| <<<Back 1 day (to 2014/03/16) | 2014/03/17 |
henrys | Robin_Watts: I sent mail about the issue | 00:20.24 |
Robin_Watts | paulgardiner, sebras, tor8: (For the logs) All the commits on robin/master up to 3cfd87d (Don't drop objects if they have been modified) are good to go. | 09:47.31 |
kens | Hmmmm.... | 10:27.38 |
| I just ran a cluster push which only affected pdfwrite, and I get 19 seg faults on files not associated with pdfwrite and 20 diffs in non-pdfwrite/ps2write output. And yes, I did update my repository first | 10:28.40 |
| Someone broke somethng ? | 10:29.07 |
kens | tries a simple up-to-date repository | 10:29.29 |
| OK so an up-to-date repository with no changes from me behaves the same, 19 segfaults and 23 diffs. | 10:46.59 |
| Time for git bisect I think | 10:47.44 |
Robin_Watts | Morning tor8. | 11:58.25 |
| I've got a batch of commits on robin/master for your consideration. | 11:58.50 |
| Everythng up to and including "Don't drop objects if they have been modified" should be good to go. | 11:59.08 |
tor8 | hi Robin_Watts. | 11:59.10 |
| okay, will look | 11:59.23 |
Robin_Watts | The only potentially contentious one is Rework fz_streams | 11:59.33 |
tor8 | the first one must be a rare occurrence, but I can see how it would happen if you're using "borrowed" references from fz_dict_gets | 12:00.25 |
Robin_Watts | tor8: My cleanup code hits it. | 12:00.43 |
tor8 | so to trigger, the best test would be to fz_dict_puts(dict, "foo", fz_dict_gets(dict, "foo")) | 12:00.48 |
Robin_Watts | I have a /Filter [ /Fl ] that I replace with /Filter /Fl | 12:01.56 |
| or /Filter [/Ascii85Decode /Fl] that I replace with /Filter /Fl | 12:02.10 |
tor8 | yeah, that would trigger it | 12:02.18 |
| I've got a similar printf-issue with libjs that I need to solve | 12:05.33 |
| a %g that stops when there are enough digits to losslessly convert back to the same double | 12:05.58 |
Robin_Watts | tor8: Can you adopt my fz_vsnprintf ? | 12:06.18 |
tor8 | so 0.4 would be "0.4" not "0.40000000000000002" | 12:06.22 |
Robin_Watts | ah, mine suffers from that. | 12:06.38 |
| I avoid the 0.40000000 problem. | 12:06.55 |
tor8 | so maywe we should hold off on the printf commit until I've solved it for libjs | 12:06.58 |
Robin_Watts | Can we get the current commit in, and then revise it later ? | 12:07.13 |
tor8 | I'm not thrilled about reusing the system printf for the implementation of our own printf, didn't that have setlocale issues? | 12:07.49 |
| like using ',' as decimal separators etc | 12:07.56 |
Robin_Watts | I fix that. | 12:08.03 |
| I use the same nasty trick that gs does. | 12:08.10 |
kens2 | was about to ask that.... | 12:08.26 |
Robin_Watts | You print "1.5" to a string, then look to see what the second char of the string is. | 12:08.34 |
tor8 | I need to study the stream rework a bit more, but the "premature dropping in pdf_dict_put", "BDC operators get both params" and "don't drop" all LGTM | 12:08.54 |
Robin_Watts | That gives you the char the locale uses for the decimal point. | 12:09.09 |
| I then print, and replace that char. | 12:09.18 |
| The alternative is to swap locales which is horrible. | 12:09.27 |
chrisl | So if you sprinf a string with comma in it? | 12:09.38 |
Robin_Watts | chrisl: The %g (or %f) section is done separately. | 12:10.06 |
chrisl | Cool, sounds sane | 12:10.27 |
kens2 | Yeah, its kind of an ugly hack though | 12:10.42 |
Robin_Watts | kens2: I don't deny that. | 12:10.51 |
kens2 | I'm not complaining, it works.... | 12:11.01 |
| Its just a pity we can't insisit on a 'C' locale when we need it | 12:11.13 |
chrisl | Well, that's where I say glibc is broken..... | 12:11.36 |
Robin_Watts | There are calls to do "printf in this specific locale", but you can't rely on them being there. | 12:11.51 |
chrisl | glibc shouldn't do locales at all..... | 12:12.23 |
Robin_Watts | The alternative is to do our own handling for printing floating point numbers, but that would be painfully slow. | 12:12.28 |
