| <<<Back 1 day (to 2012/10/11) | 2012/10/12 |
vtorri | hey | 09:19.15 |
ghostbot | que tal, vtorri | 09:19.15 |
vtorri | does the sumatrPDF lurk here sometimes ? | 09:19.38 |
Robin_Watts | vtorri: no. | 09:21.31 |
| I've never known zeniko to log in here. | 09:21.40 |
vtorri | ok | 09:22.42 |
| so mail is the best solution | 09:22.58 |
| it's for epub support | 09:23.49 |
| zeniko == Krzysztof Kowalczyk ? | 09:25.33 |
chrisl | Robin_Watts: if you want to give it a try before I push to the main repo: http://git.ghostscript.com/?p=user/chrisl/ghostpdl.git;a=commit;h=ba731ff7 | 09:27.29 |
Robin_Watts | vtorri: No, zeniko = Simon | 09:35.35 |
sebras | vtorri: no, zeniko is a different guy from kowalczyk. | 09:35.49 |
| vtorri: send a mail to zeniko, he usually responds within a day. | 09:36.18 |
Robin_Watts | chrisl: Looks good. Let me pull it to peeves and try it out. | 09:36.51 |
| Still seems to be trying to build gdevmag | 09:46.19 |
chrisl | Oh, maybe I missed the MISC_DEVS list.... | 09:47.17 |
Robin_Watts | I noticed yesterday when updating the msvc files that the list of default devices is very different for pcl etc and gs. | 09:48.30 |
chrisl | The PDL builds only use a small subset of the devices | 09:48.57 |
| Robin_Watts: revised version in my repo | 09:49.28 |
Robin_Watts | I tweaked the VS projects/makefiles a bit yesterday, so there there are 'Debug-contrib' and 'Release-contrib' configurations that build with all the contrib devices in. | 09:50.09 |
| OK. With mag16 and mag256 added, that works perfectly. | 09:50.27 |
chrisl | Having all the contrib devices in the Windows build seems a bit redundant - they're mostly for printers that have "real" drivers on Windows | 09:50.59 |
Robin_Watts | Right, but it's a much better environment to debug on :) | 09:51.34 |
chrisl | <cough><splutter> that's a matter of opinion! | 09:52.03 |
Robin_Watts | Same reason as for having Debug-cups. | 09:52.04 |
chrisl | That's not quite the reason for having cups in there....... | 09:52.43 |
| Okay, pushed to the main repo...... | 09:56.40 |
Robin_Watts | chrisl: Crumbs. | 11:27.20 |
chrisl | Robin_Watts: huh? | 11:27.34 |
Robin_Watts | Turns out that it was easy to remove all those gs_lib_ctx_get_non_gc_memory_t calls after all. | 11:27.58 |
| so it's 'just' the global usage we need to sort. | 11:28.38 |
chrisl | In some cases, there was a mem pointer "near by", but I assume you means something else? | 11:28.53 |
Robin_Watts | no, I mean that in all but 2 cases, there was a mem pointer available in the functions alread. | 11:29.16 |
| already. | 11:29.18 |
| and those 2 functions just required it to be passed in. | 11:29.28 |
| i'd thought there was going to be much more fiddling required. | 11:29.38 |
chrisl | Ah, I thought there was more than just the two - I only took a quick look though. Sounds good :-) | 11:30.20 |
oy | tkamppeter, your coming is much appreciated :-) | 11:40.35 |
tkamppeter | oy, thanks for making the sponsoring possible. | 11:59.03 |
vtorri | Robin_Watts, sebras : thanks | 12:44.32 |
kens | chrisl Robin_Watts got a minute ? | 13:28.48 |
chrisl_r61 | kens: sure | 13:31.30 |
Robin_Watts | back, sorry, sure. | 13:31.44 |
kens | Having trouble with msys git and ssh/cluster | 13:31.48 |
Robin_Watts | Are you using plink or ssh ? | 13:32.04 |
kens | First thing is how do I export GIT_SSH so that msys will remember the environment variablke ? | 13:32.15 |
