| <<<Back 1 day (to 2016/05/15) | 20160516 |
HenryStiles | kens: I saw the new coverity problems for the macro expansion: no good deed unpunished ;-( | 13:16.38 |
kens | Actually I fully expected it | 13:17.24 |
| THere were already those same warnings, but hidden in macros. I figured it would be easier to fix them by sorting the macros out first | 13:17.53 |
HenryStiles | oh good , btw I think you were missing an inline keyword in one of the declarations, I said so in the logs but it was buried in a lot of other stuff. | 13:20.00 |
kens | Chrisl mentioned it, I had spotted it at the time, and carefully kept the email, then threw it away somewhere along the line. If you can remember which one it would be good to know, otherwise I'll just have to look | 13:21.00 |
Robin_Watts | store_sample12 IIRC? | 13:27.03 |
| or sample_store12 ? | 13:27.11 |
HenryStiles | kens: sample_store_next12 line 277 | 13:27.15 |
tor8 | Robin_Watts: okay, I've hooked up the display list images for SVG images in HTML/EPUB | 14:52.48 |
| getting garbage images out :( | 14:52.57 |
| Robin_Watts: nvm, I figured it out. we forgot to clear the pixmap before rendering, and I'd misunderstood the intent of the transform matrix entry | 15:01.36 |
Robin_Watts | tor8: Ah, fab :) | 15:01.44 |
tor8 | now, to figure out how to get it to render at a decent resolution rather than just 72 dpi :) | 15:02.06 |
Robin_Watts | I have the pixmap stride stuff working. | 15:02.30 |
tor8 | nice! | 15:02.37 |
Robin_Watts | and the alpha removal stuff is getting there. | 15:02.47 |
tor8 | Robin_Watts: so... the caching of decoded pixmaps from images is preventing us from re-rendering the svg at the device resolution each time the pixmap is requested | 15:17.13 |
Robin_Watts | tor8: Oh, cos we rescale from the bitmap we have? | 15:17.59 |
| hmm. | 15:18.02 |
tor8 | yeah. | 15:18.05 |
Robin_Watts | So, maybe we shouldn't cache the decoded pixmap for that sort of pixmap. | 15:18.26 |
| Ok, so fz_pixmap has alpha, interpolate and free_samples. | 15:20.07 |
| All of which could be uchars rather than ints. | 15:20.21 |
| (actually, yours won't have alpha). | 15:20.28 |
| We could add an 'uncacheable' one in there too? | 15:20.44 |
| and then in fz_get_pixmap_from_image we skip the last fz_try block if it's set? | 15:21.39 |
| Or rather than fiddling with fz_pixmap, we could just return a flag from the get_pixmap call ? | 15:21.59 |
| which might be nicer. | 15:22.06 |
tor8 | yeah, it feels a bit icky to set that flag there. I'd rather have the flag in the fz_image then | 15:22.25 |
Robin_Watts | tor8: It doesn't need to be an externally visible flag. | 15:22.44 |
| It's just an extra int *cache field to the internal image->get_pixmap function. | 15:23.13 |
tor8 | a dontcache flag in fz_image we set for display list images would work for me | 15:23.20 |
| or an extra out-parameter from image->get_pixmap | 15:23.37 |
| either one | 15:23.43 |
Robin_Watts | the latter is nicer, I think. | 15:23.44 |
tor8 | yes | 15:23.51 |
Robin_Watts | cos the behaviour of the image cache is no ones business but image. | 15:24.04 |
tor8 | another problem is we clamp the requested w,h to image->w,h | 15:25.55 |
| in the subarea stuff | 15:26.03 |
| we can solve that by setting the w,h to be something at 600 dpi and the xres/yres to 600 | 15:26.23 |
| but that also feels a bit hacky | 15:26.29 |
Robin_Watts | Yeah. | 15:26.33 |
| We can have a flag in the image to say "variable size image", and use that to not clamp the image w/h. | 15:27.28 |
| Under such sircumstances the w/h would just indicate the size at the "native" dpi (for aspect ratio purposes) | 15:28.18 |
tor8 | if we do have a 'resolution independent image' flag, we can use that to avoid caching as well | 15:28.19 |
Robin_Watts | yeah. | 15:28.25 |
gregf_ | hello | 15:50.27 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 15:50.27 |