| and/or would build knowledge of the particular fp formats used into the system, which feels bad. | 12:12.51 |
kens2 | chrisl did you say you had cherry-picked Ray's commit f18fac0 "Fix sporadic vertical stripes..." ? | 12:13.27 |
tor8 | Robin_Watts: the libc handling for printing floating point numbers *is* painfully slow | 12:13.38 |
Robin_Watts | tor8: The next commit along the chain is causing differences, and I can't easily bmpcmp it until I get this one in. | 12:13.43 |
chrisl | kens2: yes, I did | 12:13.48 |
kens2 | chrisl you might want to rethink that, its causing a number of seg faults, one with a PDF file, the rest with PCL input | 12:14.16 |
Robin_Watts | tor8: Right, well, we can't reasonably hope to be any faster. And you have to figure that if any platform has dealt with the problem already, we'd be best placed to get it from that. | 12:14.26 |
kens2 | Naturally I can't reproduce the faults locally..... | 12:14.39 |
| I'm just doing a final check, and then a bmpcmp, but it looks quite definitive that its a problem | 12:15.00 |
tor8 | Robin_Watts: I suspect we should just go to the source for all the libc floating point printing/parsing and reuse that: | 12:15.05 |
| http://www.netlib.org/fp/dtoa.c | 12:15.13 |
chrisl | kens2: yeh, I saw that. I'll probably pull it out, but I'll what see what Ray says first | 12:15.17 |
kens2 | I'll finish this up and send an email round to Ray and tech | 12:15.37 |
Robin_Watts | tor8: That doesn't feel like something that fits naturally within mupdf. | 12:16.21 |
tor8 | Robin_Watts: it's two public functions, strtod and dtoa | 12:16.58 |
| sadly, dtoa as implemented isn't part of libc, but strtod is... | 12:17.26 |
Robin_Watts | tor8: That, if I am not mistaken, relies on the specific format of fp numbers used. | 12:17.52 |
tor8 | Robin_Watts: it relies on IEEE floating points | 12:18.21 |
Robin_Watts | "Exactly one of IEEE_8087, IEEE_MC68k, VAX, or IBM should be defined." | 12:18.26 |
| so 4 possible representations. Do you want to get into the business of detecting which one at compile time? I don't. | 12:18.52 |
tor8 | Robin_Watts: no, we strip out the three that don't exist :) | 12:19.11 |
| Robin_Watts: that file is pointed to by the ECMAScript specification | 12:20.12 |
| and is included in various hacked up forms in all of the javascript interpreters (JSCore, spidermonkey, v8) | 12:21.00 |
Robin_Watts | I can't see this as being a good way to proceed (for mupdf itself) at least. | 12:21.41 |
| This isn't speed critical, and even if it was, this isn't going to perform noticably faster than the alternative version. | 12:22.11 |
| My code (reusing the system printf) is far simpler, and more portable. | 12:22.39 |
tor8 | Robin_Watts: it makes us independent of libc vagaries, but no, I'm not thrilled about the prospect of taking on this code either | 12:22.42 |
Robin_Watts | If we cared about speed, then I could understand taking this on, and optimising it. but speed is not what mujs is about, right? | 12:23.25 |
tor8 | Robin_Watts: the benefit of dtoa is we could get the output we wanted ("0.4" rather than "0.40000000002" without trickery) | 12:23.59 |
| speed is not of concern | 12:24.16 |
| the api we want is | 12:24.20 |
Robin_Watts | tor8: mmm. | 12:24.29 |
tor8 | Robin_Watts: there is a much cleaned up variant of dtoa in the plan9 source tree, but that has the BSD advertising clause in the license | 12:27.19 |
kens2 | lunches | 12:34.14 |
henrys | never gives up⦠Did I ask about the first weekend in June? | 13:59.06 |
Robin_Watts | henrys: What days are sabrinas shows? | 14:11.06 |
henrys | there are two problem sabrina and ray_laptop which make that particular weekend inconvenient. | 14:12.08 |
| friday - sunday | 14:12.27 |
Robin_Watts | henrys: We could do a sunday/monday meeting? | 14:12.46 |
| The Euros could fly out on saturday. | 14:12.59 |
| If we do the "come a day early to see miles' place" then we could even come on saturday and do a monday/tuesday meeting. | 14:13.23 |
| a sunday/monday monday meeting would be easier for ray right? And would solve it for you? | 14:14.00 |
