IRC Logs

Log of #ghostscript at

 <<<Back 1 day (to 2015/05/04)20150505 
Robin_Watts kens: We have a 3 year code sharing cert.08:58.53 
kens Cool!08:58.58 
  WHen you say 'code sharing' does that mean we cna sign devices and stuff ?08:59.20 
Robin_Watts code signing, sorry.08:59.29 
kens :-D08:59.35 
  So we cna sign GSView now, and ghostpdl.inf, good news08:59.57 
Robin_Watts actually, it seems that joann is having some trouble with the site, so though we have bought it, she can't log in to get it :)09:00.21 
kens Oops oh dear09:00.29 
  Well I guess there's no desperate rush09:00.36 
Robin_Watts I'll forward details as soon as it shakes itself out.09:00.38 
kens Hm, does that mean I just got the job of signing the .inf file ? :-)09:00.57 
Robin_Watts I was hoping so :)09:02.07 
kens Well I guess I don't have to worry about it until the next release09:02.24 
  Currently trying to figure out what I've broken in the GC this tme09:02.42 
Robin_Watts Joann got a receipt webpage, and it says "Please click on the link below to complete the order process".09:10.45 
  So she printed the webpage out, scanned it as a PDF and mailed it to me.09:11.00 
kens Ah....09:11.10 
Robin_Watts I'm not sure how I'm supposed to click on the link :)09:11.11 
kens Grr, both devices are pointing to the same stype, I was sure I'd fixed that :-(09:23.51 
  Sigh, missed a copy.....09:25.15 
Robin_Watts ok, I'm in.... do we want an SHA-1 1or SHA-2 hash to be used when signing the certificate ?09:29.18 
kens return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /*09:29.31 
  default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default case */return 0; /* default cas09:29.32 
  e */return 0; /* default case */I htnk SHA-1 is deprectaed isn't it ?09:29.32 
  Miranda having a bad day.....09:29.53 
  As I was trying to say, I tink SHA-1 is deprecated isn't it ?09:30.04 
  Looks like we should use SHA-209:31.02 
  SHA-2 sounds good to me.09:44.44 
kens Yeah if MS is going to deprecate SHA-1 it as well I think its important that we use SHA-209:45.45 
  Oooh, down to 4 seg faults, and a handful of diffs, that's a big improvement10:11.40 
tor8 Robin_Watts: found a serious bug in fz_pack_path which caused the annotation crashes from the email10:20.01 
Robin_Watts tor8: crashes?10:20.17 
tor8 out of memory crash10:20.24 
Robin_Watts tor8: easily fixable?10:20.29 
tor8 typo in fz_malloc_array10:20.31 
  it's allocating path_len * path_len * sizeof(float)10:20.39 
Robin_Watts D'Oh!10:20.45 
  bloody hell.10:20.54 
  And yet I was still seeing significant memory savings.10:21.04 
tor8 my review process has failed :(10:21.07 
  the test file from the user has a path on the page which is enormous10:22.04 
  50000 points or so10:22.14 
  moveto lineto closepath moveto lineto closepath with a single fill at the very end10:22.41 
Robin_Watts yeah, I was feeding in the file that was taking 7.5Gig or something10:22.53 
  and torture-test.10:23.07 
  Well, regardless, it's got to be good news now it's fixed.10:23.36 
  Thanks for spotting that.10:23.39 
tor8 wonder how things would fare now that it's been fixed10:23.55 
  commit on tor/master10:23.57 
Robin_Watts tor8: Looks great.10:29.39 
  stupid bloody cluster insists on building gxps for a gs test.10:30.34 
chrisl It has to10:30.49 
Robin_Watts I've broken gxps. I know I've broken gxps, and pcl etc, that's why I just wanted to test gs.10:30.55 
  chrisl: why for so now?10:31.00 
chrisl xpswrite10:31.06 
Robin_Watts oh, for testing xpswrite.10:32.15 
kens We have nothing else to read and render XPS files with10:32.32 
chrisl We should tweak it so if you're not testing xpswrite, it doesn't bother with gxps10:32.37 
Robin_Watts chrisl: I started to fiddle with refactoring for the language switch stuff while I was in in London on saturday.10:33.32 
  and i spent yesterday bashing it through a bit more.10:33.49 
  Those are some (brief) notes.10:34.03 
chrisl How do you define 'topmost language'?10:35.47 
Robin_Watts chrisl: in a language switch build, the topmost language would be the switcher.10:36.10 
  in gs builds the topmost language would be gs, etc.10:36.37 
chrisl That's going to be a *big* change for how the Ghostscript API works10:40.00 
Robin_Watts chrisl: Really?10:40.15 
  The attraction of this is that it's invisible from outside the ghostscript API.10:40.31 
  unless we are talking about different APIs.10:40.48 
chrisl So you'd still envisage those being in place - okay, that makes sense10:41.22 
Robin_Watts chrisl: Yes.10:41.36 