gregf_ | is this channel active please? | 15:50.36 |
| im using ghostscript to convert from pdf to jpeg, but on the production box a conversion takes far too long(approx 20 minutes) | 15:51.34 |
| s/20/30/ | 15:51.42 |
kens | Why do you thnk that's too long ? | 15:51.50 |
| Its going to depend on the size of the PDF< the complexity of the content and the resolution of the output | 15:52.12 |
gregf_ | on my dev box(where gs version is 9.10 ubuntu 4.04 LTS) it takes 1.29 minutes | 15:52.13 |
| kens: sure | 15:52.21 |
kens | THen there's somethign different about the other system | 15:52.25 |
gregf_ | the live box is ubuntu 12.10(gs version 9.06) | 15:52.40 |
| kens: sure, there is a diff in the boxes | 15:53.03 |
kens | Well there's problem 1, you aren't comparing the same versions of Ghostscript, and both of them are out of date | 15:53.04 |
| Current version is 9.19 | 15:53.16 |
gregf_ | kens, well sure | 15:53.17 |
| kens: so another question then :) | 15:53.28 |
kens | Well if it were me, the fist thing I would do is make sure I was comparing apples to apples | 15:53.36 |
gregf_ | is it possible to upgrade to 9.19? | 15:53.37 |
kens | Yes | 15:53.40 |
gregf_ | er, i;ve asked too many questions. was the Yes for the upgrade? | 15:54.05 |
kens | Yes, you can upgrade | 15:54.11 |
gregf_ | *smiles* | 15:54.19 |
chrisl | You probably can't upgrade via apt though | 15:54.33 |
gregf_ | but kens, another question before the upgrade | 15:54.35 |
| yeah, yeah, sure - not via apt-get | 15:54.47 |
| the next question: the docs says under #Improving performance# : -c 30000000 setvmthreshold -f | 15:55.16 |
| that does_not produce the output file | 15:55.30 |
| it runs in 30 seconds, sure, but no output file | 15:55.48 |
| :| | 15:55.50 |
kens | And no error ? | 15:55.55 |
gregf_ | nope | 15:56.09 |
kens | Well if you can reproduce that on current code you can report it as a bug and someone (probably not me) will look at it | 15:56.28 |
chrisl | With "-c" etc the parameter order is important | 15:56.59 |
gregf_ | oh | 15:57.14 |
| well this is the command i'm running: | 15:57.23 |
| time gs -c 30000000 setvmthreshold -f -dUseFastColor=true -dNOGC -dPDFFitPage -dBATCH -dNOPAUSE -dNOPROMPT -dTextAlphaBits=4 -dGraphicsAlphaBits=4 -dUseCIEColor -r150 -sDEVICE=jpeg -sOutputFile=/tmp/vms/1463394710577412/%d.jpeg /tmp/vms/1463394688349499.pdf -c quit | 15:58.08 |
| s/time// | 15:58.13 |
chrisl | Yeh, not going to work | 15:58.17 |
kens | Yeah bad order | 15:58.20 |
gregf_ | oh | 15:58.24 |
kens | Also *don't* use -dUseCIEColor unles you know *exaclty* why you are using it. | 15:58.48 |
gregf_ | please if you could tell the proper order | 15:58.54 |
| ah - sure. i'll take that one off | 15:59.04 |
kens | Move the -c...-f to the and, just before the input file | 15:59.12 |
| drop the -c quite off the end | 15:59.25 |
gregf_ | ok | 15:59.28 |
chrisl | gs -dUseFastColor=true -dNOGC -dPDFFitPage -dBATCH -dNOPAUSE -dNOPROMPT -dTextAlphaBits=4 -dGraphicsAlphaBits=4 -r150 -sDEVICE=jpeg -c "30000000 setvmthreshold" -f /tmp/vms/1463394688349499.pdf | 15:59.31 |
kens | -c quite* | 15:59.32 |
gregf_ | *gives that a whirl* | 15:59.36 |
kens | bah *quit* | 15:59.36 |
gregf_ | sure :) | 16:00.01 |
chrisl | I'd also urge not using -dNOGC - that's a really bad idea....... | 16:00.26 |
kens | There's probably no point in -dTextAlphaBits or -dGraphicsAlphaBits for JPEG output, all that will do is make the output still blurrier IMO | 16:00.47 |
gregf_ | oh | 16:00.57 |
| well, im just doing it by trial and error(from the docs( | 16:01.10 |
| s/(/)/ | 16:01.16 |
kens | But its subjective, and people disagree with me. But I'd try it without | 16:01.21 |
gregf_ | sure, you know better than i do ;) | 16:01.37 |
kens | Try the simplest possible command line is my advice, only add other stuff if you are sure you know why you want it | 16:01.47 |