| oh, sorry, sunday is still bad for you. | 14:14.22 |
henrys | I think Miles will suggest Denver which is better for you guys anyway, fixes my problem, but I don't know about ray, and I don't want to mess up his kid thing. | 14:14.39 |
Robin_Watts | AIUI, ray has a commitment on friday evening. | 14:15.44 |
| And was planning to catch a late night/early morning plane to SF. | 14:16.07 |
henrys | right miles and I talked about it and ray would be a bit late, fly in in the morning. | 14:16.41 |
| but better for all would be may 31 jun 1 | 14:17.13 |
| does denver save you guys some air time, let me check | 14:17.56 |
| ? | 14:17.58 |
Robin_Watts | I'd have no problem with 31st May/1st June. | 14:18.40 |
| henrys: No direct flights. | 14:19.34 |
chrisl | I find that really surprising...... | 14:20.08 |
Robin_Watts | Actually... BA/American fly direct. | 14:20.10 |
| But not virgin. | 14:20.18 |
henrys | a solid hour and 20 minutes savings to denver - is that something worth pitching to Miles or is it not worth it? | 14:20.22 |
Robin_Watts | 1001 quid. | 14:21.09 |
henrys | that way you could have jet lag and elevation sickness on the same day. | 14:21.11 |
kens2 | 1001 ? Wow | 14:21.23 |
henrys | I'm not familiar with the prices, wow? | 14:22.27 |
kens2 | normally about 600 | 14:22.40 |
chrisl | I really thought Denver would be easy.... | 14:23.00 |
Robin_Watts | Can save money by not flying direct, but... | 14:23.27 |
kens2 | no way | 14:23.42 |
Robin_Watts | 618quid fare, 383 taxes. | 14:23.46 |
ray_laptop | henrys: actually, I can fly even fly in late Fri evening. My work is wrapped up by ~ 5pm, which means I can make a flight after about 7:30pm (allowing for Fri nite traffic to the airport) | 14:26.55 |
| henrys: and that works for me on the west coast. | 14:27.25 |
henrys | Before I send out email does 5/31 and 6/1 work for everyone? That would fix everything also. | 14:28.26 |
ray_laptop | but if the meeting is Sun/Mon (the euros would still stay over Sat) then I am flexible | 14:28.39 |
henrys | kens2, chrisl ray_laptop ^^^ | 14:28.41 |
kens2 | I'll have to check | 14:28.42 |
chrisl | That's fine with me | 14:28.49 |
ray_laptop | henrys: I've already left the house and our master calendar is home. I'll have to let you know | 14:29.25 |
| it is memorial day weekend, but I doubt that affects international flights, and that weekend isn't (AFAIK) a particularly big air travel date domestically | 14:30.24 |
henrys | ray_laptop: I think that is the week before | 14:30.48 |
| ray_laptop: memorial day is the 26th | 14:31.24 |
ray_laptop | henrys: oh, yeah. I forgot the business of moving the holiday to the monday before. | 14:32.14 |
henrys | tor8 is that date good for you? | 14:34.45 |
kens2 | OK apparently I can make the 31/5 and 1/6 at the moment | 14:34.51 |
| I wouldn't mind visiting Denver, never been to Colorado before. Looks expensive though | 14:35.56 |
Robin_Watts | I'd love to visit Denver, but as you say... | 14:36.18 |
| I suspect that changing to Denver would only happen if we change the dates. | 14:36.36 |
henrys | I wonder why that is, less competition? | 14:36.59 |
Robin_Watts | Would that be ski season? | 14:37.14 |
tor8 | henrys: any date is good for me | 14:37.36 |
henrys | Robin_Watts: nope ends in early may | 14:37.36 |
| Robin_Watts: there is high altitude skiing but that doesn't draw a lot of folks | 14:38.26 |
| tor8:all equally horrid? ;-) | 14:39.38 |
tor8 | henrys: denver is awkward to get to. | 14:39.56 |
Robin_Watts | tor8: transfer via London? fly on BA with us? | 14:40.50 |
tor8 | as well ass a fair bit more expensive. the only 1-stop trip (that doesn't have a stop in the US, which I would rather avoid...) I can find is via Reykyavik | 14:40.53 |
henrys | we change it to the 31st and I'll pitch London - we need to meet our new friends at pedro_ and Jogux ;-) | 14:41.19 |
| s/at/: | 14:41.32 |
pedro_ | sounds like a plan :) | 14:41.40 |
ray_laptop | I have a conflict on 31st (actually 29th, 30th and 31st). That's my daughters last show choir show dates | 14:41.57 |
