| <<<Back 1 day (to 2011/07/12) | 2011/07/13 |
henrys | just leavin' were you able to reproduce the problem with the bbox? | 00:14.54 |
mvrhel2 | henrys: yes | 00:15.56 |
| just ran it now | 00:16.00 |
| all sorts of ugly complaints | 00:16.06 |
| I will see what is going wrong | 00:16.15 |
henrys | okay I don't know if that will address the X11 problems also, I hope so. | 00:16.33 |
mvrhel2 | I will check that after I figure this out | 00:16.46 |
henrys | okay | 00:16.56 |
mvrhel2 | ok. the bbox proc is set correctly at least to start with | 00:18.06 |
| and it found the profile and processes the first page | 00:18.43 |
| oh something is wrong with the null device though | 00:19.04 |
| the dir_name is screwed up on this device | 00:19.38 |
| I see the issue with this one | 00:22.28 |
| when the null device is set, we need to get the directory set up also | 00:22.46 |
| ah a chicken and egg problem | 00:23.46 |
| ok got that fixed | 00:27.14 |
| let me check on linux now | 00:27.21 |
| how do I build in linux with compile inits off? alexcher are you around? | 00:32.07 |
| Robin_Watts is probably sound asleep, or should be... | 00:33.48 |
alexcher | mvrhel2: ./autogen.sh --disable-compile-inits | 01:26.53 |
mvrhel2 | ah. ok | 01:27.19 |
| thanks | 01:27.21 |
| .alexcher are you still there | 01:39.17 |
alexcher | yes | 01:40.56 |
| mvrhel2: yes | 01:41.04 |
mvrhel2 | so what magic path option do I need to give it so that so that it finds gs_init.ps | 01:41.48 |
| I try -I/home/mvrhel/Artifex/git_ghostscript/ghostpdl/gs/Resource/ | 01:43.07 |
| uhoh have to go | 01:43.16 |
| dinner time. | 01:43.18 |
| bbiaw | 01:43.20 |
alexcher | -I/path/to/Resource/Init | 01:43.28 |
mvrhel2 | ok that fixed that | 01:52.44 |
| intersesting. still a path issue with bbox | 01:53.04 |
| I have the windoze case working | 01:53.11 |
| henrys: So I fixed the bbox issue that I saw on windows but it appears that this is different than the issue on linux. I have to step out right now. I will be back on line late tonight | 02:29.32 |
| or I am doing something wrong.. | 03:22.52 |
| henrys: I am off to bed. I will work with you on the linux icc profile stuff tomorrow if you would like. | 05:07.33 |
henrys | okay I just started I'm a little confused the profile directory is being set as both a user parameter and a device parameter. | 05:15.44 |
| from .setuserparams and .putdeviceprops. | 05:17.39 |
mvrhel2 | henrys: i am up for a bit more | 05:27.49 |
| yes. there is some funniness there. let me explain | 05:28.11 |
| the icc profile directory was originally only as a user parameter and set in the icc manager | 05:28.36 |
| when we moved the device profile to the device, we needed to also have a directory set up there too. this is due to the fact that during clist rendering the icc manager is gone | 05:29.28 |
| but we still may need to find the profiles for the device. in particular this could occur if we are playing back a saved clist | 05:29.58 |
| so we needed to also store it in the device. | 05:30.11 |
| in the discussions with ray it was thought that this should be a device parameter | 05:30.27 |
| in the normal mode this is set to be in-sync with the profile dir that is set in the manager. this particular device parameter is not typically set by the user | 05:31.02 |
| unless they happened to know exactly what they were doing. | 05:31.14 |
henrys | okay thanks waiting for the dreaded -Zi debug output to finish. | 05:37.12 |
ray_laptop | I assume mvrhel2 has stopped on the linux COMPILE_INITS=0 issue | 05:43.55 |
mvrhel2 | I am off to bed in a sec | 05:44.06 |
| writing a quick email to alexcher | 05:44.16 |
ray_laptop | henrys: what are you looking for with -Zi ?? | 05:44.25 |
| mvrhel2: can you cc tech ? | 05:44.34 |
mvrhel2 | yes | 05:44.38 |
ray_laptop | I may be up for a while. | 05:44.45 |
mvrhel2 | I did get bbox working on windoze though | 05:44.46 |
| i didnt get a chance to debug linux though | 05:45.10 |
ray_laptop | mvrhel2: BTW, your patch for gs_nulldevice is similar to what I had clusterpushed this AM, but I had some unexpected SEGV and had to set it aside for cust 532 | 05:45.34 |
mvrhel2 | oh ok | 05:45.53 |
ray_laptop | mvrhel2: does -dNODISPLAY also work ? | 05:45.54 |
mvrhel2 | in windows? | 05:46.07 |
ray_laptop | (that installs the 'nullpage' device) | 05:46.09 |
mvrhel2 | ray_laptop: what is it that I should run? | 05:46.27 |
ray_laptop | mvrhel2: Windows -- I assume that it is still broken on linux | 05:46.30 |
mvrhel2 | so bbox device? | 05:46.38 |
| or display device? | 05:46.47 |
ray_laptop | bin/gswin32c -IResource/Init -dNODISPLAY -c quit | 05:47.02 |
mvrhel2 | let me try | 05:47.20 |
henrys | ray_laptop:to see what in the start up SLOP is responsible for setting the icc directory. | 05:47.29 |
ray_laptop | mvrhel2: that's OK. I was going to beat on it anyway (with your patch) on windows and look into linux | 05:47.47 |
| henrys: -Zi is sort of the hard way to find it in the Resource/Init files | 05:48.26 |
mvrhel2 | ray_laptop: that runs fine in windows | 05:48.45 |
ray_laptop | mvrhel2: good -- it worked for me as well. | 05:49.20 |
| henrys: the ICCProfilesDir is never set as a device parameter. grep -l ICCProf Resource/Init/*.ps gives gs_lev2.ps and gs_icc.ps (12 references total including some comments) | 05:51.06 |
| ICCProfilesDir is confined to gs_lev2.ps | 05:51.48 |
henrys | well maybe you and michael should get your stories straight ;-) | 05:53.13 |
mvrhel2 | ha | 05:53.24 |
| it is true that it is never initialized by setting a device parameter | 05:53.41 |
| by the interpreter. | 05:54.20 |
ray_laptop | the profiles directory is set by gsicc_init_device_profile_dir in gsicc_manage.c | 05:54.44 |
henrys | yes and that procedure call is called when device parameters are set. | 05:55.20 |
ray_laptop | this is called from (one place in) gsdparam.c | 05:55.22 |
mvrhel2 | it is really set with gsicc_sync_iccdir | 05:55.42 |
| which occurs when we have both the device and the graphic state | 05:55.54 |
henrys | so gsicc_init_device_profile_dir is called when user parameters are set and wehn device parameters are set. What did you intend? | 05:55.58 |
mvrhel2 | see zdevice.c line 412 | 05:56.36 |
| and the comment above | 05:56.46 |
| that line | 05:56.48 |
| I don't remember why we wanted to make is a device parameter itself. I thought is was for the case if we ever wanted to play back a save clist and wanted to set the directory as a device parameter | 05:58.35 |
| refresh my memory ray_laptop | 05:58.51 |
| s/make is/make it/ | 05:59.06 |
| s/save/saved/ | 05:59.14 |