gregf_ | ok | 16:01.59 |
| so do you suggect just : -dNOGC -dBATCH -dNOPAUSE -dNOPROMPT | 16:02.24 |
| *suggest | 16:02.29 |
kens | do *NOT* use -dNOGC | 16:02.33 |
gregf_ | ok, the docs say otherwise :| | 16:02.47 |
chrisl | *Our* documentation says about -dNOGC: "Useful only for debugging" | 16:02.47 |
kens | That's a debugging only switch, the documentation says so | 16:02.49 |
gregf_ | aha - only debugging | 16:02.54 |
| by, debugging, does it mean the stuff printed to the STDOUT? | 16:03.26 |
kens | If you use -o instead of -sOutptuFile you get all of -dNPAUSE and -dBATCH for free | 16:03.27 |
| No, by debugging it means, well, debugging | 16:03.40 |
| Trying to find and fix faults | 16:03.50 |
gregf_ | well, i'm a developer, so for me debugging is like stepping through/over. *you probably too are a dev* | 16:04.14 |
chrisl | It means if you are debugging the "internals" of Ghostscript | 16:04.14 |
kens | we're all developers here :D | 16:04.45 |
gregf_ | heh | 16:04.49 |
| o/ | 16:05.02 |
| let me try what you've suggested | 16:05.14 |
| also, upgrading it to 9.19 would improve performance? | 16:05.29 |
kens | Maybe, maybe not | 16:05.36 |
gregf_ | hmm, ok | 16:05.50 |
kens | Its hard to remember all the changes over the years | 16:05.53 |
| 9.10 is 2 years old, I can't really remember if there were any performance changes | 16:06.25 |
gregf_ | ok, its surely faster than 9.06 apparently | 16:06.43 |
kens | 9.06 is 3 years old | 16:06.44 |
| There's no reason to expect dramatic improvements in performance unless you hit a corner case and we happen to have worked on it. | 16:07.09 |
chrisl | 9.06 was probably using jasper..... | 16:07.11 |
kens | The code is mature, mostly stable, and doesn't often get startling performance enhancements | 16:07.29 |
gregf_ | hmm, stability is more important :| | 16:07.43 |
kens | Well, if it was using JasPer, and the input has JPEG2000 images, then there will be an improvement | 16:07.52 |
| But 3 years on, I can't remember | 16:08.20 |
| Frankly I have trouble rememberig last Monday, so..... | 16:08.33 |
gregf_ | heh | 16:09.50 |
| ok, i've run this now: time gs -dNOPROMPT -dTextAlphaBits=4 -dGraphicsAlphaBits=4 -r150 -sDEVICE=jpeg -o/tmp/vms/1463394710577412/%d.jpeg -c "30000000 setvmthreshold" -f /tmp/vms/1463394688349499.pdf -c quit | 16:10.45 |
| completed in 36 seconds. which is faster by 1.30 minutes | 16:11.02 |
kens | Almost certinaly down to not converting all colours into CIE colour spaces and back to device space | 16:11.26 |
gregf_ | is that all i need, would: -dUseFastColor=true -dPDFFitPage help? | 16:11.42 |
| hmm | 16:11.44 |
| oh - ok | 16:11.59 |
kens | -dPDFFitPage won't do anything useful unless you set a fixed media, which you aren't doing | 16:12.10 |
gregf_ | sure | 16:12.18 |
kens | You could use -dUseFastColor, at the expense of colour fidelity | 16:12.26 |
gregf_ | and the textAlphaBits.. | 16:12.28 |
| ok | 16:12.46 |
kens | Text and Graphcs AlphaBits are a matter of taste, in general I don;t like the result. When going to JPEG I really don;'t like the result | 16:12.55 |
gregf_ | well, yeah, i dont need them either | 16:13.34 |
kens | I rather doubt the vmthreshols is helpiung, but t might, by reducing memory thrashing. YOu could try taking it off, but it shouldn't cause nay harm | 16:13.42 |
HenryStiles | hmm I just noticed I'll be at the lake district over the holiday, is it very crowded, is the traffic bad? I guess it doesn't much matter now other than to know what I'm in for. | 16:13.59 |
kens | HenryStiles : it will be more crowded over a holiday, is it the late May Bank Holiday you will be there ? | 16:14.22 |
| If so, same rules apply as in the US, go 200yards from the car park and it will be just fine | 16:14.40 |
| But tourist traps will be busier | 16:15.11 |