  The big change is that this blows away the pl_universe stuff.10:41.48 
chrisl Yes, that's pretty flawed. IMHO....10:42.14 
Robin_Watts chrisl: As soon as I get something worth showing you, I'll let you rip it to shreds :)10:43.15 
  I'd like that to be "when I get a first test to run through the cluster".10:43.33 
chrisl So each language would have its own 'gs_main_instance', then an "umbrella" one for language switching10:43.44 
sebras tor8: no errors when viewing the .epub I complained about yesterday!10:45.01 
  tor8: good work! :)10:45.05 
Robin_Watts chrisl: yes.10:49.32 
  The wrapper would start up the language switcher language, and (somehow) tell it what sub languages to use.10:50.03 
  the language switcher would then look after the gs_main_instance's for them.10:50.31 
  any calls made to the language switcher would be passed through appropriately.10:50.51 
  I'm envisaging that the language switcher might have the freedom to deactivate/closedown languages if memory got tight.10:51.38 
chrisl Robin_Watts: For language choosing, I think the basic idea of the current code is about the best we can expect to do - with the exception that I'd make the "magic string" an array rather than a single string10:51.59 
Robin_Watts chrisl: I haven't looked at the current code for ls in any detail.10:52.25 
  I was assuming that it would 'peek' at the input and spot patterns that would suggest which language to use ?10:52.46 
chrisl Yes, exactly, and each language has a single pattern to peek for - I'd like to allow for multiple patterns10:53.24 
Robin_Watts chrisl: Sure.10:53.57 
  Or, better, to call each language to say "give me a score that represents how likely this input is to be in your language"10:54.20 
  That way knowledge about each language is kept in each language.10:54.40 
  That's what SOT and MuPDF do.10:54.47 
chrisl That could work....10:55.42 
tor8 sebras: good!10:56.09 
Robin_Watts tor8: So are we looking at 1.7b ?10:56.30 
chrisl Robin_Watts: My concern is that I don't like the way the current code just uses "%" to spot Postscript or PDF....10:56.43 
Robin_Watts chrisl: ew. indeed.10:57.07 
chrisl One problem is that PCL will basically always match....10:57.32 
sebras tor8: I just ran through my library of ebpus using mudraw and got a few errors.10:58.39 
  tor8: I'll have a look at those later (need to run an errand)10:58.52 
Robin_Watts chrisl: yeah. We have scope to do better by building the assessment into the language.10:59.13 
  peek at the first (say) 4k, then do a quick parse forward.10:59.46 
chrisl Robin_Watts: no, that's not what I mean: PCL has to literally match anything you throw at it11:00.10 
Robin_Watts yeah, I see what you mean.11:00.28 
chrisl Even if you throw random bytes at it, it's still "valid" PCL....11:00.44 
kens PCL can take just text, so there's no way to spot PCL really11:00.46 
Robin_Watts So it's a question of the other languages having to spot their own stuff, and PCL having to accept anything that's not recognised.11:01.07 
kens I would say so yes11:01.22 
chrisl Yes, I think that's the best we can do11:01.25 
kens If you exspress it as confidence PCL would always say, 50% for example11:01.41 
Robin_Watts We have a similar situation in SO, I believe, where the text agent can accept anything.11:01.44 
kens Then other languages could have a go, if they beat that, then its up to them11:01.56 
Robin_Watts where we do... exactly what kens just suggested.11:01.58 
kens heads for lunch11:02.31 
Robin_Watts Stupid bloody linker.11:02.42 
chrisl We could also have a fallback, where if we think it's Postscript, if the interpreter throws an error, try it as PCL (at least one printer I've worked with worked that way)11:03.01 
Robin_Watts I have a gsinit2.c that has just 2 functions in it, which use "gs_language". Neither of those functions are referenced from the pcl build, and yet, the linker is complaining that gs_language is missing when it links.11:03.37 
  If I comment out those 2 functions, it links fine.11:04.02 
  so it clearly doesn't need them.11:04.08 
tor8 Robin_Watts: yeah, a 1.7b by the end of the week?11:22.25 
Robin_Watts tor8: yeah.11:22.39 
JeeWee_ goodday to everyone12:16.56 
Robin_Watts JeeWee_: Welcome. If you have a question, ask it, don't wait to get a response before asking, or you could be hanging around for a while :)12:18.05 
JeeWee_ fair enough! :)12:18.21 
  I have just started learning to write some postscript12:18.31 
  and I am wondering why my code is not working12:18.38 
  im certain its something trivial 12:18.45 
  let me throw it online real quick12:18.51 
