| | <<<Back 1 day (to 2011/11/15) | 2011/11/16 |
| Robin_Watts | henrys: You're welcome. Glad to hear you think it'll fix it. | 01:24.51 |
| arthurf | tor8: Hi. Tried to do some quick searching with the iOS app on a PDF file, but not much happened. I'm guessing there is more work to do? | 02:52.24 |
| tor8 | arthurf: yes. it still doesn't show any results :) | 02:52.50 |
| arthurf | tor8: Okay thanks. Just wanted to check that I wasn't completely missing something. Thanks. :) | 02:53.22 |
| robingower | I'm having a bit of nightmare compiling ghostscript on a mac | 02:54.44 |
| | I filed this ticket at macports months ago | 02:54.54 |
| | https://trac.macports.org/ticket/28825 | 02:54.55 |
| | I can't get it to build outside of macports either | 02:55.18 |
| | any ideas of what else I could do to diagnose my problem? | 02:55.49 |
| tor8 | robingower: that's odd, many of us (developers) use macs as development machines when working on ghostscript | 02:56.02 |
| robingower | indeed. I've tried the usual suspects like fiddling with arch settings and I've removed fink and macports etc | 02:57.41 |
| tor8 | my normal procedure is: git clone, ./autogen.sh, make | 02:59.30 |
| | I don't use macports or anything like that, just a plain mac install with Xcode | 03:00.10 |
| robingower | 'k thanks tor8 I'll give that a try | 03:08.37 |
| | I've cloned from git://git.ghostscript.com/ghostpdl.git and done autogen & make | 03:23.01 |
| | but I still get | 03:23.04 |
| | ../gs/base/gstype42.c:68: error: conflicting types for âgs_type42_read_dataâ | 03:23.09 |
| | ../gs/base/gxfont42.h:141: error: previous declaration of âgs_type42_read_dataâ was here | 03:23.14 |
| | I've checked the offending lines (dozen's of times!) and they look the same to me | 03:23.36 |
| | i.e. same parameter types etc | 03:23.49 |
| AlecTaylor | hi | 04:53.39 |
| ghostbot | hola | 04:53.39 |
| AlecTaylor | Any MuPDF developers around? | 04:53.48 |
| | How would I go about reverse-engineering an XML file back into the PDF? | 05:06.41 |
| alexcher | AlecTaylor: XML describes a general structured document, e.g. an IP datagram. | 05:15.12 |
| AlecTaylor | just posted on the mailing-list | 05:15.30 |
| alexcher | AlecTaylor: PDF is a page description language. | 05:15.46 |
| | AlecTaylor: How one can be converted to another? | 05:16.48 |
| AlecTaylor | alexcher: I was able to output an XML document using MuPDFs pdfextract tool. I have extended the output to detect header/footers and page numbers. Now I want to push the *proper* page numbers and additional per/page logical structure information back into the PDF | 05:17.18 |
| alexcher | AlecTaylor: MuPDF developers will be here in a few hours. | 05:20.18 |
| AlecTaylor | kk | 05:20.22 |
| | BTW: Did I post to the right mailing-list? | 05:20.32 |
| | http://ghostscript.com/pipermail/gs-devel/2011-November/009091.html | 05:23.14 |
| alexcher | AlecTaylor: yes | 05:23.56 |
| AlecTaylor | Great :) | 05:24.07 |
| AlecTaylor | has been writing these extensions for the poppler libraries, but they don't seem to have anything for moving from XML->PDF | 05:24.29 |
| kens | Morning marcosw_ | 08:46.36 |
| AlecTaylor | How can I reverse-engineer an XML file generated with MuPDFs pdfextract tool back into the PDF? | 09:01.06 |
| kens | Write a tool to do it ? | 09:01.37 |
| | Open it with a XML reader and print to PDF ? | 09:01.53 |
| AlecTaylor | kens: Back into the PDF it was created with | 09:04.01 |
| kens | You can not recover teh *same* PDF (as requested somwhere else), its been interpreted and information has been lost. | 09:04.44 |
| | You can create a simliar, visually more or less identical PDF, but its not the same PDF. | 09:05.07 |
| | And we don't (as far as I know) provide a means to put it back. Why would you boother, you have the original. | 09:05.37 |
| | Actually, I guess you could do so, but I still don;t think there's a tool do it. I still don't see why you would want to eitehr, you have the original file | 09:07.22 |