| sigh | 05:59.16 |
henrys | also I'm getting confused because HEAD seems different than 9.03. | 05:59.18 |
ray_laptop | mvrhel2: a saved clist only has default colorspaces and ICC based colorspaces | 05:59.23 |
mvrhel2 | but it has to map too something, which is the device profile | 05:59.49 |
| profile(s) | 05:59.54 |
ray_laptop | it is true that we would like to preserve the ability to play back a clist to devices with different device profile(s) | 05:59.58 |
mvrhel2 | yes | 06:00.02 |
| and we need to know where to find those | 06:00.12 |
| hence the directory as a device parameter | 06:00.25 |
| that can be set | 06:00.29 |
| for that case | 06:00.34 |
ray_laptop | for example, a saved clist may play back on different media or ink/toner set | 06:00.47 |
mvrhel2 | yes | 06:00.52 |
| exactly | 06:01.02 |
ray_laptop | mvrhel2: as long as the playback can get the profile directory from somewhere during playback, it doesn't need to be in the device cmm_dev_profile_t struct | 06:01.54 |
mvrhel2 | true, but we don't have an icc_manager for playback | 06:02.19 |
| which is where it has been residing | 06:02.28 |
ray_laptop | when playing back a clist, there is an imager state | 06:02.28 |
mvrhel2 | yes | 06:02.33 |
| but it is not really initialized with a valid icc manager | 06:02.52 |
| since all the source profiles are packed in the clist | 06:03.00 |
ray_laptop | mvrhel2: what do you mean we don't have an icc_manager during playback? | 06:03.03 |
mvrhel2 | the icc_manager during playback is NULL | 06:03.13 |
| or least its entries are NULL | 06:03.24 |
| excuse me | 06:03.26 |
ray_laptop | what determined where to get the profiles needed to build the link profiles ? | 06:03.29 |
mvrhel2 | they are in the clist | 06:03.36 |
| those are the source profiles | 06:03.41 |
| all the source colors are defined in the clist | 06:03.59 |
ray_laptop | doesn't the icc_manager manage the device AND source (input) profiles ? | 06:04.24 |
mvrhel2 | icc_manager has nothing to do with the device profiles | 06:04.43 |
| those are left to the device | 06:04.49 |
ray_laptop | mvrhel2: oh, that's right -- when a link is needed, the device provides the output profiles | 06:05.31 |
mvrhel2 | yes exactly | 06:05.36 |
| early on I had the device profile in the icc manager and that was a mistake | 06:05.46 |
henrys | Is the initialization somewhat dependent on whether the device is opened or closed - this could be why X11 is slightly different I know it's open/close semantics are slightly different thant the other devices? | 06:07.22 |
mvrhel2 | oh it could be when its procs are set up | 06:07.58 |
| hold on | 06:08.37 |
henrys | on a working device pbmraw for example I get the icc directory initialized from putdeviceprops according to -Zi I don't get that for X11 - both initialize the setuserparam stuff. | 06:09.23 |
ray_laptop | in particular, I had tripped over the nulldevice not setting up the get_profile early enough | 06:09.24 |
mvrhel2 | put a break point at line 435 in gsdevice.c | 06:09.37 |
ray_laptop | I have to put my 5yr old to bed -- bbiab (or if mvrhel2 has to leave, we can take it up in the AM)... | 06:10.08 |
henrys | okay | 06:10.30 |
mvrhel2 | if the procs.get_profile is not set up then we miss getting things initialized | 06:10.35 |
| I would prefer to work on this in the morning. | 06:10.50 |
| I am fading right now | 06:10.54 |
| and fighting stupid cold | 06:11.00 |
henrys | yeah me too I hate this suff I saw it coming when you said you were going to put the stuff in the device but I suppose there's no other way to do it. | 06:11.42 |
mvrhel2 | the disconnect between the source and device profiles while both needing to have access to the same directory makes it tricky | 06:15.29 |
| good night | 06:27.57 |
ray_laptop | I do want to make one point -- the userparams (and PS setting) are NOT the root issue. ALL parsers need to operate the same way, and know the ground rules for setting/changing the icc profilles directory. The zusparam.c 'set_icc_directory' simply calls "gsicc_set_icc_directory" to set the path (profile directory into an IMAGER state -- which may be a graphics state) | 06:30.53 |
| the issue is to make sure that all devices have access to an imager state with this profile directory (pis->icc_manager->profile_dir) when they are initialized so that the device can load its profiles | 06:32.30 |
| the PS userparam/systemparam issue is irrelevant, but device initialization is the important point. Devices keep a copy of the profile directory, but they can only get it from the imager state _or_ another device (but the imager state version is the 'master', so using another device is "risky") | 06:35.30 |
| enough for now -- I'll be around tomorrow AM to discuss this. If I'm not paying attention to IRC, please call me. | 06:36.37 |
| morning, chrisl , kens | 06:49.46 |
kens | Hi Ray | 06:49.51 |
henrys | looks like a little postscript debugging for x11 gs_init.ps putdeviceprops isn't called - see line 1854 for x11 we just get the { pop pop } and when setdevice is fired off the icc directory is not set. For printer devices this works and the device profile is set. | 06:52.09 |
chrisl | 'evening Ray! | 06:53.28 |
| henrys: I wonder if we're trying to deal with this in the wrong way - maybe a better approach would be be more tolerant of devices *not* having an ICC profile. | 06:54.28 |
henrys | so he needs to do that setdeviceprop even if there isn't new parameters if he wants it to work as he has it coded up now. | 06:54.55 |
| s/setdeviceprop/putdeviceprop/ | 06:55.18 |
| chrisl:maybe | 06:56.53 |
chrisl | henrys: The problem I see is that we have no way to automatically apply a profile on device creation, so there will always be the chance of a device existing without a profile. | 06:58.36 |
henrys | actually we are doing this setup once on the null device - that is the defaultdevice when this code in gs_init.ps is running, I haven't checked how the null device icc profile setting is percolated to other devices . X11 just fails on the setdevice right there in gs_init.ps because the directory wasn't set. | 07:05.37 |
| anyway I should go to bed. | 07:07.36 |
| good night see you in a few hours. | 07:11.43 |
chrisl | 'night henrys | 07:13.05 |
kens | chrisl ? | 07:20.37 |
chrisl | kens: yo! | 07:21.31 |
kens | Shelly has been looking at bug #687593. I'm inclined to ask him to close it, what do you think ? | 07:22.01 |
chrisl | Well, he can't close it, but yes, I think we should close it. | 07:22.46 |
kens | Fair enough, I'll close it then. | 07:23.01 |