Robin_Watts use or something rather than pasting it here.12:19.07 
JeeWee_ yes12:19.42 
  as you can see its extremely simple12:19.57 
  just started doing some tutorials12:20.03 
kens What did you want that to do ?12:20.27 
JeeWee_ write a function to draw a box12:20.42 
  but even without it using any arguments12:20.56 
  it errors out on me12:20.59 
  i tried that code on its own.. i.e. not as a function12:21.08 
  and then it draws the box12:21.13 
  but as soon as i try to make it into a function12:21.20 
chrisl Remove the "end"12:21.21 
kens Yeah12:21.26 
  end closes teh current dictionary12:21.39 
  and removes it from teh dictioanry stack12:21.53 
JeeWee_ facepalms12:21.53 
chrisl The Blue Book can be useful.... "The PostScript Language Tutorial and Cookbook"12:22.33 
JeeWee_ i have it :) Currently doing "learning postscript by doing" from andre heck12:23.02 
  where i write postscript and preview it directly with gsview12:23.17 
  thankyou for the link though12:23.38 
chrisl And what you have there is a PS file, not an EPS12:23.47 
JeeWee_ i assume there is a specific reason you mention that12:24.32 
  could you elaborate?12:24.39 
chrisl EPS file aren't supposed to use certain operators - in this case, showpage12:25.04 
JeeWee_ ok12:25.27 
  i am not familair yet with the differences between them12:25.38 
  except for the fact i would like to end up, eventually, with eps files12:25.49 
  i am not learning eps so much as a programming language12:26.12 
chrisl Basically, EPS is intended to be "placed" inside another PS job (as a logo, or similar), so doing something like emitting the page would be bad12:26.38 
JeeWee_ but im trying to learn it so i can generate graphics without having to depend on a vector drawing program such as ilustrator12:26.42 
  yes.. hence the "encapsulated" part12:27.17 
  i must admit i am wondering wether I have bitten of more then i can chew on though12:27.41 
  the entire way of working with postfix notation, the stack etc12:27.59 
  seems ancient to me12:28.06 
chrisl Just takes some practice - and a warped mind.....12:28.25 
JeeWee_ haha ^_^12:28.31 
  thanks for not taking offense and yes.. it does feel that way12:28.40 
  i just hope i can generate ps files that are good enough for production12:29.07 
  i opened an eps file generated from illustrator and there is sooo much junk in there12:29.24 
  it looks like 90% in the illustrator eps is function definitions12:29.38 
  its amazing12:29.40 
  there is even an xml embedded in there12:29.48 
  pretty much impossible to reverse engineer12:30.01 
chrisl Yes, that's true! Fonts will be your biggest challenge for portable Postscript12:30.08 
JeeWee_ yeah im pretty afraid of fonts12:30.24 
chrisl Oh, I have to head out.... back in a couple of hours12:30.27 
JeeWee_ ok thanks for you help so far!12:30.37 
Robin_Watts JeeWee_: Any reason you aren't generating PDF?12:31.07 
JeeWee_ Robin_Watts: not really no12:31.24 
Robin_Watts *Much* simpler.12:31.29 
JeeWee_ there is a good chance we will convert the ps to pdf in the end12:31.39 
  Robin_Watts: forgive my ignorance12:31.48 
Robin_Watts And PDF is graphically more powerful (transparency etc)12:31.57 
JeeWee_ but is a pdf not just a container for postscript?12:32.00 
Robin_Watts JeeWee: No.12:32.07 
kens : Definiely nt12:32.12 
JeeWee_ today i learned i was wrong12:32.18 
  ok then12:32.23 
Robin_Watts Postscript is a programming language. PDF is not.12:32.37 
kens PostScript is a programming language, PDF is not, it is a page description language12:32.37 
JeeWee_ im going to ask a really open question then12:32.45 
kens No control operators in PDF for example, such as while, repeat loop etc12:32.59 
JeeWee_ what do you suggest i do? is PDF a language?12:33.01 
kens PDF is a page description language12:33.08 
  Originally PS and PDF had the same graphcis model, PDF has been extended since then12:33.22 
  PDF now includes transparency whcih PostScript does not (mostly)12:33.39 
Robin_Watts JeeWee_: PDF files contain (at their heart) streams of page marking operators.12:33.50 
JeeWee_ transparancy is not important to me at this stage12:33.56 
Robin_Watts 'move here, draw here', etc.12:33.57 
  'Put this text here in this font, in this colour'.12:34.11 
JeeWee_ however if PDF is simpler (and i know its more modern) it seems to me thats the way to go12:34.14 
kens Because the original graphcis model is the same, and PDF is a superset, its possible to take a PostScript program and convert it to a PDF file.12:34.15 
  It depends how you define simpler.12:34.26 