| Robin_Watts : plink | 13:32.21 |
| I have GIT_SSH=plink.exe | 13:32.31 |
Robin_Watts | You edit a file... let me try and figure out which one :) | 13:32.40 |
kens | ssh casper.ghostscript.com gives me 'permission denied (publickey) | 13:32.57 |
chrisl_r61 | It will be either .profile or .basshrc in your home directory | 13:33.06 |
kens | But PuTTY is fine doign the same thing | 13:33.06 |
Robin_Watts | plink casper.ghostscrip.com | 13:33.23 |
kens | chrisl : Aha, Let me look at .basshrc | 13:33.25 |
Robin_Watts | C:/Users/Robin Watts/.bashrc is what I have | 13:33.48 |
kens | Robin_Watts : plink casper.ghostscript.com works OK | 13:33.49 |
Robin_Watts | kens: Fab. | 13:33.54 |
kens | But git pull --rebase does not | 13:34.01 |
chrisl_r61 | kens: with ssh you need to specify a user name - ssh ken@casper...... | 13:34.03 |
kens | nor does a straight ssh | 13:34.09 |
Robin_Watts | kens: Forget ssh, stick to plink for everything. | 13:34.21 |
kens | Hmm I tried -l ken, let me try the other way | 13:34.23 |
Robin_Watts | msys git ssh is broken on windows sadly, due to permissions problems. | 13:34.47 |
kens | ssh ken@casper.ghostscript.com fails with the publickey error | 13:34.54 |
| Robin_Watts : I thought it worked on my other PC just a moment | 13:35.10 |
| Ah, no it doesn't, my mistake | 13:35.36 |
Robin_Watts | Maybe you've replaced the ssh within msys with a cygwin one. That works, apparently, though I don't trust it. | 13:35.38 |
| ok, so: git remote -v and paste the results please. | 13:36.10 |
chrisl_r61 | kens: so it looks like ssh is not finding your key file | 13:36.11 |
kens | Robin_Watts : no it doesn't work on either, so that's OK | 13:36.21 |
| chrisl : not worried about ssh now, onl;y about Git ;-) | 13:36.36 |
chrisl_r61 | Well, as git needs ssh of some kind...... | 13:37.00 |
kens | git pull --rebase says 'connection abandoned. fatal: The remote end hung up unexpectedly" | 13:37.14 |
Robin_Watts | chrisl_r61: That's why he has export GIT_SSH=plink.exe in his .bashrc | 13:37.23 |
chrisl_r61 | Robin_Watts: right | 13:37.42 |
kens | $ git remote -v | 13:37.45 |
| casper ken@ghostscript.com:/home/ken/repos/ghostpdl.git (fetch) | 13:37.45 |
| casper ken@ghostscript.com:/home/ken/repos/ghostpdl.git (push) | 13:37.45 |
| cluster regression@ghostscript.com:/home/regression/cluster/gitbridge/ghostpdl ( | 13:37.45 |
| fetch) | 13:37.45 |
| cluster regression@ghostscript.com:/home/regression/cluster/gitbridge/ghostpdl ( | 13:37.45 |
| push) | 13:37.46 |
| origin ken@ghostscript.com:/home/git/ghostpdl.git (fetch) | 13:37.46 |
| origin ken@ghostscript.com:/home/git/ghostpdl.git (push) | 13:37.47 |
Robin_Watts | OK, so try: plink ken@ghostscript.com | 13:38.03 |
chrisl_r61 | Do you perhaps need the full patch to plink? | 13:38.19 |
Robin_Watts | chrisl_r61: Not if it's on his path. | 13:38.30 |
| git is finding plink successfully I believe. | 13:38.50 |
kens | chrisl : plink is in the path (an earlier fix) | 13:38.52 |
chrisl_r61 | Okay | 13:39.00 |
kens | Robin_Watts : I had to agree to store the key in the cache, but plink ken@ghostscript.com works fine | 13:39.12 |
Robin_Watts | Now try git pull --rebase again. | 13:39.23 |
kens | Aha! | 13:39.56 |
| Now its OK | 13:40.00 |
| OK so all I need to do now is edit .bashrc | 13:40.16 |
Robin_Watts | cool. I had to go through this a while ago while setting up git on helens machine. | 13:40.18 |
