thing I'd like to see in comments :) | 23:43.32 |
| But maybe that's just me showing my ignorance here. | 23:43.45 |
rayjj | you can see the code in gx_render_device_DeviceN that compultes the base 'color' and then the 'l_color' (level) based on the shade. | 23:44.46 |
| color is the base color (eg. for 2 -bit 0, 1, or 2) and l_color is the 'level' above that to be dithered in by setting bits to color+1 | 23:46.04 |
| but the thresholding code doesn't care about 'color' | 23:46.54 |
Robin_Watts | rayjj: I *could* dig around for the info, or it could be in idiot proof comments there. | 23:47.08 |
rayjj | it just does the dithering based on the level | 23:47.11 |
Robin_Watts | (I am happy to be the idiot to test the idiot proofing :) ) | 23:47.22 |
rayjj | Robin_Watts: point taken. I'll add that kind of stuff in | 23:47.37 |
Robin_Watts | Thanks. | 23:47.47 |
| I must go to bed. I'll try and read the code again in the morning. | 23:49.56 |
rayjj | The actual bug fix is to handle the > 256 case without the non-monotonic behaviour that the previous code had. The == 256 case was the only one that worked, but since the default lowres screen and stocht.ps are both == 256 levels, they worked. | 23:51.19 |
| The > 256 level case was fairly rare and is what shows up with the bug and with the few progressions | 23:52.02 |
| I suspect we've never tested with a 16-bit PS threshold array | 23:52.47 |
rayjj | makes a point to make one of those to try | 23:53.11 |
| Forward 1 day (to 2015/05/21)>>> | |