JeeWee_ kens: ah! :)12:34.31 
kens PDF is a binary structured format, PostScript is an unstructured program12:34.44 
JeeWee_ ok let me clarify12:34.45 
  i am not at all in terested in programming in postscript12:34.53 
  i want to define graphics for production12:34.59 
  from what you have told me that means i should be working with pdf12:35.09 
  not ps12:35.11 
Robin_Watts Yes.12:35.15 
JeeWee_ this chat is very clarifying12:35.26 
kens The *content* of a PDF file is easier since it is merely descriptions, not a program, however you need to create the structures that define the PDF file, such as the cross-refernce table for example12:35.29 
Robin_Watts There is an initial hurdle with using PDF in that you will need to generate the PDF file structure (as opposed to PDF which is just a stream).12:35.56 
  But that's not such a big deal.12:36.13 
kens THe PDF structure isn't that complex12:36.17 
JeeWee_ Robin_Watts: i asssume i could use libraries for whatever programming/scripting language i end up using12:36.26 
Robin_Watts JeeWee_: You certainly can.12:36.34 
JeeWee_ which will probably streamline that process to a huge extend12:36.36 
kens Don't use Cairo though, unless its suddenly improved a lot12:37.06 
JeeWee_ ^_^ i am certain there are good and bad libraries.. i will put that one on my bad list12:37.34 
  so i assume i should start working my way through this then? > 
kens If you want to understnad the structure of a PDF file, yes12:38.11 
  Much of it will not be of interest to you,probably12:38.23 
  Things like forms, annotations, possibly security etc.12:38.45 
JeeWee_ well its a reference so i need to have it anyway12:38.51 
  is there something like a pdf cookbook?12:38.59 
  similair to what i was using for postscript?12:39.08 
kens Sure, and if you intend to work with PDF files its good to have at least an outline idea of how they are constructed12:39.24 
  THere is no cookbook though12:39.32 
JeeWee_ agreed12:39.33 
  ill find something though12:40.02 
kens There is a great szeries of PostScript and PDF articles on John Deubert's site, just a second12:40.03 
JeeWee_ thankyou12:41.20 
kens Other than that, take a PDF apart and see what;s in it.12:41.39 
  FOr a simple start, take the PS file you have, run it through GS with :12:42.06 
  -sDEVICE=pdfwrite -o out.pdf -dCompressPages=false12:42.06 
  That will get you a PDF file you can read (uncompressed pages)12:42.18 
  And you already know what the content should do12:42.30 
JeeWee_ that is true!12:42.36 
kens You can also use MuPDF to uncompress a PDF file, though I forget the magic runes of the top of my head, RObin probably knows them12:43.10 
Robin_Watts mutool clean -difggg in.pdf out.pdf12:43.24 
JeeWee_ thankyou12:44.58 
  time for me to read now12:45.04 
kens "Enjoy".....12:45.15 
JeeWee_ hahaha12:45.27 
  one last question12:45.30 
  which i know im going to regret12:45.34 
kens :-D12:45.38 
JeeWee_ I have no experience with PDF at this point12:45.52 
  is it realistic to expect to be creating production ready pdf files in say.. a week?12:46.12 
kens Probably not no12:46.21 
  Unless you use some kind of librry to do most of the work in creatng the file12:46.36 
Robin_Watts JeeWee_: From scratch? No. Using a library? Maybe.12:46.37 
JeeWee_ fair enough12:46.52 
Robin_Watts What platform are you working on?12:47.13 
JeeWee_ Robin_Watts: you are going to laugh but that is still to be decided12:47.25 
  in my dreams i'd love for it to be web based12:47.34 
Robin_Watts And what are you wanting to generate from? Is this content you already have in an application?12:47.36 
JeeWee_ say python and django12:47.47 
Robin_Watts django is a text handling thing, right?12:48.02 
JeeWee_ Robin_Watts: 
  and my goal at first would be to mutate a pdf12:48.34 
  for example turn a single element into a strip of x meters long12:48.54 
  in the end i would like to create content from scratch12:49.03 
Robin_Watts So, it's something that you write websites/webapps in ?12:49.19 
JeeWee_ currently i have created a tool within adobe illustrator through extendscript which "draws" what we need12:49.28 
  however that environment is "locked" obviously12:49.40 
  Robin_Watts: correct12:49.43 
Robin_Watts 'element' in this case, being ?12:49.55 
  an input PDF? or an input image? Or a series of drawing commands that you have as coords etc ?12:50.17 
  an element being a single description of whats in that pipe12:50.39 
  Robin_Watts: input eps or pdf and yes also a series of drawing commands12:51.04 