| AlecTaylor | kens: I am adding logical structure information and *proper* page numbers into the PDF. No actual text will be changed, just some internal information | 09:35.45 |
| AlecTaylor | has developed a solution using just the stdlib and the header-only RapidXML library | 09:36.15 |
| kens | Ah, editing a non-editable format :-) | 09:40.05 |
| AlecTaylor | kens: eh? | 09:40.48 |
| kens | Well PDF files aren't really meant to be edited, even for metadata. | 09:43.07 |
| AlecTaylor | kens: True, but there are billions of PDFs without proper metadata. I would like to fix them all up! | 09:50.41 |
| AlecTaylor | is sure there's a way | 09:50.55 |
| kens | Creating a PDF file from the XML output of pdfextract is possilbe, but you need to regenerate the xref. SO you need an application to do it, and we haven't written one. | 09:52.50 |
| | You could use such a thing to create a Linearized PDF as well mind you | 09:53.07 |
| | But that might require more analysis | 09:53.25 |
| AlecTaylor | Hmm | 09:56.49 |
| | kens: Would there be another open-source project which provides the XML->PDF feature? | 09:57.48 |
| kens | I doubt it | 10:09.54 |
| AlecTaylor | :( | 10:10.28 |
| kens | But you know, you can be the first to write one :-) | 10:10.43 |
| AlecTaylor | yeah, but been working the last few months on this project, want it completed already! | 10:11.02 |
| kens | will be back later | 10:25.27 |
| Robin_Watts | marcosw_: ping? | 12:58.50 |
| | Hmm. The cluster code is going to need some changes to cope with building on a windows system. | 13:49.03 |
| kens | I would expect so | 13:49.12 |
| Robin_Watts | Presumably we want to call the projects rather than the makefiles. | 13:49.15 |
| | (otherwise we'd be testing a cygwin build rather than an MSVC one, and defeating the point) | 13:49.37 |
| kens | Yes, not good | 13:49.51 |
| | Unless you explicitly call nmake and force use of VC instead of gcc ? | 13:50.41 |
| Robin_Watts | I think an explicit call to nmake is the way to go. | 13:50.59 |
| | Or, probably better, a command line call to msdev, because that checks the projects. | 13:51.28 |
| kens | Will it use the right comiler though ? | 13:51.31 |
| | OK sounds good | 13:51.40 |
| Robin_Watts | Normally, you have to make something run once a minute on a crontab... that's going to need tweaking too. | 13:52.06 |
| | I am NOT having 2 copies of tests_private checked out on my machine. Time for some symlinks :) | 13:53.02 |
| kens | can't use task scheduling ? | 13:53.11 |
| Robin_Watts | I'd rather run something to start the cluster, and kill it to stop it. | 13:54.20 |
| | (I've just spent 250 quid on a graphics card - no point in that if I'm going to have the machine drop to a crawl when someone commits :) ) | 13:55.09 |
| | (this is not an Artifex owned machine :) ) | 13:57.00 |
| kens | So buy another one for Artifex. | 13:57.13 |
| Robin_Watts | If we're going to invest in a dedicated windows node, then putting somewhere with faster network access might make sense. | 13:58.09 |
| kens | Chris's office ;-) | 13:58.26 |
| chrisl | as long as it's *very* quiet! | 13:58.55 |
| kens | Or perhaps Alex's basement | 13:58.56 |
| | being slightly more serious for a minute | 13:59.15 |
| Robin_Watts | My new graphics card is slightly more noisy than my last one :( | 13:59.31 |
| | Helens new passport has just been delivered. I guess she'll be coming to Miami after all... | 14:12.33 |
| kens | Glad to hear it :-) | 14:12.55 |
| tkamppeter_ | kens, can you have a look at bug 692687? If there is a simple fix for it I would like to issue this fix as a patch release for Oneiric, too. | 14:20.43 |
| kens | OK | 14:21.56 |
| | I owuld always advise against 'round tripping' files though | 14:23.20 |