chrisl | I'm amused at the way people keep saying how important hinting is for kanji glyphs at small sizes, given Ian's solution for improving (many/most) kanji fonts | 07:24.07 |
kens | Absolutely. Technically they should be correct, of course, but it relies on well made fonts. | 07:24.34 |
chrisl | Yeh, I'm sure there must be some out there, but they seem to be even more elusive than Brigadoon.......... | 07:25.36 |
kens | :-) | 07:25.49 |
| Well, that reduces my assigned bug count by one :-) | 07:26.29 |
chrisl | That really ought to have been mine, anyway. I wondered why I hadn't seen it before | 07:27.46 |
kens | Yes, I didn't realise until just now, I obviously missed it previously. | 07:28.11 |
| You can have #687475 too if you like :-) | 07:28.41 |
Robin_Watts | It's like top trumps with bugs. | 07:29.06 |
kens | chrisl you might want to look at #690340 as well, I think it can probably be closed. | 07:30.58 |
chrisl | kens: Okay I'll take 687475. Strictly speaking, I have an enhancement project to support MM font substitution if the fonts are available, so....... | 07:31.16 |
kens | chrisl basically I think that bug *is* an enhancement request for MM substitution :-) | 07:31.42 |
| The rest of the bug is just nonsense | 07:31.51 |
| If you already have an MM enhancement bug, make this one a duplicate of it. | 07:32.09 |
chrisl | I assumed the MM instance wasn't being created properly - I don't think there's a bug open for the enhancement yet. | 07:32.38 |
kens | I think the bug is just 'we need MM substitution'. | 07:33.02 |
| #687475 that is. | 07:33.13 |
chrisl | Yep, it does look that way. It's a good enough reminder anyway. | 07:33.34 |
kens | That's why I asked Shelly to leaver it open when he asked earlier | 07:33.49 |
| My bug list decreases by 2, result! | 07:34.25 |
chrisl | As for 690340, given it is FAPI, it is clearly in my arena, too. | 07:35.46 |
kens | Well, it sort of is and sort of isn't. | 07:36.11 |
| You could investigate what I was talking about (since I no longer remember), but it looks like maybe the bug should simply be clsoed. | 07:37.22 |
chrisl | I guess I'll get to look at it at some point. | 07:38.17 |
kens | There's definitely no rush | 07:38.35 |
chrisl | Grr, I keep getting weird, spurious build failures on the cluster :-( | 07:38.54 |
kens | Hmm, you're altering the build system though ? | 07:39.13 |
chrisl | Yes, but this was just cherry-picking from master to the 9.03 branch. | 07:40.18 |
kens | <sigh> Windows wants to reboot, it can wait until I go off riding | 07:40.42 |
ray_laptop | mvrhel: henrys: I think chrisl's mention of our graphics library tolerating NOT having a profile or profile directory makes sense. Why not have 'simple' defaults always available built into the code that we can use whether or not we have the %rom% file system. | 07:40.49 |
kens | ray_laptop : But maybe we shuold generate a warning ? | 07:41.11 |
chrisl | I think I understand it, though. If you run clusterpush.pl from the gs directory, it only syncs GS, but tries to test all the languages :-( | 07:41.16 |
ray_laptop | That way we can more closely do what old GS used to do (dumb colors) | 07:41.20 |
kens | chrisl that's bad, I thought it only tested GS from the GS directory | 07:41.48 |
chrisl | kens: apparently not, I'll send a mail to marcos about it | 07:42.12 |
ray_laptop | kens: if you specify 'gs' as a parameter, yes. Otherwise I think it does "all" | 07:42.23 |
kens | Oh, I thought it was sensitive to the directory | 07:42.36 |
chrisl | ray_laptop: do you mean if we don't have a device profile, fall back to the stock Postscript color conversions? | 07:43.25 |
ray_laptop | it may differ between gitpush.pl and clusterpush.pl (two slightly different scripts) | 07:43.32 |
| chrisl: or at least whatever corresponds to the iccprofiles/default.. profiles | 07:44.20 |
kens | My bug list is down bby 3 now. I wonder if I can get below 80 by palming stuff off to other people :-) | 07:44.28 |
ray_laptop | ducks | 07:44.41 |
Robin_Watts | goes for a run. back in a bit. | 07:44.44 |
| Clearly I'm up too early if ray_laptop is still here :/ | 07:44.56 |
ray_laptop | the staff runs away as kens looks for victims... | 07:45.08 |
chrisl | ray_laptop: the problem I saw with that is we'd need individual profiles for every potential number of components. | 07:45.18 |
ray_laptop | Robin_Watts: my wife and kids just got home from HP 3 and 4 movies | 07:45.40 |
kens | thinks Robin_Watts is early and ray_laptop is late. | 07:45.49 |
chrisl | I think we may have changed the cluster behaviour because changes in base affect all languages, not just GS | 07:46.13 |
ray_laptop | tomorrow and Thursday are my turn -- 5 and 6, then the two "Deathly Hallows" movies back to back (they start even later 9pm and 12:01AM PDT) | 07:47.13 |
Robin_Watts | commiserates. | 07:47.46 |
kens | Haven't seen those, doubt if I will, didn't enjoy the book. | 07:47.52 |
ray_laptop | I am fading fast, so any technical comments are heretofore BS | 07:48.04 |
Robin_Watts | disliked the first 2 movies. Haven't seen the rest. | 07:48.16 |
ray_laptop | I've read all of the books at least twice | 07:48.39 |
kens | Early books, yes, later books, no. Melanie didn't finish the last one. | 07:49.02 |
ray_laptop | mostly because I like to have them fresh in my mind before seeing the movie | 07:49.10 |
| the last book really D...R...A...G...G....E...D | 07:49.39 |
kens | I feel that way from 4 onwards, it just got progressively worse | 07:49.57 |
ray_laptop | I could have written it better in 50 pages or so | 07:50.01 |
kens | I think that's the problem, the earlier books show signs of an editor's hand, the later ones I don't think she would let anyone touch. | 07:50.36 |
ray_laptop | but on the plus side, that means that it is easier to not leave parts out of the 4 hours of film | 07:50.52 |
kens | Well, that's a bonus. But why make the last book into 2 ? Oh yes, to spin out the gravy train, silly me.... | 07:51.29 |
ray_laptop | kens: how innocent you sound ;-) | 07:51.52 |
| off to my pillow ... | 07:52.53 |
Robin_Watts | reboot | 09:16.18 |
| tor8: you here? | 12:55.36 |
tor8 | yup | 12:55.43 |
Robin_Watts | How hard would it be for me to reproduce your testing setup here? | 12:56.01 |
tor8 | fairly trivial :) | 12:56.28 |
Robin_Watts | Will it work on windows? | 12:56.37 |
| or macos ? | 12:56.43 |
| or do I need to boot Ubuntu ? | 12:56.49 |
tor8 | mac or unix | 12:56.51 |
Robin_Watts | laptop might not be a bad choice actually, so mac. | 12:57.08 |
| What's the best way to proceed then ? | 12:57.49 |