tor8 | Robin_Watts: I don't get the BA trips when searching from CPH, but that would definitely be an option | 14:42.25 |
ray_laptop | (I just called my wife and those are the dates she recalled. I'm going on to the choir website to double check) | 14:42.34 |
| I had to go to the first night show of hers in December, but the Sat night show is better and special for her. | 14:43.35 |
henrys | what is unfortunate is we cannot do weekdays. That's why the planning sucks. | 14:44.43 |
Robin_Watts | henrys: Well, we can do sunday/monday, or (even in this case) monday/tuesday | 14:45.28 |
| Miles invited us to fly out a day early for the next trip and go see his cabin. | 14:46.05 |
| If we do that, we can fly in saturday, see the cabin sunday, and that leaves us having a meeting on monday/tuesday. | 14:46.25 |
| Which avoids your problems with sabrinas show, avoids ray having to get a silly flight etc. | 14:46.49 |
| the only question is, can people manage the monday/tuesday? | 14:47.00 |
kens2 | Well I assume we Euopeans all can | 14:47.28 |
henrys | but I assume all of you don't want to come out early so some would come sunday night? Does that get you the lower airfare? | 14:48.33 |
tor8 | henrys: I think all Euros would have to fly out on saturday for the lower airfare | 14:49.40 |
kens2 | no we need staurday night stay | 14:49.46 |
Robin_Watts | Would any of us *not* be happy to come out a day early? | 14:50.07 |
tor8 | I'm fine with it this time, just don't make it a habit :) | 14:50.24 |
kens2 | fine by me | 14:50.28 |
ray_laptop | The cabin is fun. My kids liked it when we stayed there last summer | 14:50.41 |
tor8 | going up to Miles cabin in the summer sounds like fun, it's been several years since I last was there | 14:50.51 |
ray_laptop | henrys: so Sabrina's show is Sat and Sunday ? | 14:51.12 |
| the river *might* have water in it still | 14:51.44 |
henrys | right I'll get a late night flight sunday or even monday morning | 14:51.45 |
ray_laptop | henrys: if we are meeting Monday, I recommend late Sunday -- you at least get to sleep in a bed for a while before the meeting :-) | 14:52.49 |
| rather than at the meeting | 14:52.58 |
| The choir my daughter is in won 1st place at both of the competitions so far. It's not in the 'advanced' category, but she is excited with how they are doing | 15:02.03 |
| The other more advanced choirs are also doing really well. I'm glad my daughter is not in the choir going to "nationals" in Nashville, however. That competition is INTENSE | 15:04.00 |
Robin_Watts | paulgardiner: Would you have a problem if for the next staff meeting we flew out on saturday, went to Miles' cabin on the sunday, then had the meetings monday/tuesday ? | 15:09.07 |
henrys | oh I forgot about paulgardiner ⦠I just told miles all the euros agree he thinks it a great plan. | 15:13.50 |
paulgardiner | Fine with me too | 15:18.53 |
chrisl | Hmmm, another cluster failure due to "log file(s) too big" on macpro...... | 15:20.59 |
Robin_Watts | tor8: Did you have an opinion on the fz_stream stuff yet? | 15:22.04 |
kens2 | I mailed that to marcos chrisl | 15:22.22 |
henrys | chrisl: a ulimit thing? why only Mac Pro? | 15:22.44 |
kens2 | henrys no idea | 15:22.54 |
chrisl | henrys: no idea..... | 15:23.00 |
| kens2: I see there's now a list of jobs that triggered the problem, which is useful | 15:23.31 |
kens2 | there is ? | 15:23.50 |
chrisl | kens2: I just looked at the report on the regression dashboard | 15:24.24 |
kens2 | I thought that was the large log files | 15:24.59 |
Robin_Watts | temp/tests_private__comparefiles__Bug694353.pdf.ppmraw.300.0 would appear to be the offender. | 15:25.04 |
kens2 | It seems always ot be the one for me | 15:25.22 |
chrisl | I wondered if it was dumping a lot of warnings or something | 15:26.01 |
kens2 | Its possible of course | 15:26.15 |
Robin_Watts | THat's generally what causes log file overflows. | 15:26.54 |