  the current process uses existing elements which get embedded and some elements are programatically drawn12:51.29 
  but as i said thats all within the adobe illustrator environment12:51.47 
Robin_Watts JeeWee_: One possibility would be to use the ghostscript library.12:51.51 
JeeWee_ so created a pdf is as easy as calling the saveas function12:51.56 
Robin_Watts Start gs, using the pdfwrite device, and then send a sequence of PS commands to render the required graphics.12:52.53 
  The pdfwrite device in MuPDF isn't quite in advanced enough state for that to be an option.12:53.24 
JeeWee_ silly question perhaps12:54.02 
kens If you are going to use EPS, then you need a PostScript interpreter12:54.09 
JeeWee_ why would i not use something like this: 
  since ill be writing in python for django anyway?12:54.30 
Robin_Watts No reason at all.12:54.41 
JeeWee_ ok. just checking :)12:55.02 
kens A library is a library, if it does what you need, then use it :-)12:55.21 
JeeWee_ time to brush up on my python, learn django and then learn pyx12:56.37 
kens Wow, I htnk I just got a clean run off the cluster :-O12:56.44 
JeeWee_ haha ^_^ (/me cries softly)12:56.48 
kens chrisl changing so that each device is registered with the garbage collector does (after a lot of tinkering) seem to be the right solution. Still some tidying up to do I htink, but it looks like its actually working now12:58.16 
JeeWee_ thank you all for your help again12:59.16 
kens you're welcome12:59.24 
Robin_Watts chrisl: OK, I got a run through that died on the gxps tests, but has passed all the others.13:06.25 
  Commit on robin/master for you comments.13:06.40 
  And from anyone else interested, obviously.13:07.03 
kens Drat, I just realised chrisl's last commit is showing me all kinds of diffs. I'd better rebase my brach13:12.22 
henrys looks like Robin_Watts and marcosw can save their sunlight now with Musk's new battery ...14:17.21 
kens The reviews aren't flattering about it14:17.35 
henrys kens: yeah and I don't think folks understand how big the storage problem really is...14:19.10 
Robin_Watts henrys: Seems like a boondoggle to me. As ever, maybe the second version will be worthwhile.14:19.11 
kens henrys the main problem sedems to bt eh max 2KW drain14:19.33 
  Which means you can't actually draw that much from it, you'd still need the grid14:19.58 
  Vanadium is a MUCH smarter idea than lithium.14:20.14 
  I'd be much more interested to know how much energy can be stored (and for how long) in one Williams flywheels.14:20.52 
  s/one/one of/14:21.01 
henrys this is an old article but most of it still ringst true: 
Robin_Watts henrys: Dunno if you read the logs from earlier.14:26.03 
  I've been playing with the start of a refactor of some of the code with a view to language switching.14:26.31 
  I've got a first version up on robin/master now.14:26.47 
henrys Okay I'll read that and the logs.14:27.00 
Robin_Watts gs runs fine, but pcl etc are all broken, cos I need to redo the pl_universe stuff.14:27.07 
  The idea is that the gsapi interface will continue to work as is.14:27.33 
henrys great I'll read it after the meeting, the only customer concern I know about it norbert backwards compatibility wise.14:28.23 
Robin_Watts henrys: norbert uses a vanilla PCL build, right?14:28.39 
henrys Robin_Watts: yeah but I think he's hacked on the "Universe" stuff.14:29.16 
  Robin_Watts: we'll just tell him and we can help if he needs anything.14:29.44 
chrisl I thought they used a modified language_switch for PCL and XPS14:29.46 
henrys it's fine14:29.47 
  yeah he has pcl, xps and pxl going.14:30.10 
Robin_Watts henrys: all from us?14:30.21 
rayjj Robin_Watts: one of the issues was the 'push' vs. 'pull' difference. GS pulls data from the file/stream you provide. The pl approach pushes buffers to the language.14:30.42 
henrys Robin_Watts: yup and postscript from adobe, I imagine that is completely separate.14:30.50 
Robin_Watts rayjj: I've not looked at anything at that level yet.14:31.10 
chrisl rayjj: I thought both did both push and pull14:31.14 
rayjj gs can be run in a 'push' mode with the run_string_continue approach14:31.39 
henrys right the string stuff is really the same thing.14:31.53 
  meeting started topic 1: language switching.14:32.13 
  we'll all read Robin_Watts twiki and comment, agreed?14:32.47 
chrisl Clearly Robin_Watts has volunteered for that, so I'm off the hook....14:32.48 
marcosw morning14:32.50 
Robin_Watts henrys: The twiki is really just some notes - there isn't a hell of a lot there yet.14:33.13 