tor8 | I'm finding the most up-to-date version to send to you now | 12:58.33 |
| it's basically just two scripts that run from a directory where it stores all the temporary files | 12:58.57 |
| and it predates git ... now I have to remember how darcs works again :) | 12:59.37 |
Robin_Watts | Is it worth making a git repo out of it and putting it on ghostscript.com ? | 13:00.29 |
| The cluster is in such a repo now. | 13:00.35 |
tor8 | yeah, I should do that | 13:00.53 |
| in the meanwhile: http://ghostscript.com/~tor/stuff/sane.zip | 13:01.13 |
| if you run on mac, you need to change line 3 of sane.sh to read: | 13:01.34 |
| oh, drats, that's another script I have locally | 13:01.59 |
| anyway, readlink -f $0 doesn't work on macosx | 13:02.11 |
| #!/usr/bin/env python | 13:02.20 |
| import os,sys | 13:02.21 |
| print os.path.realpath(sys.argv[1]) | 13:02.21 |
| that's the python one-liner I use on osx, perhaps there's a way to make it a real one-liner | 13:02.43 |
| SANE=$(dirname $(python -c 'import os,sys;print os.path.realpath(sys.argv[1])' $0)) | 13:03.43 |
| run sane.sh from the mupdf directory, and it'll build and test and diff against previous runs | 13:04.24 |
Robin_Watts | so the python one liner should be saved as readlink.py ? or readlink ? | 13:12.37 |
| in the same directory as sane.sh ? | 13:12.44 |
tor8 | see my SANE= line, no need to keep it as an external script | 13:15.49 |
Robin_Watts | tor8: so I put that in the environment? | 13:16.44 |
| or add that to sane.sh ? | 13:16.48 |
tor8 | replace line 3 of sane.sh | 13:16.55 |
Robin_Watts | ok | 13:17.01 |
| md5sum: command not found | 13:21.34 |
| Should macosx have that? | 13:22.04 |
tor8 | it's "md5 -r" on macosx | 13:23.04 |
Robin_Watts | ok, that seemed to work. | 13:26.51 |
| Now to add my changes in. | 13:27.01 |
kens | Hmm, that's better. I've got my linked list workling properly. I can now extract the text from the Adobe Type Manager 4.6 flyer. Seems to be OK, including correct extraction of 'f' and 'ffi' ligatures. It also extracts the 'text' for the type 3 font containing the Adobe logo :-) Doesn't make any sense, but its there. | 13:30.53 |
| There is one problem with location so I'll have to work on that. | 13:31.35 |
Robin_Watts | I just applied my patch, reran sane, and it rebuilt and told me there were no differences. | 13:33.57 |
| which is bad, because Isolated-rgb.pdf definitely looks different. | 13:35.37 |
| tor8: Any ideas? | 13:36.04 |
tor8 | it takes two diffs, one for error logs, and one for images | 13:36.39 |
| it'll say "no differences" if the error logs are the same (since it just calls "diff" for that) | 13:37.01 |
| even if the images are different | 13:37.06 |
Robin_Watts | Ah, right. | 13:37.19 |
tor8 | do you get any lines with ! and some checksums after that? | 13:37.21 |
Robin_Watts | where ? | 13:37.34 |
| When running sane.sh to stdout ? no. | 13:37.55 |
tor8 | hm, you may need to install imagemagick (or replace the "compare" command line with something else at the top of sane-diff.py) | 13:38.14 |
Robin_Watts | I have compare. | 13:38.33 |
tor8 | it should print lines like "! file.pdf pagenumber" for all the pages that differ | 13:40.01 |
| sane diff old-version new-version (with the git-describe tag names used by sane to save the logs) | 13:40.32 |
Robin_Watts | Diffing 0.8.179-2c4bbbf 0.8.180-a28240f | 13:42.07 |
| 0 files changed. | 13:42.16 |
| Done. | 13:42.18 |
tor8 | hmm, odd | 13:43.26 |
Robin_Watts | Where are the output png files stored ? | 13:43.48 |
tor8 | it saves the md5 sums for the renders in $SANE/log/ | 13:43.49 |
| png files are generated on the fly when diffing by rerunning pdfdraw | 13:44.05 |
| and cached in $SANE/png | 13:44.12 |
Robin_Watts | All I have in log is 0.179.blah | 13:44.44 |
| and in there I have pdfdraw | 13:44.49 |
| and that's all. | 13:44.51 |
| (and I have 0.180.blah too) | 13:45.03 |
tor8 | there should be one *.log and one *.sum for each test file run | 13:45.37 |
Robin_Watts | nope. | 13:45.50 |
tor8 | you got a $SANE/test directory with pdf files? | 13:46.13 |
Robin_Watts | To be clear... I'm in mupdf.git | 13:46.17 |
| I do ../sane/sane.sh | 13:46.25 |
tor8 | that's where you should be, and exactly what you should be doing | 13:46.33 |
Robin_Watts | In sane I have a test directory, with lots of pdf files in. | 13:47.07 |
tor8 | it should create symlinks in ../sane/sym/ to the test files when you run sane.sh | 13:47.18 |
Robin_Watts | Ah. It doesn't. | 13:47.36 |
tor8 | in retrospect, I should just use the original pdf file names instead of md5-summing them | 13:47.40 |
Robin_Watts | All I have in sym is a .pdf file. | 13:47.44 |
tor8 | hm, probably the $(md5sum ...) command that fails still | 13:48.04 |
| oh... could it fail because there's a function md5 in the script, that's called the same as the md5 command you try to call on osx? | 13:49.18 |
| I solved it (a long time ago) by making a md5 shell script in my $HOME/bin | 13:49.33 |
Robin_Watts | will try tweaking in a mo. brb. | 13:50.00 |
| ok, this seems much more likely. | 13:57.09 |
| I globally replaced md5 with md6. | 13:58.31 |
| and then just edited the actual command back to md5 | 13:58.40 |
| excellent. | 14:00.02 |
| On macos, there is a handy command "open" that I recently discovered. | 14:00.30 |
| open report/blah.html will do whatever clicking the file in the finder window would do. | 14:01.00 |
| hence in that case, it will kick the report into a browser. | 14:01.15 |
| No diffs (except in the isolated case), but 1 segv. | 14:01.38 |
tor8 | Robin_Watts: yeah, the same as the "start" command in windows | 14:12.38 |
Robin_Watts | ooh. | 14:12.58 |
tor8 | open -a "application name" is handy, if you can't be bothered to dig through the applications directory | 14:13.01 |
Robin_Watts | I may attempt to make sane.sh open the report after it's generated. | 14:14.11 |
| Just trying to understand how I can have a pop_clip without a mask | 14:14.35 |
| oh, empty image. | 14:15.58 |
mvrhel2 | henrys: I see you tracked down the issue | 14:55.30 |
| I am trying to make sense though of how gs and the devices are initialized. surely there must be some commonality amongst devices | 14:58.06 |
| what ever that commonality is that is where the sync of of the iccdir should occur | 14:58.54 |
| bbiab need to eat | 14:59.52 |
Robin_Watts | fears mvrhel2 may be hunting for the missing link. Or for dark matter. Or for a snark. | 15:01.08 |
| Hmm. shadings and shapes... | 15:03.26 |
chrisl | The closest thing to such device commonality I've seen is gx_device_(forward_)fill_in_procs(), but even that isn't sure to used. | 15:03.36 |