Jogux | I already sent my notes on how to setup a mac build machine and do ios releases of the picsel stuff to Paul, but if anyeone else wants them they're on casper: /home/joseph/ghostdocs-ios-build-release-instructions.docx | 15:55.49 |
henrys | mvrhel_laptop: I'm not sure how we can schedule things such that euros fly saturday without the meeting being monday and tuesday. We have everyone signed up for this can you change something? | 15:56.17 |
mvrhel_laptop | henrys: did you see my email about having the meeting before the 8th and 9th | 15:56.20 |
| henrys: what is different about them arriving saturday for a meeting on monday and tuesday vs having a meeting thursday and friday and heading home sunday morning? | 15:57.11 |
| I guess one extra night perhaps | 15:57.25 |
henrys | ray and I can't do those days | 15:57.40 |
mvrhel_laptop | ah ok | 15:58.08 |
henrys | does your son like meetings ? ;-) | 15:58.48 |
mvrhel_laptop | he has a Concert also on the 10th unfortunately | 15:59.25 |
henrys | okay the gods are against me, back to the original | 15:59.55 |
mvrhel_laptop | henrys: I am sorry | 16:00.19 |
henrys | I understand | 16:00.37 |
mvrhel_laptop | if I could bring him with I probably would but the concert is the issue with that | 16:00.53 |
Robin_Watts | henrys: Would doing the meeting sunday/monday still work? | 16:01.07 |
| That would mean you only missed 1 day? | 16:01.28 |
mvrhel_laptop | robin_watts I would have to do some checking about what I would do with my son on Monday | 16:02.32 |
Robin_Watts | oh, monday is the problem? | 16:02.47 |
mvrhel_laptop | well my wife is gone monday and tuesday | 16:02.59 |
| with my daughter for camp | 16:03.08 |
Robin_Watts | I see. | 16:03.15 |
mvrhel_laptop | and my son would be home alone | 16:03.17 |
| and is not quite old enough for that yet | 16:03.26 |
| I might be able to figure out something but it would take some time | 16:05.44 |
henrys | mvrhel_laptop: what time is the concert? | 16:07.27 |
| I guess we can do the entire plan one week earlier also. | 16:09.23 |
| mvrhel_laptop: does that work? | 16:10.08 |
mvrhel_laptop | that would be much better. doing the 9th and 10th would be a mess for me | 16:10.30 |
henrys | this is my new job now I can see why miles gave it to me. | 16:10.32 |
Robin_Watts | henrys: At some point, it becomes easiest to get everyone to mail their availability in for every day in june :) | 16:10.57 |
mvrhel_laptop | we get back from Japan on the 30th | 16:11.09 |
Robin_Watts | . for OK. x for bad. X for impossible. | 16:11.20 |
henrys | luckily we have no choice about september with print in chicago | 16:11.30 |
Robin_Watts | Then for each person we'd have: ....xXXXXx.....xxXXXXXxx..... etc | 16:11.40 |
| Then we could line 'em all up and look. | 16:11.46 |
henrys | Robin_Watts: actually marcosw recommended setting up a web page. | 16:11.48 |
Robin_Watts | doodle.com? | 16:11.54 |
mvrhel_laptop | henrys: so you are thinking the 2nd and 3rd? | 16:12.46 |
henrys | mvrhel_laptop: yes with optional fly out Saturday for the russian river. | 16:13.11 |
mvrhel_laptop | since I will be flying back from japan on the Friday after being gone for what might be 9 days if we go to korea I will probably miss that | 16:13.54 |
henrys | Robin_Watts: actually the weekend is quite a constraint, we've exhausted everything in June except this last one and the current set dates. | 16:14.33 |
mvrhel_laptop | there is also the 16 and 17th. not ideal for me due to fathers day and my daughters birthday (17th) but I could make it work | 16:16.01 |
kens2 | Its 3x the price for air fares if we don't stay over Sturday night | 16:16.02 |
mvrhel_laptop | henrys: what was the issue with the 14th and 15th? | 16:16.36 |
henrys | mvrhel_laptop:miles son graduates. | 16:16.48 |
mvrhel_laptop | ah ok | 16:16.53 |
| so that probably also kills the 16th and 17th | 16:17.10 |
| with the russian river | 16:17.15 |
henrys | mvrhel_laptop: right, the 2nd and 3rd don't work? | 16:17.39 |
mvrhel_laptop | ok, so the 2nd and 3rd might work, with the russian river as long as miles is up to entertaining after japan | 16:17.49 |