| tkamppeter | kens, the problem here is that we are in a transition between PostScript-centric and PDF-centric CUPS filtering. In Debian and Ubuntu we have switched to PDF-centric filtering, but there are still some few applications which send print jobs in PS and some few printer drivers which rerquire PS input. | 14:27.04 |
| kens | Yeah, just saying | 14:27.37 |
| | Each conversion potentially throws stuff away | 14:28.39 |
| AlecTaylor | bn | 14:32.29 |
| tor8 | Robin_Watts: looking through your patches now. they look good to me, but I must have something to complain about! I really do prefer: while (span) { ... span = next; } to the do { ... span = next; } while (span != NULL) form for loops like in the "stack overflow in text handling" commit | 14:36.43 |
| kens | tkamppeter it seems to be somthing to do with multiple subset TrueType fonts. It looks like a lot of glyphs are being replaced with .notdef | 14:40.48 |
| | Its going to take a while just to reduce the original PostScript file to somethign simple enough to investigate. I'm rather busy with a different problem right now, how urgent is this ? | 14:41.40 |
| daviddoria | In the help of a latex package it says I can run a line like this to convert all of the pages to separate png files: gs -sDEVICE=png16m -dTextAlphaBits=4 -r300 -dGraphicsAlphaBits=4 -dSAFER -q -dNOPAUSE -sOutputFile=equation%d.png Equations.pdf - however, when I do that, i just get GS> prompt? | 14:43.23 |
| kens | Any error message ? | 14:46.00 |
| | Did you properly specify the input file ? | 14:46.07 |
| | Oh, and you need to add -dBATCH to exit afterwards or type 'quit' | 14:46.32 |
| daviddoria | kens, oh haha it actually worked... I am just suprised it didn't exit automatically but rather took me to a GS> prompt? | 14:46.39 |
| kens | See comment above | 14:46.50 |
| daviddoria | yep, that did the trick, thanks | 14:47.02 |
| kens | no problem | 14:47.10 |
| | tkamppeter did you see my earlier question ? | 14:54.35 |
| tkamppeter | kens, it is not overly urgent, but if it is done in a week it would be great. | 15:09.36 |
| Robin_Watts | tor8: I did it that way, because it exactly matched what was there before. | 15:12.12 |
| kens | tkamppeter, well maybe, I can't be certain | 15:12.29 |
| Robin_Watts | But, yes, it would be nicer to have used while () { ... } as that would have protected against span being NULL on entry. | 15:12.39 |
| kens | It'll tkae me several hours just to reduce the problem until I can work on it | 15:12.42 |
| Robin_Watts | tor8: I'll fix that if you want. | 15:13.06 |
| kens | The file contains 4 fonts, 3 of tehm are compostie nad contin up to 15 sub-fonts | 15:13.09 |
| tor8 | Robin_Watts: right. in other places where I've used an iteration to free a linked list instead of the lazy recursion I use while (foo) {} | 15:13.09 |
| kens | tkamppeter, also this file started life as a PDF file. | 15:13.34 |
| | So ita PDF->PS->PDF->PS | 15:13.44 |
| tor8 | Robin_Watts: also get rid of the != NULL explicitness, we don't do that in mupdf ;) | 15:14.04 |
| Robin_Watts | but we should :) | 15:14.24 |
| tor8 | then we may as well rewrite it all in pascal... | 15:14.41 |
| | do you want to amend your commit or make another one? | 15:15.07 |
| | (ie, should I push to master or wait) | 15:15.16 |
| Robin_Watts | IF you haven't merged yet, I'll fix mine. | 15:15.28 |
| | but if you have, push what you've got and I'll do new ones. | 15:15.43 |
| | I can't do this immediately - in the middle of fighting the cluster. | 15:15.52 |
| tor8 | I haven't merged anything yet | 15:15.53 |
| Robin_Watts | by "merged" I mean "pulled my changes in", sorry. | 15:16.12 |
| tor8 | I pulled them in but I haven't done anything other than looked at them | 15:16.29 |
| Robin_Watts | ok, I'll redo them. Will be a nicer history that way. | 15:16.40 |
| tor8 | so if you want to amend, that's no trouble for me | 15:16.43 |
| tkamppeter | kens, the problem here is also that okular sends print job in PS format and it should send PDF, see https://bugs.launchpad.net/ubuntu/+source/okular/+bug/891199. | 15:44.43 |