henrys | in the interest of getting a release out I think we need a hack to get what you have working. | 15:05.46 |
Robin_Watts | tir8: Does fitz have a function to fill a given rectangle of a pixmap with a value ? | 15:08.52 |
| tor8, even. | 15:08.59 |
henrys | marcosw:could you test a handful of random files and make sure compile inits 0 and 1 output match? | 15:12.07 |
Robin_Watts_ | They are hooking up the solar panels to the mains now, so expect me to be dropping in and out of here for a while. | 15:14.48 |
| router isn't on the UPS. | 15:14.54 |
marcosw | henrys: do I need to run the compile inits 0 vs 1 test with a particular revision? I ran a compile inits 0 regression yesterday morning. | 15:26.41 |
henrys | mvrhel2:I guess 9.03 we know printer devices work. | 15:27.21 |
marcosw | that was with a06634a969ea9b0db2d0840d8641847247056145 | 15:27.34 |
| btw, I'm also going to be dropping off in about 30 minutes due to the mains being cut. I'm having my electric car charger installed today. | 15:28.46 |
mvrhel2 | I am back. That is funny Robin_Watts | 15:29.07 |
Robin_Watts | mvrhel2: ? | 15:29.22 |
mvrhel2 | what I am hunting for (with respect to device) | 15:30.03 |
| devices | 15:30.07 |
Robin_Watts | oh, right. | 15:30.09 |
henrys | alexcher:do you have any ideas how to gurantee the put params is called? | 15:30.42 |
| disk space problems on mac pro - I wonder if the regression have gone wild or something. | 15:32.28 |
alexcher | henrys: at what moment? | 15:33.29 |
marcosw | henrys: I jsut checked and the cluster doens't appear to be using more space then normal on your macpro. | 15:34.24 |
henrys | gs_init.ps:1865 see your email | 15:34.27 |
| marcosw:thanks it might be me, I'll check. | 15:34.45 |
marcosw | you might take a look at /var/log/system.log. it's huge. | 15:34.58 |
mvrhel2 | ah some good news from customer 711 | 15:35.27 |
henrys | bbiab I'm going out for breakfast this morning be back in an hour or so. | 15:40.19 |
kens | henrys I see the PCL type 3 thing just came back again | 15:42.24 |
| marcosw any results with support from Adobe on that ? | 15:42.35 |
Robin_Watts | Solar Panels up and running. | 16:04.05 |
ray_laptop_ | hi, all | 16:04.44 |
mvrhel2 | Robin_Watts: you have solar panels? | 16:05.01 |
| hi ray_laptop | 16:05.08 |
Robin_Watts | I do now :) | 16:05.15 |
mvrhel2 | cool | 16:05.20 |
| I have a nice south facing roof here. If it made sense $$ wise I would do it | 16:05.44 |
chrisl | 'morning ray_laptop | 16:05.49 |
Robin_Watts | Getting 2.4kW now, which isn't bad on a fairly hazy day. | 16:05.54 |
| mvrhel2: Does Washington have subsidies? Those can really make the difference. | 16:06.11 |
ray_laptop_ | it occurred to me this morning that we could punt on the icc profile directory and put it in the pseudo-globals (gs_lib_ctx_t) that is pointed to by the gs_memory_t | 16:06.12 |
mvrhel2 | oh yes that is true | 16:06.31 |
ray_laptop_ | we already have that polluted with things like gs_font_dir and accurate_screens | 16:06.39 |
Robin_Watts | http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=115x40597 | 16:07.08 |
ray_laptop | that way there is no nonsense about when a device is intialized, etc. | 16:07.31 |
mvrhel2 | right | 16:08.05 |
ray_laptop | henrys: are you here ? | 16:08.22 |
mvrhel2 | he stepped out for breakfast. said he would be back in about 1 hour | 16:08.37 |
ray_laptop | I see. (I hadn't read carefully thru the logs). | 16:08.59 |
| moving the profile_dir to the gs_lib_ctx is sort of a sledgehammer, but with the number of bugs that have kept showing up due to the way it's done now, a sledgehammer might be just the tool we need | 16:10.15 |
mvrhel2 | I want to check one thing real quick before we go down that route | 16:12.11 |
chrisl | ray_laptop: do you know what the target platform for cust 532 is? | 16:17.45 |
ray_laptop | chrisl: it's a Power PC | 16:18.05 |
| so it is big endian | 16:18.13 |
chrisl | Right, and OS? | 16:18.24 |
ray_laptop | chrisl: I think it is Wind River | 16:18.52 |
chrisl | ray_laptop: VxWorks, or embedded Linux? Sorry, but I'm having much luck getting info from them. | 16:19.44 |
| s/I'm/I'm not | 16:19.53 |
ray_laptop | I'll call Len -- hold on | 16:20.05 |
| chrisl: it is Monta Vista linux | 16:21.18 |
chrisl | ray_laptop: cool thanks! | 16:21.36 |
kens | TIme for me to go, night all. | 16:22.02 |
ray_laptop | chrisl: he is going to try setting for 'SPARC' so that the byte order comes out correct (High-Low), what we call big endian, but he is going to got to unsigned chars | 16:27.32 |
| s/got/go/ | 16:27.39 |
| chrisl: he thinks that is needed to get past the 5100 error | 16:28.12 |
| mvrhel2: what approach are you looking into ? | 16:31.09 |
mvrhel2 | ray_laptop: I am just looking at this to understand better why we fail with X11 and not the printer ones. I don't see why they have to be different | 16:32.28 |
| s/ones/devices/ | 16:32.45 |
ray_laptop | mvrhel2: I'm just concerned that with all the various devices that (sometimes) act differently, having it handled by put params and stored in the device is going to continue to bite us | 16:34.37 |
mvrhel2 | ok. | 16:35.20 |
| You understand how all the initialization stuff works better than me so I will take your word that it is better to be done this way | 16:36.13 |
ray_laptop | mvrhel2: that's part of the problem -- I don't understand the vagaries of the the initialization well enough. I do know that we have to do this so it works for the other parsers as well (not rely on the way PS does it) | 16:37.51 |
| the plmain (PCL, l-s, xps) uses a totally different method of device initialization | 16:38.34 |
chrisl | ray_laptop: I just replied to Len. I think they really need to actually create a definition for their target, the ones in ufstport.h are only common examples, you are expected to tailor it for your own platform. | 16:38.59 |
ray_laptop | because it wants to have a single device that is shared in the language_switch approach | 16:39.18 |
| chrisl: that's a good recommendation. thanks | 16:39.42 |
| chrisl: I'm glad that I'm not having to field these questions | 16:40.05 |
chrisl | ray_laptop: I just can't see how using the SUN/SPARC definitions on a Linux/PPC target is doing the "right thing"! ;-) | 16:40.34 |
ray_laptop | well, at least the are both big endian unix variants | 16:41.10 |
| and the SPARC is picky about alignment, as is the PPC | 16:42.00 |
| IIRC | 16:42.12 |
chrisl | True, but still, I think an accurate platform definition would be a good idea | 16:42.25 |