| henrys: I am fine with the 2nd and 3rd | 16:18.36 |
| I will just probably arrive sunday evening though | 16:18.45 |
henrys | mvrhel_laptop: what are the proposed Japan dates - scott might be tired too? | 16:18.46 |
mvrhel_laptop | Japan is fly out 24th and get back home on the 30th | 16:19.11 |
| we may fly out on the 21st though if we are going to korea | 16:19.27 |
| scott was a pilot he would be fine | 16:19.41 |
| or is a pilot | 16:19.46 |
Robin_Watts | scott can sleep through the meeting :) | 16:19.53 |
| how would we tell the difference? :) | 16:20.05 |
henrys | jesus you should be excused from the meeting I've done a meeting as a talking head. | 16:20.14 |
| mvrhel_laptop: ^^^ | 16:20.36 |
mvrhel_laptop | well that would put scott miles and I in the same boat as the rest the euros | 16:20.37 |
Robin_Watts | <tor8>We could just cancel the meeting?</tor8> | 16:20.46 |
| mvrhel_laptop: Flying westwards by 8 hours is easier than flying eastwards by 8 hours. | 16:21.30 |
mvrhel_laptop | yes | 16:21.37 |
| as long as the room is not dark, I will be fine | 16:21.46 |
henrys | are all euros good with pushing back a week to the 2nd and 3rd flyout Saturday? | 16:22.27 |
kens2 | Provisionally yes, I would need to check that too | 16:22.46 |
Robin_Watts | I seem to be. | 16:23.40 |
henrys | chrisl, paulgardiner ? | 16:23.56 |
paulgardiner | I believe I am | 16:24.06 |
henrys | I will set up doodle.com or something to make this simpler in the future. | 16:24.22 |
paulgardiner | It's within my original "leave it free" zone | 16:24.36 |
chrisl | That's fine by me..... | 16:24.50 |
paulgardiner | Do we have the Nas we believe to contain the wiki backed up? | 16:26.25 |
Robin_Watts | marcosw: See paulgardiner's question | 16:26.50 |
paulgardiner | marcosw: we wish to fiddle with it, but daren't until we know there's no risk of loosing it | 16:27.39 |
henrys | paulgardiner: I just spoke to marcosw by email, shoot him a mail he may be away from irc. | 16:29.27 |
paulgardiner | henrys: will do | 16:29.53 |
tor8 | mvrhel_laptop: flights east are almost always overnight flights, so thurs/friday means waiting around all sunday without a hotel room for the flight | 16:31.22 |
mvrhel_laptop | tor8 roght | 16:33.35 |
| right | 16:33.37 |
| brb | 16:36.35 |
| robin_watts: what 16 bit device do we have to check out the issue from Customer #885 | 17:00.57 |
Robin_Watts | mvrhel_laptop: Bug number ? | 17:01.15 |
mvrhel_laptop | no bug yet. email to support | 17:01.29 |
| I saw it come in and I am sure it will land in my lap. | 17:01.58 |
Robin_Watts | mvrhel_laptop: No idea. | 17:02.57 |
| He does say "You might not have been noticed this issue as no current GS driver supports 16 bits depth in a functional way" | 17:03.29 |
mvrhel_laptop | he said this, and it actually surprised me | 17:03.52 |
| I do know that I have an open bug to add 16bit support to psdcmyk | 17:04.09 |
| which might be a good idea | 17:04.14 |
chrisl | tiff64nc | 17:04.32 |
mvrhel_laptop | ah | 17:04.39 |
| yes | 17:04.40 |
Robin_Watts | mvrhel_laptop: If it was me, I'd be hooking the lcms calls. | 17:04.42 |
mvrhel_laptop | good | 17:04.42 |
| hooking? | 17:04.52 |
Robin_Watts | breakset. | 17:04.59 |
mvrhel_laptop | yes | 17:05.02 |
Robin_Watts | his file only contains 1 color, AIUI :) | 17:05.14 |
mvrhel_laptop | yes. Let me see what tiff64nc gives me. thanks chrisl and robin_watts | 17:05.36 |
chrisl | mvrhel_laptop: since I read it like they were happy before the ICC stuff, and they are in a hurry, would it be an option to point them at UseFastColor as a temp workaround? | 17:07.04 |
mvrhel_laptop | chrisl: good point. let me first dig into this and see how both behave | 17:07.31 |
Robin_Watts | mvrhel_laptop: It's possible that our existing profiles are off fractionally at the top and bottom end. If they were correct in 8 bits, we'd never have noticed. | 17:08.36 |
mvrhel_laptop | right | 17:09.01 |
| hmmm this device must require something special. running it gave me "Invalid compression setting for this bitdepth" | 17:11.17 |