kens | Robin_Watts : I knew there was stuff I hadn't done, but I couldn't for the life of me remember what | 13:40.36 |
| It didn't help that I was (wrongly) sure that ssh should work.... | 13:40.50 |
Robin_Watts | S'OK. Between us we have most of a functioning memory :) | 13:40.53 |
chrisl_r61 | So, git doesn't like the ssh delegate requiring interaction..... | 13:40.55 |
kens | It seems like it | 13:42.24 |
| Great, that seems to fix everything, I think I'm about ready to turn off the old PC and just use this new one now | 13:45.14 |
Robin_Watts | kens: Do you use Console 2 ? | 13:45.52 |
kens | No, because I have no idea what it is :-) | 13:46.05 |
Robin_Watts | It's a better windows command window. | 13:46.19 |
| Still runs cmd.exe (so exactly the same behaviour as before in that regard) | 13:46.43 |
kens | What's different then ? | 13:46.54 |
Robin_Watts | but supports tabs (so I can have cmd in one, an msys git in another in the GhostPDL directory, another msys git in the MuPDF directory, a cygwin,. ...) | 13:47.19 |
| And has sane cut and paste. | 13:47.30 |
kens | Well, I'm no that bothered, I mostly launch msys whenI want it and then close it down | 13:47.57 |
Robin_Watts | Ah, I run msys all day long; I do all my work from it. | 13:48.12 |
kens | Whatever suits best.... | 13:48.26 |
| OK still to install Acrobat 9 and X | 13:48.37 |
Robin_Watts | configure: WARNING: Due to the --enable-threadsafe option, the following devices have been excluded because they are not threadsafe: | 14:16.37 |
| x11alpha x11 x11 x11 x11 x11 x11 x11 x11 x11 x11 x11 x11 x11 hpdj1120c hpdj890c hpdj870c hpdj855c hpdj850c hpdj690c hpdj680c hpdj670c hpdj660c hpdj600 hpdj560c hpdj550c hpdj540 hpdj520 hpdj510 hpdj500c hpdj500 hpdj500 hpdj400 hpdj340 hpdj320 hpdj310 hpdjportable hpdjplus pcl3 uniprint stcolor opvp lxm3200 fmlbp cdj970 cp50 bjccolor bjccmyk bjcgray bjcmono | 14:16.39 |
| chrisl: It really thinks x11 is unsafe :) | 14:16.52 |
chrisl_r61 | That's confusing..... | 14:17.30 |
Robin_Watts | I've changed the list :) | 14:17.39 |
chrisl_r61 | Still, each device should only be listed once | 14:17.59 |
Robin_Watts | yeah, I figured. | 14:18.32 |
chrisl_r61 | The "NTS_EXCLUDES=`echo $NTS_EXCLUDES | uniq`" line should prune any duplicates | 14:19.23 |
Robin_Watts | http://git.ghostscript.com/?p=user/robin/ghostpdl.git/.git;a=commitdiff;h=6bc255daf1bdd196ec4e41fd6f3f19e22bf78f12 | 14:19.29 |
chrisl_r61 | I'll take a look at it..... | 14:22.21 |
Robin_Watts | Do we ever use the LogLuv stuff in libtiff? | 14:37.48 |
| (For high dynamic range images) | 14:37.55 |
chrisl_r61 | I'm not aware that we explicitly use it, no | 14:38.45 |
Robin_Watts | Cos it's not threadsafe. | 14:39.02 |
| If we don't use it it doesn't matter. | 14:39.41 |
chrisl_r61 | libtiff has some vestigaes of non-thread safety. | 14:39.52 |
Robin_Watts | We could change to explicitly disabling it when building libtif. | 14:39.55 |
chrisl_r61 | Yeh, wouldn't be a big deal. | 14:40.21 |
| Robin_Watts: http://git.ghostscript.com/?p=user/chrisl/ghostpdl.git;a=commitdiff;h=c9cfaca4 | 14:40.24 |
Robin_Watts | Want me to roll that into my patch? Or just test it? | 14:40.53 |
chrisl_r61 | Up to you..... | 14:41.17 |
Robin_Watts | I'll just test it. | 14:41.31 |
chrisl_r61 | In terms of what ends up in the device lists, it's no different, just the warning has changed | 14:42.00 |