gregf_ | heh, the US is *huge* ;) | 16:15.12 |
kens | Yeah, but everyone drives everywhere, just like here, so if you walk away from the car park it magically becomes pleasant | 16:15.45 |
gregf_ | well, folks cycle in the UK *ducks* | 16:16.04 |
kens | Sure, but those people will be out whether its a holiday or not | 16:16.21 |
| Again, same as in the US | 16:16.28 |
gregf_ | yeah - true that | 16:16.31 |
kens | On holidays, the tourist places will be busy, but I strongly suspect Henry wants to go running (or walking) | 16:16.58 |
| Or something energetic anyway | 16:17.05 |
Robin_Watts | HenryStiles: If you're lucky, it'll be raining, and no one will dare leave home for the mud. | 16:17.45 |
kens | Ah, the late May Bank Holiday is much more likely to be nice | 16:18.07 |
| OTOH lake district, there is a clue in the name :) | 16:18.19 |
HenryStiles | I guess this year your spring bank holiday is the same day as memorial day, in theory tor8 and sebras should be the only artifex staff members working | 16:18.20 |
kens | I always work statutory holidays | 16:18.35 |
Robin_Watts | Likewise. | 16:18.41 |
| It's not like we can do anything fun (like visit the bank) on those days. | 16:19.05 |
HenryStiles | yes I've noticed that. | 16:19.06 |
kens | HenryStiles : I'm pretty certain it will be good for you no matter what, but avoid the towns and tourist destinations on that weekend if you are at all worried | 16:19.24 |
HenryStiles | heading to Ambleside, I imagine that is a tourist destination. | 16:20.45 |
kens | Um, yeah I would say so | 16:21.10 |
| Pretty though | 16:22.13 |
HenryStiles | kens: I may rent a wet suit and swim, they say it's popular there. | 16:23.32 |
Robin_Watts | HenryStiles: I dunno how easy it is to rent wet suits. | 16:24.07 |
kens | No idea, but I doubt many people will be with you. Best be careful where you swim, there are ferries there | 16:24.14 |
Robin_Watts | When I went to the Lakes (which was admittedly 20+ years ago), I don't recall wet suits being available. We just swam without. | 16:24.53 |
kens | There's a place that renst them | 16:25.07 |
HenryStiles | I got a brochure, I believe from Helen, that advertised it | 16:25.46 |
kens | Capernwray dive centre, but that's not Ambleside obviously | 16:25.48 |
| http://www.mytrievents.co.uk/wet-suit-hire/ | 16:26.10 |
| They do advertise swimming in Lake Windermere | 16:26.19 |
HenryStiles | Robin_Watts: I'd think in that kind of lake you could dip in for a while, but a long swim wouldn't be comfortable. | 16:27.35 |
Robin_Watts | I suspect you're right. | 16:27.46 |
kens | Swim faster to keep warm :) | 16:27.46 |
Robin_Watts | I just wonder if it might be easier to bring your own. | 16:29.54 |
| After all, the weight is nothing compared to the shoes Sabrina will be bringing, right? :) | 16:30.13 |
HenryStiles | Robin_Watts: very good point | 16:30.31 |
gregf_ | oh - he's left :| | 16:35.32 |
| anyways, kens, chrisl : thanks very much for your inputs. much appreciated! | 16:36.02 |
chrisl | gregf_: NP | 16:36.27 |
gregf_ | chrisl: please if you could thank kens too! ;) | 16:38.21 |
chrisl | gregf_: this channel is logged, kens will see it when he reviews the logs tomorrow | 16:39.05 |
gregf_ | cool, cheers! | 16:41.04 |
ferter | Hi everyone! i've faced a stumbling block and i'm wondering if you could help. i'm working with mupdf and what i'm trying to do is to save highlight annotation. everything works just fine, except for one particular thing, - when i open my saved document in foxit reader, instead of blending colors, i see my highlight laying over the text. so i'm assuming i'm making a mistake in my AP. and thus, i seek your help. thanks for your ti | 16:45.06 |
| http://my.jetscreenshot.com/demo/20160516-7pdd-24kb | 16:45.23 |
Robin_Watts | Yeah, that looks wrong. | 17:39.40 |
| Forward 1 day (to 2016/05/17)>>> | |