| followed by a range check error | 17:11.29 |
Robin_Watts | mvrhel_laptop: What command line ? | 17:11.32 |
mvrhel_laptop | -sDEVICE=tiff64nc -o ./myoutputs/output.tif -f ./myinputs/input.pdf | 17:12.36 |
Robin_Watts | gs/debugbin/gswin32c.exe -sDEVICE=tiff64nc -o out.tif gs/examples/tiger.eps works for me. | 17:14.01 |
mvrhel_laptop | hmm weird | 17:14.19 |
| fails at line 158 in gdevtifs.c | 17:14.34 |
Robin_Watts | pulling/rebuilding now. | 17:14.52 |
mvrhel_laptop | tiff_compression_allowed does not have a test for depth == 16 | 17:15.30 |
| maybe I am out of date | 17:15.42 |
chrisl | Hrm, I may have been the last person to touch that :-( | 17:16.23 |
| mvrhel_laptop: this is probably what's needed: http://pastebin.com/5jim7xT8 | 17:19.30 |
mvrhel_laptop | so what compressions should be allowed with 16 bit data? | 17:19.34 |
| oh | 17:19.36 |
| great | 17:19.50 |
| let me give that a try | 17:20.01 |
chrisl | I'm in the middle of something else just now, so I'll test it properly when I dig my way out of where I am..... | 17:20.24 |
mvrhel_laptop | chrisl: no problem | 17:20.40 |
| thank you | 17:20.44 |
chrisl | mvrhel_laptop: in the interim the default of "no compression" is definitely valid | 17:21.03 |
mvrhel_laptop | yes | 17:21.08 |
henrys | paulgardiner: is there api documentation for using the embedded general stuff? | 17:24.33 |
chrisl | thinks henrys is being terribly optimistic...... | 17:25.00 |
kens2 | goodnight all | 17:25.15 |
paulgardiner | Not that I've seen so far. I'll take a look | 17:25.24 |
Jogux | the header files are pretty devoid of comments :( | 17:26.19 |
paulgardiner | I guess it wouldn't be in the epage repo. Might be hiding on the NAS | 17:26.20 |
mvrhel_laptop | ok so tiff64nc shows the problem, so I should be able to fix this | 17:28.51 |
Robin_Watts | It could be in epage-docs ? | 17:29.50 |
| Can't obviously see it there. | 17:30.41 |
mvrhel_laptop | so lcms is returning 65307 from some reason.... | 17:31.52 |
| s/from/for/ | 17:32.00 |
| need to dig into the profile robin_watts like you said | 17:32.12 |
Jogux | henrys: I can't find anything obvious in git or the old CVS | 17:33.27 |
mvrhel_laptop | ok this is likely an issue with the gray to K mapping | 17:39.52 |
| ok so the gray to k profile looks ok. Outputs 65535 for black input. looks like I will need to dig through lcms | 17:49.01 |
ray_laptop | mvrhel_laptop: I am seeing a really strange threshold array from gx_ht_construct_threshold. Do you have any time to look into this with me. | 17:49.48 |
| mvrhel_laptop: in particular I am seeing the t_level go from 4 to 9 then DOWN to 7 :-/ | 17:50.19 |
| the original 'threshold_from_order' doesn't have any of the delta, delta_sum or t_level_adjust "magic" | 17:51.29 |
mvrhel_laptop | ray_laptop: in the middle of debugging this right now. let me get to a stopping point. ray_laptop: that code was all "borrowed" from your code in that tiff device | 17:52.08 |
| if I am thinking of where you are | 17:52.41 |
| correctly | 17:52.44 |
| let me step through lcms first though. half way through | 17:53.00 |
ray_laptop | mvrhel_laptop: no problem. Please finish your debugging and get to a stopping point. Please call me (I may miss IRC) | 17:53.40 |
mvrhel_laptop | ok will do ray_laptop | 17:55.01 |
| crap so the table entries in the link that lcms created appear to be wrong. so this is going to take a bit more digging | 18:04.15 |
Robin_Watts | mvrhel_laptop: Are you testing with lcms 2.6? | 18:13.11 |
mvrhel_laptop | robin_watts no | 18:17.28 |
Robin_Watts | the release of 2.6 came out today. | 18:17.49 |
| if it looks like a bug *in* lcms, then perhaps it might have been fixed? I doubt it, frankly though. | 18:18.15 |
mvrhel_laptop | robin_watts first I need to find where things are going wrong in the current code. since that is what we will have to supply to the customer | 18:19.19 |
| then I will show that to marti and let him decide if it was fixed or not | 18:19.44 |