Robin_Watts | ok. | 14:42.21 |
| chrisl_r61: Urm... | 14:49.18 |
| ./configure: command substitution: line 8122: syntax error near unexpected token `)' | 14:49.28 |
| ./configure: command substitution: line 8122: `echo "$NTS_EXCLUDES" | tr " " "\n" | sort | uniq | tr "\n" " ")' | 14:49.30 |
chrisl_r61 | Oh, sorry..... | 14:50.12 |
Robin_Watts | Testing removing just the ) now. | 14:50.46 |
chrisl_r61 | Yeh, that should work | 14:51.34 |
Robin_Watts | That does indeed work. | 14:51.46 |
| Thanks. | 14:51.54 |
| ooh, an mvrhel_laptop | 14:52.06 |
| I remembered something to speak to you about earlier... and now I've forgotten it again. | 14:52.22 |
chrisl_r61 | Robin_Watts: do you want me to push that to the main repo? | 14:52.29 |
mvrhel_laptop | good morning | 14:52.32 |
Robin_Watts | chrisl_r61: Would make sense. | 14:52.38 |
| Morning | 14:52.43 |
mvrhel_laptop | Robin_Watts: I have to run the kids to school and will be back on shortly after that | 14:53.24 |
Robin_Watts | mvrhel_laptop: No worries. It could be weeks before it surfaces again. | 14:53.40 |
mvrhel_laptop | ok. | 14:53.53 |
| hopefully I can get this overprint stuff working today then work on documenting all the color changes this weekeend | 14:54.24 |
Robin_Watts | cool. | 14:54.45 |
mvrhel_laptop | then get customer 330 working with the code | 14:54.48 |
Robin_Watts | Who 'owns' the X device? Is that you chrisl? | 14:54.56 |
| mvrhel_laptop: say it quick and it sounds easy :) | 14:55.36 |
henrys | I have x11 but it scares me. | 14:57.26 |
| for x11alpha michael owns the alpha part. | 14:58.40 |
Robin_Watts | I was going to ask about x_error_handler in gdevxini.c as it looks not to be threadsafe. | 14:59.11 |
| but then I read the comment just before it, where it says "this isn't threadsafe". | 14:59.24 |
| :) | 14:59.27 |
mvrhel_laptop | you own that part Robin_Watts | 14:59.33 |
| ;) | 14:59.56 |
henrys | right you own x11threadunsafe | 14:59.59 |
Robin_Watts | mvrhel_laptop: Gee, thanks :) | 15:00.03 |
mvrhel_laptop | bbiaw | 15:00.25 |
henrys | Robin_Watts:but x11 itself must run in separate processes right? | 15:01.28 |
Robin_Watts | I'm still trying to understand the issue. | 15:02.06 |
| If we have multiple threads each calling XSetErrorHandler, there is no way for the callback that comes in to know what thread the error was called by, I guess. | 15:03.29 |
| ok, essentially, nothing Xlibby can be thread safe. http://stackoverflow.com/questions/6096605/is-handler-set-via-xseterrorhandler-global-or-threadlocal | 15:04.37 |
henrys | well there is an xinitthreads() call but we don't use it so I don't thing it will work | 15:04.44 |
Robin_Watts | googles for that. | 15:04.56 |
| ok, we'd need to lock/unlock around every Xlib call to use that. | 15:06.20 |
chrisl | That probably isn't too horrific given how the x11 devices work | 15:07.03 |
Robin_Watts | and presumably we'd only get error callbacks while locked, so we'd know which one caused it. | 15:07.04 |
| Yeah, but I think it can stay a thread unsafe device. | 15:07.30 |
| I mean, I don't really see print servers wanting to use the standard X device. | 15:07.55 |
chrisl | No, but god knows, some fool will want a multithreaded viewer, or something! | 15:08.23 |
Robin_Watts | We can burn that train when it sails. | 15:08.50 |
henrys | with out crappy X11 display one is more than enough | 15:09.18 |
| s/out/our | 15:09.22 |