ray_laptop | it's been about 11 years since I played around on a sparc | 16:42.45 |
chrisl | Yes, SPARC is picky about alignment, I can't remember about PPC, but for UFST you set the alignment explicitly in your platform definition. | 16:43.32 |
ray_laptop | what kind of electric car does marcos have ? | 16:43.36 |
| Artifex should get some kind of 'green' award -- I drive a hybrid, marcos and electric, Robin and Ken have solar panels, ... | 16:45.32 |
Robin_Watts | Ah, but hybrids are evil... | 16:46.05 |
ray_laptop | Robin_Watts: more so than electric ? | 16:46.30 |
Robin_Watts | no. | 16:46.36 |
chrisl | I need to head out, I'll check back later....... | 16:46.46 |
ray_laptop | chrisl_away: OK | 16:46.52 |
| I have a good south facing roof as well, and we are considering PV panels. It's just that our electricity usage is already pretty low and has continued to improve as we make "little changes", so we're right on the cusp of it making $$ sense | 16:48.42 |
| but my wife wants it just 'cause she wants to help the environment | 16:49.08 |
| mvrhel2: cust 850 asked again today aobut 9.03 -- what do you think about doing the lib_ctx hack for 9.03 at least so we can get them a low risk OOB release. Then we can decide later what to do for the Augst release | 16:56.30 |
mvrhel2 | I am fine with that | 16:58.29 |
| ray_laptop: do you want to make the fix? | 16:58.40 |
ray_laptop | I'd like to wait for henrys to see if he buys into that -- can you believe it -- me wanting consensus ;-) | 16:59.20 |
| mvrhel2: but I'll go ahead and start changes on the 9.03 branch so we can have a look at it (and test it) | 16:59.55 |
mvrhel2 | ok sounds good | 17:00.08 |
ray_laptop | that's one nice thing about a published branch | 17:00.26 |
| oops. the patch for cust 532 didn't work for them -- I have to look at it. (they say it is URGENT) :-/ | 17:10.55 |
| alexcher: I just made a trivial change to gs_res.ps to add a "flush" after the -sGenericResourceDir=... message because I saw it getting garbled when the COMPILE_INITS=0 failed (the other error messages on stderr were garbling it) | 17:13.57 |
alexcher | ray_laptop: that's great. | 17:18.25 |
ray_laptop | great may be a bit of an overstatement | 17:20.59 |
henrys | I'm back | 17:21.02 |
| I have relatives here and am swapping my hours around a bit. | 17:21.19 |
ray_laptop | henrys: I had thought about moving the profiledir to the gs_lib_ctx (see above) so we don't have this init problem and multiple copies (potentially getting out of sync). | 17:22.50 |
henrys | the simple hack to always call putparams before setdevice in gs_init.ps fixes all cases doesn't it, presumably we could read a set parameter in and set it ot the same value and at least the call would be made. | 17:23.15 |
ray_laptop | I'm in the middle of doing that for the 9.03 branch in case we want to get it out quick | 17:23.16 |
henrys | but yes setting it in gs_lib_ctx seems good to me also. | 17:23.55 |
ray_laptop | it's sort of non-intuitive that we have to call putparams on a device before the setdevice, isn't it ? | 17:24.18 |
henrys | I agree. | 17:25.20 |
ray_laptop | henrys: if you have done that for 9.03 and are confident of it, I'm fine with that for 9.03, however -- then we can discuss it for the Aug release. | 17:25.49 |
henrys | also the parameter is shared by the interpreter and the device, so it should be stored in a single location accessable by both right? | 17:26.31 |
| no I haven't done anything - I just analyzed it. | 17:27.21 |
ray_laptop | well, the gs_lib_ctx _is_ accessible by everyone (via the gs_memory hook) so we'd only need one copy | 17:27.52 |
henrys | you can't do a putparams with 0 parameters so something more tricky is needed. How long will it take to do gs_lib_ctx changes do you think? | 17:28.32 |
ray_laptop | henrys: as I mentioned above there are other things in the lib_ctx like font_dir and accurate_screens stuff, so it isn't any more ugly :-/ | 17:29.40 |
| henrys: probably about another 20 minutes or so | 17:30.04 |
henrys | okay let's go with lib_ctx changes then, | 17:30.16 |
ray_laptop | (if I stay off the phone and IRC) | 17:30.21 |
ray_laptop | hunkers down, then | 17:30.32 |
| henrys: how does plmain set the ICCProfilesDir ? | 17:31.23 |
henrys | alexcher:I am curious why I need '-I' even if I set GenericResourceDir. | 17:31.47 |
| ray_laptop:ha the other languages always embed the profiles so it probably just falls back. | 17:32.30 |
ray_laptop | henrys: you don't need -I if the gs_init.ps file is already on one of the paths, but if you haven't installed, it probably isn't | 17:32.40 |
| henrys: "falls back" ? you mean "falls over" (dies) ? | 17:33.15 |
mvrhel2 | ray_laptop: let me know if you need anything from me | 17:33.38 |
henrys | no falls back to %rom% profiles which are always present. | 17:33.49 |
ray_laptop | with COMPILE_INITS=0 the device needs to know where to find the profiles | 17:33.52 |
| henrys: so we only support PCL with COMPILE_INITS=1 ??? | 17:34.10 |
henrys | yes - then there is another option BUNDLE_FONTS to not include the fonts. | 17:35.25 |
| from the makefile: | 17:35.52 |
| # This is constant in PCL, XPS and SVG, do not change it. A ROM file | 17:35.53 |
| # system is always needed for the icc profiles. | 17:35.53 |
| COMPILE_INITS=1 | 17:35.53 |
ray_laptop | mvrhel2: that was another idea that I was thinking -- always have a simple, minimal, set of profiles built in so that we fall back to the old PS color model | 17:35.55 |
henrys | | 17:35.58 |
ray_laptop | mvrhel2: so that if we can't find a iccprofiles directory and don't have %%rom then we can warn and go on with the simple defaults | 17:37.35 |
| darn chatzilla -- sometimes swallows my % but doesn't accept %% as an escaped % :-( | 17:38.22 |
mvrhel2 | yes we could always have the simple ps profiles available | 17:39.20 |
| that I made | 17:39.24 |
| need to think about that a bit | 17:39.40 |
ray_laptop | henrys: I won't worry about the other parsers with COMPILE_INITS=0 for now, | 17:39.46 |
| mvrhel2: OK. not for 9.03, but maybe for the release | 17:40.03 |
henrys | yes I'll take care of that as there is interest - I just wasn't anticipating complex profile setups for the other languages, which is why I left it in the state it's in. | 17:41.24 |
| I don't know about that, I'd rather have an obvious error vs. a bug of using the wrong profile. | 17:48.12 |
alexcher | henrys: -I adds directories to the search path used by findlibfile. | 17:51.33 |
henrys | yes I think it should be enough to set GenericResourceDir but nvm. | 17:54.38 |