| AlecTaylor | Reverse XML containing amended page number and new logical structure information (partition header/footer from other page content) using MuPDF? - If so, how? | 15:45.03 |
| Robin_Watts | AlecTaylor: You're starting from the wrong place. | 15:46.26 |
| | When you convert from pdf to xml you are getting out a certain subset of the information from within the PDF. | 15:46.58 |
| kens | tkamppeter it just increases the complexity, and the number of conversions. Makes it harder to sort out the real problem. Its going to take me quite a while just to simplify the problem. A file with 1 font instead of 4 and no grpahics, would help. | 15:47.03 |
| Robin_Watts | If you discard the PDF you are then losing all the data. | 15:47.13 |
| AlecTaylor | Robin_Watts: I needed XML so that I could have per page in "normal" text | 15:47.21 |
| Robin_Watts | Far smarter to get the xml out, process it to get what you want to add, then do a process to 'add' your new information back into the original pdf. | 15:47.43 |
| AlecTaylor | Robin_Watts: Also, can't I just modify the PDF that the XML was generated from | 15:47.49 |
| | yeah, that | 15:47.59 |
| | How can I do that? | 15:48.09 |
| Robin_Watts | AlecTaylor: Someone suggested on the mailing list that you add it into the end of the contents streams. | 15:48.30 |
| | That's possible, but a bad idea, IMHO. | 15:48.44 |
| kens | Which will break teh stream length (need to recalculate it) and the xref. | 15:48.50 |
| Robin_Watts | A better idea would be to add your new content as annotations. | 15:48.56 |
| AlecTaylor | So what should I do? | 15:48.57 |
| Robin_Watts | kens: Yeah, but easy enough to do with a modified pdfclean. | 15:49.08 |
| AlecTaylor | Robin_Watts: As annotations? | 15:49.12 |
| Robin_Watts | Echo... | 15:49.26 |
| | PDF provides the facility to add annotations to a document. | 15:49.54 |
| | Read the PDF spec for more details. | 15:50.19 |
| | I suspect you'd end up taking pdfclean (which reads a pdf in, 'does stuff' to it, and then writes it out) and tweaking it. | 15:51.01 |
| | You want to read a pdf in, add annotations to it, and then write it out. | 15:51.13 |
| kens | Could use pdfwrite and pdfmarks | 15:51.43 |
| Robin_Watts | A 'free text' annotation seems like the one you want. | 15:52.51 |
| AlecTaylor | Hmm, I don't think annotation is what I am after. The displayed page number I'd like changed to whatever I have heuristically determined. | 15:54.20 |
| Robin_Watts | You use an /AP entry to point at a new xobject that you define, and that xobject can have a stream of pdf operators in it. | 15:54.21 |
| kens | What do you mean by 'displayed page number' ? | 15:54.43 |
| | The rendered value, or the ordinal displayed b Acrobat ? | 15:54.57 |
| Robin_Watts | AlecTaylor: AIUI, you want to add specified content to every page (different on each page), right? | 15:55.33 |
| kens | thinks not | 15:55.41 |
| AlecTaylor | kens: The number displayed on each page in the PDF viewer | 15:56.10 |
| kens | So like I thought, not the page content stream | 15:56.42 |
| AlecTaylor | The other feature I want to add is a per page logical structure, which specifies what's page content, and what header/footer | 15:56.49 |
| kens | I think Acrobat get s that from teh Outlines tree if present | 15:56.53 |
| | So you need to add an Outlines | 15:57.08 |
| Robin_Watts | (or amend an existing one) | 15:57.19 |
| AlecTaylor | has written and prototyped algorithms that do this, and is outputting to an XML file. Now wants to push that to a PDF | 15:57.19 |
| | yeah | 15:57.32 |
| kens | As for your other info, what PDF structure do you propose to use to hold that inrformation, marked content ? | 15:57.44 |
| Robin_Watts | Right, so you want some code that takes the pdf and the xml and outputs a new pdf. | 15:58.05 |
| AlecTaylor | Yeah | 15:58.14 |