chrisl | The trouble is, if the distros want to build in x11, our Linux libraries can never be thread safe. | 15:10.58 |
Robin_Watts | Well, we can enable the device, and people will only ever hit problems if they try and use more than 1 x display at once. | 15:15.15 |
henrys | Robin_Watts:yeah you're right sons of anarchy has improved just started the 3rd season. | 15:15.26 |
Robin_Watts | has just started Justified. | 15:15.40 |
henrys | now we have dexter, homeland and sons I'll never fix my bugs | 15:17.00 |
Robin_Watts | Homeland S2 has started here too. We have to wait ages for dexter and SoA. | 15:19.16 |
| The Internet has had a great effect on TV though; because of torrents they generally try and air shows as soon as possible here now. | 15:19.46 |
henrys | yes I would expect people to download it illegally if it isn't even available in some commercial form. | 15:21.43 |
| Robin_Watts:so you can't subscribe to showtime? | 15:23.56 |
Robin_Watts | showtime isn't available in the UK. | 15:24.25 |
| They sell their shows to various other channels. | 15:24.39 |
| Dexter is shown on FX (which also shows NCIS, The Walking Dead etc) | 15:25.10 |
| FX is a satellite only channel. | 15:25.23 |
| And about a year after they show stuff on FX, they come to terrestrial TV, generally. | 15:25.54 |
| Most of HBOs drama stuff is shown on Sky Atlantic (again, a satellite only channel) | 15:26.33 |
vtorri | what are the new features in the current mupdf repo (compared to mupdf 1.1) ? | 15:50.46 |
Robin_Watts | vtorri: Forms. | 15:52.10 |
| (and misc bugfixes) | 15:52.21 |
vtorri | web forms ? | 15:53.05 |
Robin_Watts | PDF forms. | 15:53.44 |
| (and associated javascript) | 15:53.49 |
vtorri | which javascript engine do you use ? | 15:55.21 |
| V8 ? | 15:55.26 |
Robin_Watts | The technology is designed to be engine agnostic, but the current implementation is for v8, yes. | 15:55.52 |
vtorri | ok | 15:56.12 |
Robin_Watts | vtorri: The forms stuff is still under rapid development. | 15:58.53 |
| but we're approaching the point of doing an alpha release. | 15:59.04 |
vtorri | ok | 16:01.25 |
| about graphisms for, for example, the combo box, is it also agnostic to any graphic engine ? | 16:01.53 |
Robin_Watts | vtorri: All such rendering is done using the mupdf lib. | 16:02.15 |
vtorri | ok | 16:02.30 |
Robin_Watts | (i.e. we generate appearance streams for them as PDF operators, and we then draw them like that) | 16:02.39 |
vtorri | i would really like to see a screenshot :) | 16:02.51 |
Robin_Watts | vtorri: Do you have an android device? :) | 16:03.26 |
vtorri | no :) | 16:03.42 |
Robin_Watts | Well, it works under windows/linux too. | 16:03.58 |
vtorri | i have to compile mupdf :) | 16:04.10 |
| no time these days | 16:04.16 |
| we have a lot of work in oiur project | 16:04.28 |
Robin_Watts | marcosw: This customer that is building on AIX and finding problems with the contrib dir being empty... | 16:14.33 |
| presumably he's building from the commercial version which has had all the contrib stuff ripped out? | 16:15.02 |
chrisl | Robin_Watts: the current configure script detects the missing contrib contents, and disables those devices | 16:15.55 |
Robin_Watts | I was wondering if he was building from an out of date makefile for exactly that reason. | 16:16.17 |
chrisl | quite possible | 16:16.37 |