henrys maybe marcosw will buy the big battery ... he's a californian14:33.19 
marcosw I have two big batteries already, they are called electric cars.14:34.33 
rayjj marcosw: portable batteries :-)14:34.55 
mvrhel_laptop monring14:35.19 
henrys marcosw: hah I bet they look nicer than the new battery too.14:35.47 
fredross-perry morning14:36.01 
rayjj the thing that irks me is my hybrid (lexus rx-400h) won't turn on the lighter (aux power) sockets unless the car is in 'start' mode14:36.07 
Robin_Watts rayjj: That's the case with most cigarette lighters in cars, right?14:36.44 
henrys rayjj: is there somthing wrong with the string stuff I was hoping to have "push" for all the languages.14:36.56 
Robin_Watts Even with petrol ones, you need to turn the key one click.14:37.00 
henrys ?14:37.00 
rayjj and it runs only off the little 12v lead acid battery, not the 288v "main" battery14:37.03 
  henrys: no, it works, but for small buffers it's sort of inefficient since it has to 'bubble up' the e_NeedInput error at the end of the string14:38.01 
chrisl henrys: "push" won't work well for PDF nor XPS14:39.14 
rayjj henrys: and, of course, it doesn't provide random access at all, so PDF and XPS require temp file buffering14:39.16 
chrisl :-)14:39.35 
henrys mvrhel_laptop, fredross-perry : the home office is working on the certificates, I expect that should be done sometime today.14:39.36 
rayjj (what chrisl said)14:39.36 
mvrhel_laptop henrys: ok great14:39.53 
fredross-perry thank you14:39.57 
rayjj henrys: according to the logs, Robin got in (I thought)14:40.03 
Robin_Watts rayjj: I could log in, but there are issues with requesting the certs from different machines/web browsers.14:40.40 
  so I reset the password for joann and left it to her.14:40.49 
rayjj Robin_Watts: not very user friendly -- what if the requester's machine dies14:41.56 
henrys tor8: you were going to do an epub write up?14:41.58 
  Robin_Watts: oh a security thing about logging in from the same maching?14:42.21 
rayjj the problem is that Joann only works afternoons (and not every one of those)14:42.38 
Robin_Watts rayjj: Once we have the cert it's fine. It's the initial request that matters, AIUI.14:42.46 
  And they need to do security checks which will involve calling the office etc.14:43.07 
tor8 henrys: 
henrys rayjj: this is a subscription and a I didn't want to expense it otherwise I would have done it.14:43.20 
  tor8: thanks and Robin_Watts will stuff it in the twiki or I can do it.14:44.01 
Robin_Watts Will do that now.14:44.20 
henrys I wanted to ask chrisl about his recent fix and why it isn't a freetype problem, but I that's not really a meeting topic. Was there already discussion about it in the logs, if so I'll just do my reading.14:46.04 
kens No discussin no14:46.23 
  But its the font that's buggy I think,no FT14:46.39 
chrisl henrys: it's a font problem, not a freetype problem14:46.45 
henrys and it works in the other systems because they are not applying the hint?14:47.30 
kens No, because they aren't doing non-square scaling14:47.52 
chrisl Either they are not applying the hints, or they do the same as Ghostscript does now14:47.55 
henrys chrisl: okay14:48.23 
chrisl henrys: if it was just one font, I'd have punted it back as "not our problem", but it seems there are quite a few14:49.00 
kens Probably all using the same dumb tool....14:49.28 
chrisl Yep....14:49.40 
henrys Quality Logic is trying to sell us the ATS-IF thing for Office 2016, I don't think we need that. There is some argument for using it for SO thought.14:51.57 
jogux henrys: dunno what that is. :) got a link or details you can forward?14:53.09 
henrys jogux: let me forward on a data sheet to sos14:54.00 
  jogux: everyone else knows what ATS-IF is and is suddenly very quiet.14:54.34 
jogux :-)14:54.43 
kens has no useful opinions regarding anything to do with SOT14:55.02 
Robin_Watts thinks we have quite enough problems without making ourselves a checklist of every little thing that's different.14:55.40 
henrys kens: the tests are really for our printer languages. I thought they could find another use testing SO14:55.49 
chrisl henrys: Only the printer language tests are really useful for us14:56.49 
henrys chrisl: I think they are moving away from that. Having customers generate their own.14:57.30 
chrisl Well, that's completely useless, as an industry test suite14:57.48 
henrys probably the right thing to do with the ubiquity of drivers14:57.54 
kens The drivers on (for example) WIndows XP are not the same as on WIndows 814:58.17 