| oh I see my message to michael didn't go to tech@artifex.com | 17:59.59 |
| Delivery to the following recipient failed permanently: | 18:01.00 |
| | 18:01.00 |
| tech@artifex.com | 18:01.00 |
| | 18:01.00 |
| Technical details of permanent failure: | 18:01.03 |
| Message rejected by Google Groups. Please visit http://mail.google.com/support/bin/answer.py?hl=en&answer=188131 to review our Bulk Email Senders Guidelines. | 18:01.07 |
| | 18:01.10 |
| I resent the message - but this could be a problem - tech is hardly "bulk" | 18:03.49 |
Robin_Watts | henrys: AIUI, it's saying that it thought your message was spam. | 18:07.50 |
| While in some cases we may classify a message as spam and deliver it to the spam folder, we also try to find ways to reduce the amount of spam being sent to Gmail in general. If we detect that a message has a strong likelihood of being spam, weâll block the message from being sent to Gmail. | 18:08.21 |
henrys | the only thing I see that would identify the message as spam is tech consists of multiple recipients so I am concerned addresses like tech and support are going to not receive real mail. | 18:10.36 |
ray_laptop | henrys: do you have your own mailing list for tech, or are you using the 'tech' pseudo list on gmail ? | 18:12.10 |
henrys | the gmail pseudo and I am sending from the gmail web interface. | 18:13.35 |
Robin_Watts | If it's rejecting emails sent from it's own web interface, that IS a bit poor. | 18:14.48 |
henrys | I just forwarded another copy of the message with subject setdevice to tech did you guys get it? | 18:15.05 |
Robin_Watts | fetching now (have been for a while) | 18:15.20 |
| I have the "Fwd: Delivery Failure" one. | 18:20.24 |
| "Fwd: Delivery Status Notification (Failure)" rather | 18:20.47 |
henrys | I guess we'll call it an anomaly for now and see if it repeats. | 18:28.51 |
Robin_Watts | mvrhel2: What is the 'shape' of an image with alpha ? | 18:36.32 |
| Is it the outline of the image bbox? Or is it the alpha channel of the image ? | 18:37.03 |
mvrhel2 | good question | 18:37.13 |
| I had thought that alpha = shape * opacity | 18:38.19 |
| according to adobe.... | 18:38.57 |
| however I have trouble with all these definitions | 18:39.29 |
Robin_Watts | Oh, I think I see my cockup. | 18:40.05 |
mvrhel2 | I would say that the shape of the image would be the image bbox | 18:40.06 |
Robin_Watts | mvrhel2: Really? | 18:40.17 |
| That's not what I have coded here :( | 18:40.37 |
mvrhel2 | well. it really should not matter | 18:40.42 |
| the alpha value contains the shape basically | 18:40.58 |
| you could think of it as being 0 where alpha is 0 and 1 otherwise | 18:41.16 |
| but it is not really of any use of that case | 18:41.29 |
Robin_Watts | I've fixed the cockup, I think. | 18:41.49 |
| It matters when you are keeping a separate shape plane in order to blend groups back properly. | 18:42.09 |
| but I think I have it working now. | 18:42.15 |
mvrhel2 | ok good | 18:42.18 |
Robin_Watts | Thanks. | 18:42.24 |
mvrhel2 | saves me from giving the wrong answer ;) | 18:42.32 |
Robin_Watts | I'm going to assume my code is right until I find a test file that shows me otherwise. :) | 18:42.51 |
| food time. | 18:42.55 |
ray_laptop | mvrhel2: I don't have the 'master' handy anymore, but can you look at zusparam.c set_icc_directory | 18:44.37 |
mvrhel2 | sure | 18:45.01 |
| ray_laptop: what do you need? | 18:45.07 |
ray_laptop | mvrhel2: it allocates a copy of the parameter then does the gsicc_set_icc_directory then frees it. why ? | 18:46.24 |
| I thought the whole idea of a directory name length was so we could support unicode names that may have '\000 | 18:47.10 |
| '\000' characters in them | 18:47.23 |
| (UCS-8) | 18:47.32 |
henrys | alexcher my statement about gs_init.ps wouldn't make sense if you didn't get the email sorry for the confusion. | 18:48.13 |
ray_laptop | I've kept the 'len' of the profiledir along with the profiledir pointer in the gs_lib_ctx -- if we have a length, (and don't use strlen) can we avoid this and just use the length ? | 18:48.50 |
| mvrhel2: I think we can just use: | 18:50.55 |
| /* Check if it was "NULL" and ignore it if so */ | 18:50.56 |
| if (pval->size != 0 ) { | 18:50.58 |
| gsicc_set_icc_directory(pis, (const char*) pval->data, pval-size); | 18:50.59 |
| return(0); | 18:51.01 |
| } | 18:51.03 |
mvrhel2 | sorry ray_laptop on phone for sec. reading now... | 18:51.52 |
| ray_laptop: yes I agree. I don't know why I did that | 18:52.36 |
ray_laptop | mvrhel2: I'm thinking about things like CJK | 18:52.46 |
| mvrhel2: OK. Thanks. Almost done with edits and ready to test compile (and see all the things I missed). | 18:53.21 |
mvrhel2 | CJK is the sort of thing that makes me push off this task to someone else... | 18:53.40 |
| and unicode.... | 18:53.59 |
ray_laptop | we can dump it on Robin -- he was the last person to take a whack at it ;-) | 18:55.17 |
henrys | mvrhel2:I am going to go back and find the rest of 711's timing regression unless there is something else you want me to look at. I'd like to get you back to planar as soon as possible so if you can get others to help you clear your desk that would be good. | 18:55.40 |
mvrhel2 | henrys: I was just about to send you an email about customer 711 | 18:56.05 |
| this is progressing into a screen design problem with a transfer function | 18:56.17 |
henrys | meaning you would need to look at multiple output iterations on a real device etc. | 18:57.41 |
mvrhel2 | meaning that I have the feeling that they don't have the tools to do this and that I could fix up my screen creation tool so that they could use it assuming they can make measurements etc | 18:59.31 |
| but it all requires time | 18:59.43 |
| I have an appt at 12:30 I will be back at 1 | 19:01.39 |
henrys | okay | 19:01.48 |
ray_laptop | mvrhel2: I might be done with the profiledir changes by then ;-/ | 19:08.29 |
| more changes than I had thought, but seems cleaner now | 19:11.47 |
| might even work ;-) | 19:11.55 |
| the gs universe remains in balance as well -- we took BITTAG out of gs_lib_ctx and this puts something back into it ;-) | 19:13.45 |
Robin_Watts | ray_laptop: mvrhel2: In theory, all strings held inside ghostscript (like profiledirs etc) are UTF-8 encoded. | 19:23.57 |
| The wonderful thing about UTF-8 is that you can treat it exactly like ascii. | 19:24.23 |
| Same sort order, same strlen, strcpy, strcmp etc. | 19:24.33 |
ray_laptop | Robin_Watts: I thought UTF-8 had nulls in it -- I must be thinking the 16-bit Unicode (whatever that is called) | 19:25.48 |