ray_laptop | we haven't shipped a complete 'contrib' directory in the commercial version for several years (once a customer asked us about the GPL drivers and we realized the problem) | 16:31.32 |
| but he said he ran the configure, so maybe something is wrong with it ? | 16:33.52 |
| (I assume your comments were about Oana's email | 16:34.46 |
| I said 'he' but can't really tell from the name | 16:35.27 |
chrisl | ray_laptop: I just tested on Linux, and configure properly drops the contrib devices | 16:35.47 |
ray_laptop | chrisl: yeah. I can't imagine that the AIX shell wouldn't work. | 16:36.27 |
Robin_Watts | Unless he specifically requested that device in the configure step? Can you do that? | 16:37.39 |
chrisl | That might cause a problem, but then it's not our problem! | 16:38.14 |
Robin_Watts | BUGGER. | 17:05.50 |
| I pushed the windows unicode commit. | 17:06.00 |
| I'm gonna force push that out. | 17:06.26 |
| Nobody saw that. It didn't happen. These are not the droids you are looking for. | 17:08.38 |
kens | Heading off, night all | 17:14.26 |
ray_laptop | Robin_Watts: on the new website... Has the speed/responsiveness problem been resolved ? | 17:14.52 |
Robin_Watts | no. | 17:15.00 |
| and unless we change to a different virtual host, I don't believe it will be. | 17:15.16 |
ray_laptop | Robin_Watts: OK. So, can we do that ? What's involved ? | 17:15.35 |
Robin_Watts | ray_laptop: Urm... Well, we'd need to open a new godaddy account. | 17:17.14 |
| Does anyone actually find it that big of a deal? | 17:17.51 |
| Certainly I don't find the site to be too slow. | 17:18.56 |
henrys | ray_laptop:I have about as much understanding of what Miles is saying as he does. Can you see his spreadsheet? | 17:23.26 |
chrisl | Robin_Watts: I do notice it being slower than I expect, but it's not unusable. I find it surprising because the downloads are fast enough | 17:24.40 |
ray_laptop | henrys: Miles' spreadsheet ? | 17:24.51 |
Robin_Watts | chrisl: Downloads are just shifting bulk data. | 17:24.58 |
| To generate the pages, it's database access etc. | 17:25.17 |
chrisl | Oh, I didn't think that stuff was in place yet | 17:25.39 |
henrys | ray_laptop:he sent a message to us saying remove Jack's access from the spreadsheet, did you see that email? | 17:25.41 |
ray_laptop | henrys: yes, but I don't know where that is ? | 17:26.33 |
| chrisl: yes, that's the point of the new website | 17:26.56 |
| henrys: I thought you were the one that got Scott and Jack up with a spreadsheet somewhere (google ?) | 17:27.46 |
henrys | no I just told them to use it and never heard back. | 17:28.11 |
chrisl | I gotta head out - enjoy your weekends! | 17:29.08 |
Robin_Watts | you too chrisl. | 17:29.08 |
chrisl | Cheers | 17:29.08 |
henrys | ray_laptop:I responded to the email, I have to think Scott was letting him use his password. | 17:31.37 |
marcosw | Robin_Watts: if the commercial version isn't compiling under AIX does that mean it's also not compiling on other UNIX like operating systems? | 17:32.18 |
Robin_Watts | marcosw: We discussed this a bit earlier. | 17:32.58 |
marcosw | earlier today or a lot earlier? | 17:33.08 |
Robin_Watts | Earlier today. | 17:33.15 |
| Just after I brought the subject up and you fell off the net. | 17:33.26 |
| The configure script should automatically exclude the devices when it finds contrib empty. | 17:33.38 |
marcosw | I guess I look at the logs then. | 17:33.57 |