chrisl No, too much variation when you generate the actual test jobs14:58.25 
  To be sure of useful comparisons, you need to be sure you use the same Windows version (down to patch level), the application (down to patch level) and the same driver (down to patch level) with the same configuration......15:00.17 
Robin_Watts I'm with chrisl on this one.15:00.58 
chrisl Considering we often struggle to get customers to give us a Ghostscript command line......15:01.00 
Robin_Watts They should supply the original docs, plus an official set of generated ones (or maybe multiple sets)15:01.32 
henrys Robin_Watts: they do give you reference output.15:02.23 
marcosw Robin_Watts: afaik they used to do that, supply the source doc/ai/... files and generated PS/PDF files.15:02.32 
henrys I can't think of how many times I wish I had the original app file.15:02.43 
kens henrys If I generate the file on an official printer driver here in the UK I will get different otput from you in the US. Because the default media size will be different15:03.02 
rayjj chrisl: since most printer companies also provide a driver, they _do_ want to generate the output themselves15:03.37 
henrys that's a pretty easy fix. I wouldn't worry about that one.15:03.45 
rayjj using their driver.15:03.47 
chrisl rayjj: so when they ask "how do you perform with the QL test suite?" how do we answer that?15:04.13 
rayjj that's the problem I've had with cust 532 -- when they have an ATS file, I have to get their version that came from their driver15:04.21 
kens Right, whch makes it useless as an industry-wide test as Chris said15:04.48 
rayjj chrisl: we generally pick a commonly available "reference" printer and use that driver15:04.54 
henrys I need to head on over to the other meeting...15:05.01 
chrisl rayjj: it doesn't make for a good, consistent comparison15:05.39 
rayjj chrisl: for the PDF ATS, QL provides PDF's generated several ways and one can specify which of the provided ones we use15:06.17 
chrisl rayjj: yes, that's really how it should be15:06.40 
  for all the test suites15:07.00 
kens I don't have any problem with them supplying the source files as well, so that you can test your own driver, but there ought to be a 'correct' version in the output language15:07.50 
chrisl That would also let you benchmark your driver, as well as your printer/rip15:08.20 
kens And even more scope for 'optimising' the tests :-)15:08.48 
rayjj kens: for a real printer company (e.g. cust 532) they only care about what happens with their driver.15:08.51 
kens rayjj I have no problem with that, but the point is that for comparison purposes you need something concrete15:09.14 
  And a standardised peice of 'printer language' is what should be provided15:09.36 
rayjj but generally, since we only handle the PDF, their driver doesn't touch it, so we get files "from the wild"15:09.48 
  basically, whatever the customer wants is what is right (as long as they're big enough) :-)15:11.02 
kens I htnk its just cost-cutting on QL's part. THey're getting people to do the work of producing the printer output.15:11.33 
  I suppose there aren't enough of us left in the RIP business ot be worth their while.15:11.55 
chrisl rayjj: the problem is if it's dependent on their driver, then it doesn't necessarily represent a convincing validation of the implementation of the PDL - just whatever subset their driver happens to use. And that's not really the point of an "industry test suite"15:12.32 
kens Oh Good Giref. 'SHould we revert to using subsampling?' Well that depends on whether you want your images subsampled or bicubic subsampled.15:13.51 
rayjj chrisl: the ATS files are generally more of a Quality Assurance test suite to let printer makers validate the app->driver->pdl->paper process15:14.06 
  chrisl: and sometimes, they will do performance testing vs. the competing printers using that company's driver. Then if they come up short they dig deeper to see if the problem is due to the driver, the PDL or the engine15:17.15 
chrisl rayjj: which is fine, but leads to a problem when potential customers ask us, Global and Adobe for speed/memory comparisons when running that test suite....15:17.53 
kens rayjj none of ths is a problem, and a good reason to supply the source files. Not a good reason *not* to supply a set of PDL files.15:17.56 
chrisl Worse still, because of the QL licensing, those potential customers aren't allow to give their candidate vendors sets of files to test15:19.15 
rayjj kens: they do provide a set of 'reference' files for the PDL15:20.37 