Robin_Watts | mvrhel_laptop: well, we could supply the customer with a patch that upgrades lcms. | 18:19.45 |
mvrhel_laptop | yes | 18:19.49 |
Robin_Watts | but I admit that's unlikely to be the problem. | 18:19.55 |
mvrhel_laptop | that is my plan | 18:19.57 |
| robin_watts: you don't think lcms is the issue? | 18:20.08 |
Robin_Watts | oh, I see. | 18:20.11 |
mvrhel_laptop | I am looking at the profiles and they are fine | 18:20.16 |
Robin_Watts | You think it's going to be an unfixed flaw in lcms. Got it. | 18:20.27 |
mvrhel_laptop | I can see that the smashed lut that lcms is using is returning an incorrect value | 18:20.42 |
| I am now trying to dig through the lut creating in lcms to see where it is going wrong | 18:21.06 |
| this part of the code is difficult though | 18:21.12 |
| but I will find it | 18:21.22 |
Robin_Watts | Let's hope it's not because of our use of lcmsChangeBufferFmt. | 18:21.33 |
mvrhel_laptop | since it is a "black point" I suspect there is some special stuff he is doing | 18:21.48 |
| robin_watts: I don't think that is the case | 18:22.08 |
| I created the link as 2 bytes to begin with | 18:22.25 |
Robin_Watts | you're probably right of course. I had forgotten the blackpoint voodoo. | 18:22.32 |
mvrhel_laptop | I am seeing the wrong value in the actual table | 18:23.06 |
| the interpolation in the table is OK. | 18:23.16 |
| hence the link creation is the issue | 18:23.23 |
| he has some special table for the case when we are going from 1 input to N output | 18:23.51 |
| which this is. and for the black colorant in the table the output is not ffff in the table | 18:24.19 |
| need to head off to lunch. bbiab. there is some weird stuff going on in the black point compensation that makes no sense. I may end up disabling BPC when we are doing the gray to black mapping | 18:41.09 |
ray_laptop | mvrhel_laptop: I took out all the stuff in construct_threshold that I didn't understand (that wasn't in my original code) and it is working for me now. That was why the 4-bit (GrayValues=16) looked so funky | 18:42.13 |
| also now, the stocht.ps (BNM) looks good as well. | 18:45.41 |
| mvrhel: for the logs. I need to understand what that code was trying to do, but is not urgent (now). | 18:46.11 |
henrys | bbiab | 19:14.58 |
Robin_Watts | tor8: ping | 19:24.42 |
mvrhel_laptop | robin_watts: so turning off black point compensation fixes the issue | 19:38.57 |
Robin_Watts | ok. | 19:39.09 |
mvrhel_laptop | I am going to make this the default when we are doing our special gray to K ICC profile | 19:39.10 |
| so no changes to lcms just to gs | 19:39.18 |
Robin_Watts | cool. I don't understand the issues here, so I can't contribute anything :) | 19:39.40 |
mvrhel_laptop | I am going to pass the profiles to marti and let him sort out the issues that he has with this | 19:39.47 |
| The black point compensation stuff that he has in there is a little messy for me to to step through now | 19:40.23 |
| chrisl: for the logs, I tested pack and lzw with tiff64nc and they work fine. I am just going to commit the fix you gave me | 19:54.08 |
| ray_laptop: just pushed a cluster test on the lcms thing. do you want to talk about halftone now? | 20:13.38 |
ray_laptop | mvrhel_laptop: OK | 20:13.57 |
tor8 | Robin_Watts: yes? | 20:13.59 |
ray_laptop | mvrhel_laptop: in my git repo, I pushed the code that has the modified (#if 0) in gsht.c | 20:15.28 |
tor8 | Robin_Watts: the fz_stream rework LGTM | 21:16.00 |
| the ideal factor one I'll have to take your word for | 21:16.21 |
| Robin_Watts: not thrilled about decompressing the images twice though (if my reading of the 'fix operator buffering of inline images' is not way off target) | 21:20.22 |
| also, pdf_load_compressed_inline_stream leaks streams on errors? | 21:21.28 |
| I'd like to have a stab at vsnprintf myself; I've got some ideas there but if it's blocking you commit as is and I'll rework from there. | 21:22.07 |
SpNg | Is there a way to output an eps to a jpg with a background color using ghostscript? | 23:14.24 |
| Forward 1 day (to 2014/03/18)>>> | |