Robin_Watts | The best guess we could come up with is that he's specifically requesting that device in the configure invocation. | 17:34.11 |
marcosw | okay, I'll clarify what options they are using and request the configure output files (and Makefile) | 17:35.21 |
henrys | marcosw:you are still looking at Elizabeth's issue right? | 17:38.06 |
marcosw | henrys: I'm supposed to be, have been busy with other (non-artifex) stuff the last couple of days. | 17:39.08 |
henrys | marcosw:just checking you know the ball's in your court. | 17:39.54 |
marcosw | henrys: sometimes it seems that all the balls are in my court. | 17:40.37 |
henrys | I think I have more customer bugs than everyone put together - so I must have a few balls too. | 17:42.14 |
Robin_Watts | Why am I not getting the lxm3200 device in my final build? | 18:23.46 |
| oh, I see, maybe. | 18:25.09 |
ray_laptop | Robin_Watts: you were the last person to touch the image interpolation stuff. | 19:25.29 |
Robin_Watts | I probably was, yes. | 19:25.44 |
ray_laptop | Robin_Watts: what do you think of having an integer scale factor so that we only interpolate to a NxN device pixels ? | 19:26.58 |
Robin_Watts | ray_laptop: I'm not convinced it'll look good on contone devices. | 19:27.52 |
| I can possibly see that it might be worthwhile on dithered output. | 19:28.12 |
ray_laptop | Robin_Watts: this would be optional (device param) for dithered output | 19:28.31 |
Robin_Watts | but if it's to be done, it should be done outside of the interpolation code. | 19:28.34 |
| we should have another image enumerator thing that calls the interpolation classes, and then expands the results. | 19:29.13 |
| If you see what I mean. | 19:29.17 |
| The interpolation stuff is quite hairy enough already. | 19:29.29 |
ray_laptop | Robin_Watts: OK. | 19:32.57 |
Robin_Watts | ray_laptop: Actually, that may not be easy to do at the enumerated image type level. | 19:52.23 |
| Within the interpolation image type we handle the interpolation as a stream. | 19:53.37 |
| It may be possible to do a stream that wraps the interpolation processing stream that pixel doubles (or triples etc) the output. | 19:54.18 |
ray_laptop | Robin_Watts: I'm thinking that maybe finagling gs_image_class_0_interpolate and image_render_interpolate to do the N-up may not be too bad | 19:54.24 |
Robin_Watts | Ok. So that keeps the complexity out of the central interpolation processing code. | 19:55.04 |
ray_laptop | i.e. the interpolation code just gets a different WidthOut HeightOut | 19:55.48 |
Robin_Watts | Yes, so the outside code drives the core interpolator differently. No changes to the core processing. | 19:56.14 |
ray_laptop | then when we do the LINE_ACCUM_COPY we expand it | 19:56.23 |
| Robin_Watts: right -- the core code is pretty hideous | 19:56.40 |
Robin_Watts | We have a bug open for the lack of interpolation in rotated images. I have code that almost works. I should revisit that and try to solve the clist problems. | 19:56.51 |
| The core code is not hideous. It's just complex. | 19:57.01 |
| And it uses far more fp than is good for it. | 19:57.27 |
ray_laptop | hideous as in complex, not hideous as in badly written (hideous is in the eye of the beholder) | 19:57.50 |
Robin_Watts | actually, I take that back. It's far from the nicest implementation of this algorithm I've seen. | 19:57.55 |
| Forward 1 day (to 2012/10/13)>>> | |