kens henrys said they are 'moving away form that, having customers generate hteir own'15:21.15 
rayjj chrisl: actually, once we bought the ATS, they did give cust 532 written permission to give us their version of the file15:21.32 
henrys well I don't know that for sure this last data sheet I got said the masters were "optional" that's a departure from what has been done in the past.15:24.09 
kens 'optional' probably means 'costs extra15:24.52 
rayjj have to move my car. bbiab15:31.51 
Robin_Watts henrys: Oh, we forgot to say. We're going to try to get 1.7b out by the end of the week.15:39.27 
henrys Robin_Watts: oh I thought it was out...15:40.41 
Robin_Watts We got a few epub bugs in that tor has fixed.15:41.12 
  and a memory fix.15:41.25 
henrys scott and miles are still fooling with the newsletter but it's release is certainly this week, good to have it up as soon as possible.15:42.04 
Robin_Watts tor8: What remains to be done before we can tag and release? Were there other bugs/tests you wanted to fix first?15:42.51 
tor8 Robin_Watts: the commits on tor/master then I think we're in good enough shape to tag15:45.23 
Robin_Watts tor8: The top one looks strangely simple, but otherwise all 3 look good.15:47.23 
tor8 Robin_Watts: there are a few more coming soon, sebras is here today :)15:59.46 
Robin_Watts ok :)15:59.53 
henrys hi fredross-perry are you still about?16:06.41 
fredross-perry yes, here.16:06.49 
henrys fredross-perry: I meant to get to it in the meeting but we were thinking of getting you started in SOT.16:07.53 
  smart office16:08.00 
  all that discussion happens on skype. though16:08.26 
fredross-perry got it. heading there now16:08.38 
Robin_Watts chrisl: ?17:42.50 
mvrhel_laptop I hate illustrator and in-design. Trying to use these tools to create some pdf examples for this blending color space stuff without icc profiles. In the save as options I specify no profiles. But they still embed the profile for the blending spaces17:43.25 
  maybe the newer versions are better17:45.05 
Robin_Watts mvrhel_laptop: I suspect if you want fine grained control you will need to hand edit the PDF file.17:45.32 
mvrhel_laptop yes. I had been doing that17:45.40 
  but I actually want to create a bit more complicated file with pattern and shading fills of glyphs laid over various objects17:46.22 
  I will likely use the tool to get close then go and and hack it17:46.33 
  So I had been doing a series of simple tests and building a set of notes as I blended various Source colors e.g. DeviceCMYK, DeviceRGB, ICC into various blend spaces e.g. DeviceCMYK, DeviceRGB, ICC . I thought I had a good handle on it, but this one damn altona test file with its non-isolated transparency group in a DeviceCMYK blending space with a glyph that is filled with an pattern that...17:56.36 
  ...has a shading defined by RGB ICC colors is not fitting in 17:56.37 
  I will get to the bottom of this but it is going to take a bit of time17:58.58 
  one more test to make. I am getting very suspicious about the text being filled with the shading pattern18:19.55 
  ok so having the pattern in shading makes for a completely different result18:50.23 
  need to check one more thing18:50.34 
  but basically if I fill the glyph with a shading, that is quite different than filling it with a pattern that has a shading.18:50.57 
  I do see this in the spec18:51.05 
  Page 560 of PDF spec18:51.27 
  In both cases, the pattern definition is treated as if it were implicitly enclosed in a non-isolated transparency group: a non-knockout group for tiling patterns, a knockout group for shading patterns.18:51.28 
  so, right now I have a glyph filled with a tilling pattern that is filled with a shading pattern vs. a glyph filled with a shading pattern. These end up quite different through the acrobat transparency process18:58.43 
  need to try one more case18:59.19 
  ok. I think I finally found the issue with this file21:01.25 
  much less odd than I thought21:01.48 
  somehow gs is screwing up a group composition 21:02.19 
  in the altona file, the D was wrapped up in two levels of transparency groups that are getting hit with an exclusion blend21:03.43 
fredross-perry anyone on Skype?23:25.19 
henrys fredross-perry: pretty late for the euros23:25.48 
fredross-perry ok tomorrow then23:25.58 
mvrhel_laptop ok. I have beat on this to the point where it almost appears to me that Adobe is doing special color conversions based upon the blend mode. need to set up a test file for that 23:54.40 
  done for now 23:54.51 
 Forward 1 day (to 2015/05/06)>>>