| or Windows 'wide' chars | 19:26.10 |
Robin_Watts | ray_laptop: In UTF-8 a 0 byte means end of string, same as in ascii. | 19:26.32 |
| and directory separators (and indeed all chars <128) are just represented by themselves. | 19:27.11 |
ray_laptop | OK, well maybe we can get rid of the 'len' values in the ICC stuff (I, in my misunderstanding, steered mvrhel2 wrong) | 19:27.22 |
| we'll leave that cleanup for after we squeeze 9.03 out | 19:28.07 |
| (maybe the Aug release) | 19:28.22 |
| mvrhel2: when you get back, please have a look at the commented out code in zputdeviceparams (psi/zdevice.c) -- there's a commented out: code2 = gsicc_init_device_profile_struct(dev, NULL, 0); | 19:32.18 |
| hurray! have a clean build (on Windows). Now to test... | 19:44.34 |
chrisl | alexcher: The VC7 build problem you had yesterday, I think I know how to fix it - should I do so? | 19:46.33 |
| ray_laptop: looks like Len has got a few steps further after the suggestion(s) I made - yay! | 19:47.11 |
ray_laptop | chrisl: congrats ! and thanks for perservering | 19:53.14 |
chrisl | ray_laptop: it's a bit of a relief, I must say! I await the next problem, of course ;-) | 19:55.07 |
| ray_laptop: how are we looking for 9.03? I notice the customer prompted us today....... | 19:56.56 |
ray_laptop | chrisl: I'm working on a fix for the ICC profiles directory problems that we classed as a show stopper | 19:58.52 |
| chrisl: it builds and runs COMPILE_INITS=1 -- now building and trying COMPILE_INITS=0 ... | 19:59.28 |
chrisl | ray_laptop: okay, great. If it all works okay, can you drop me a mail letting me know, and with the script and instructions for the commercial release, please? | 20:00.13 |
ray_laptop | I'll do a clusterpush to make sure that I didn't mess up COMPILE_INITS=0 then also push the change to the 9.03 branch for marcos to test with COMPILE_INITS=0 and others to beat on | 20:00.53 |
| chrisl: wilco | 20:01.13 |
chrisl | ray_laptop: cheers. I'll leave you in peace now, and go and watch some television....... | 20:01.40 |
ray_laptop | it's a little nervous making in that it changes 9 files and is pretty extensive change | 20:02.03 |
| for right before a release... | 20:02.16 |
chrisl_away | needs must when the devil etc..... | 20:03.00 |
Robin_Watts | tor8: New mupdf commit waiting for you on casper. | 20:09.20 |
alexcher | chrisl: yes, please fix the VC7 problem. | 20:10.45 |
Robin_Watts | Advances VOX and isolation-rgb.pdf | 20:11.04 |
mvrhel2 | ok I am back | 20:11.07 |
ray_laptop | mvrhel2: darn. I messed something up. Somebody is stomping on the LIBPATH list :-( | 20:15.21 |
| during running the init files (fails after gs_lev2.ps) | 20:16.07 |
| need some food ... | 20:16.31 |
mvrhel2 | ray_laptop: that commented out code should be removed | 20:19.38 |
| everything is set during gs_putdeviceparams | 20:20.53 |
chrisl_away | alexcher: Will do. I'm testing it now, but it takes an age to build (VBox, shared drive.....) so I'll leave it building, if it goes right, I'll commit it tomorrow morning (my time) | 20:25.45 |
mvrhel2 | henrys: are you around? | 20:39.10 |
henrys | yes | 20:43.45 |
| I am wondering if it can't be done with just tuning parameters this is 180 dpi monochrome. | 20:44.24 |
mvrhel2 | I am wondering that too. I get the feeling they are not to familiar with hafltoning though | 20:55.49 |
| to me the lowres screen output that they showed doesn't look too bad | 20:56.14 |
| My worry is that do the expect us to tell them exactly what HT values work best for their device? | 20:56.50 |
| s/the/they/ | 20:56.57 |
| henrys^^ | 20:57.00 |
| in terms of linearity it looks pretty decent | 20:58.02 |
| there is the ever present trade off of levels versus resolution | 20:58.38 |
| and they are seeing it when they go to the high res screen | 20:58.49 |
| where they have lost levels | 20:58.54 |
| if they want perfection they will need to create their own screen that is optimal for their device | 21:00.16 |
henrys | from his email though he is complaining the 5,10,15,20 are pure white in the middle example that doesn't seems right to me either. | 21:02.37 |
| unfortunately I have to go out for a little bit and will be back this evening. | 21:05.28 |
mvrhel2 | ok. np. | 21:05.36 |
| well he hit the high res screen with a exp 0.2 | 21:05.46 |
| which means the values at high lights get pushed even lighter | 21:06.45 |
| he needs a LUT for the transfer function | 21:06.55 |
| or a different parametric function that behaves more linear in the highlights | 21:07.24 |
| I will craft a reply. How hard is it to use a LUT for a transfer function? | 21:08.27 |
| it would be better for him to generate a screen though with the transfer function baked in as alexcher had suggested | 21:08.58 |
| but first we need to determine the function | 21:09.19 |
alexcher | mvrhel2: transfer function is just a PS program. LUT access can be coded in one line of code. | 21:15.27 |
mvrhel2 | alexcher: ok so if I had a table to N float values can you supply we with a way to use that as a transfer function? | 21:16.26 |
| s/we/me/ | 21:16.31 |
| as a PS program | 21:16.38 |
| s/table to/table of/ | 21:16.51 |
alexcher | mvrhel2: yes, of course. | 21:17.07 |
mvrhel2 | ok. could you do that today? | 21:17.18 |
alexcher | mvrhel2: Yes, 20..40 min | 21:17.45 |
mvrhel2 | alexcher: very nice | 21:17.52 |
| thank you | 21:18.00 |
| bbiab | 21:37.55 |
ray_laptop | COMPILE_INITS=0 works (and =1 as well). Doing gitpush regression .... | 21:42.19 |
| mvrhel2: we should explain to cust 711 why baking the transfer function into the threshold array is a good idea -- otherwise even if they have 256 dots (16x16) threshold array, their transfer function may be such that many of the levels are all effectively black. | 21:46.53 |
| mvrhel2: with their extreme dot gain, even with 16x16 they may end up with only about 60 or so usable levels | 21:47.31 |
mvrhel2 | ray_laptop; yes I agree | 21:57.49 |
| the question is, is that something we do for them or do they do it themselves | 21:58.05 |
| to me this is getting into a gray area | 21:58.14 |
| pun intended | 21:58.24 |
| back shortly | 22:16.45 |
alexcher | mvrhel2: Done. | 22:26.59 |
ray_laptop | strange. I did a gitpush from my Windows machine and the compiles fail, but I sent the patch to peeves and applied it to 9.03 and it compiles fine ??? | 22:55.36 |
| and COMPILE_INITS=1 runs fine -- even with x11 | 22:56.45 |
| henrys: can I send the patch to 9.03 to you in case you can figure out why the cluster regression compile fails ? | 23:14.24 |
| Forward 1 day (to 2011/07/14)>>> | |