| Robin_Watts | AIUI, the complete scheme for what you have does: | 15:58.18 |
| | original.pdf -> temp.xml (using mupdf) | 15:58.33 |
| | temp.xml -> temp2.xml (using your processing code) | 15:58.51 |
| AlecTaylor | pdfextract original.pdf; someothertool original.xml original.pdf | 15:59.08 |
| Robin_Watts | Then you want to go: temp2.xml + original.pdf -> final.pdf (using the bit of code we are talking about now) | 15:59.18 |
| AlecTaylor | Now original.pdf has new information from original.xml | 15:59.23 |
| | yeah | 15:59.33 |
| Robin_Watts | OK. So you want something like a modified pdfclean. You can strip that down so it just opens and reoutputs a pdf (and lose all the garbage collection, page extraction code etc) | 16:00.45 |
| | Then you can start to build it up so that it reads your xml and adds extra objects to the PDF before it reoutputs it. | 16:01.14 |
| | But you have to realise that pdfextract goes to some lengths to present you with a 'coherent' version of content. | 16:02.09 |
| | Just because you see 2 columns of content in the xml output, doesn't mean it appears like that within the PDF source. | 16:02.57 |
| | The pdf source could have bits of column 0 and bits of column 1 interleaved with one another. | 16:03.14 |
| | hence 'tagging content' at that level is likely to be hard. | 16:03.45 |
| | I still don't entirely understand what your aim is here. | 16:04.00 |
| | I thought from what you had said before that it was to add headers/footers to pages. | 16:04.44 |
| | If you're now saying you want to 'reflow' page content that's a hugely different thing. | 16:05.01 |
| kens | AIUI the aim is to add metadata to PDF files with things like the ''tagged' (marked data) concept to delineate header/footer/ page number etc. And also to add Outlines with teh 'correct' page info (content, index etc) | 16:05.03 |
| Robin_Watts | Using marked data will require (I reckon) a modified pdf interpreter; it'd need to run the PDF, and catch the marking of the page; as it marks into the appropriate regions (identified by the previous xml processing), then it would need to rewrite the streams to include marked content markers. | 16:08.12 |
| | That's a huge job. | 16:08.18 |
| | But I'd like to hear a clear description of what's required from AlecTaylor before we go any further, cos I could be barking up the wrong tree. | 16:09.01 |
| | Is the idea to 'mark' the headers/footers/stories (say by adding visual boxes to the page)? | 16:09.43 |
| AlecTaylor | Hmm | 16:10.35 |
| | "<kens> AIUI the aim is to add metadata to PDF files with things like the ''tagged' (marked data) concept to delineate header/footer/ page number etc. And also to add Outlines with teh 'correct' page info (content, index etc)" | 16:10.46 |
| | That's what I am trying to do | 16:10.53 |
| | as well as "repair" the page numbers displayed by the PDF viewer to be the same as the page number "printed" on the page | 16:11.25 |
| Robin_Watts | So what's the purpose of adding metadata? How will my viewing experience differ by me looking at a processed file rather than the original one? | 16:12.03 |
| kens | That depends (I think) on the Outlines. If there are no Outliens Acrobat displays the ordinal page number (1/50 and so on) | 16:12.11 |
| | Robin_Watts : If you use (eg) a PDF to speech package, it won't read page numbers if they are idetified as such, for example | 16:12.59 |
| | tagged PDF is mandated by a number of governement agencies for this reason | 16:13.14 |
| | (accessibility) | 16:13.20 |
| Robin_Watts | OK. That's what I was after. Is that the reason, AlecTaylor ? | 16:13.35 |
| kens | Because they don't really understand what the creation implies | 16:13.48 |
| | Also header/footer may be omitted etc. | 16:14.16 |
| AlecTaylor | Yeah, for accessbility and searching is my purpose for specifying what on the page is the header/footer and what is the other stuff. | 16:14.45 |
| kens | Searching already works, or if it doesn't you won't be able to fix it | 16:15.05 |
| Robin_Watts | AlecTaylor: Right. So annotations are not the way to go. | 16:15.12 |
| | You do need to rewrite the content streams. | 16:15.28 |
| kens | I see it as /Outlines plus marked content (tagged PDF) | 16:15.39 |
| | I *think* you can insert marked content inside text blocks, but I'm notr certain | 16:16.03 |
| | If you cant' then that's going to be really hairy | 16:16.12 |
| Robin_Watts | And pdfclean is a reasonable starting point (it reads in/writes out, and will enable you to add Outlines). | 16:16.18 |
| | But it's far from easy. | 16:16.39 |
| AlecTaylor | Hmm | 16:16.56 |
| Robin_Watts | Because you're going to have to interpret the PDF in order to know where each text rendering operation puts its text (and hence in what section it goes). | 16:17.24 |
| AlecTaylor | Well on the bright side, at least IT'S POSSIBLE in this library (rather than poppler, which I've been working off) | 16:17.29 |
| Robin_Watts | You could easily have a text object that wrote to both header, footer, and multiple stories within in. | 16:18.02 |
| | AlecTaylor: Anything is possible with any library depending on how much you're prepared to rewrite :/ | 16:18.47 |
| AlecTaylor | Robin_Watts: I envision a find/replace function, search for this text "aaa 44" (whatever's in the XML tag), add information around that text tagging it up | 16:18.47 |
| Robin_Watts | AlecTaylor: Sadly, no. | 16:18.58 |
| AlecTaylor | :\ | 16:19.07 |
| kens | That text may not appear in the PDF | 16:19.18 |
| Robin_Watts | You might have got information in your extracted XML telling you that the word "hello" appeared on the page. | 16:19.45 |
| | and you might want to tag that as being part of the header. | 16:19.52 |
| | Within the pdf, you might have "hello", or you might have had "h", "e", "l", "l", "o" | 16:20.24 |
| | or "he" "ll" "o". | 16:20.30 |
| AlecTaylor | Ahh, because it isn't shown as a big line of text | 16:20.34 |
| Robin_Watts | or "o" "l" "e" "h" "l" | 16:20.45 |
| kens | Or even random text between, but positioned separately | 16:20.47 |
| AlecTaylor | wait | 16:20.50 |
| | what | 16:20.52 |
| | That last one, I don't get it | 16:20.57 |
| kens | Font encoding | 16:21.03 |
| | What you see in the XML is eht Unicode from teh ToUnicode cMap | 16:21.16 |
| Robin_Watts | kens: I wasn't even talking about Font encoding yet :) | 16:21.18 |
| kens | Or a guess based on other heuristics | 16:21.25 |
| Robin_Watts | AlecTaylor: Letters can be sent to the page in any order. | 16:21.37 |
| kens | Yep. | 16:21.46 |
| Robin_Watts | PDF is basically a stream of 'page marking' operations. | 16:21.51 |
| sebras | Robin_Watts: mmm, or what characters are actually part of the font embedded in the pdf... | 16:21.53 |
| kens | And often are in CJKV or right to left languages | 16:21.58 |
| AlecTaylor | :S | 16:22.30 |
| Robin_Watts | As such any marks that add up to give the correct final result are fine; letters can be send in any orders. | 16:22.39 |
| | You could send all the 'a's then all the 'b's etc. | 16:22.51 |
| | You need to redo the page interpretation, and watch where each char is placed; if it happens to be outputting a char to a place you know is in a header, then you can rewrite the stream to mark it as being in the header. | 16:24.46 |
| | It's not a trivial job. | 16:24.57 |
| AlecTaylor | How about extending pdfclean to fix up the streams for this operation? | 16:28.12 |
| Robin_Watts | AlecTaylor: pdfclean will get the streams into memory in a raw, uncompressed form. | 16:28.48 |
| | You can rewrite them there. | 16:29.02 |
| AlecTaylor | Hmm | 16:29.08 |
| kens | Its not as simple as 'fixing' them, you haev to identify where teh stream writes the glyphs you are interested in and modify it | 16:29.13 |
| AlecTaylor | Seems quite extensive and complicated | 16:29.19 |
| Robin_Watts | BUT... knowing how to rewrite them is the hard part. | 16:29.21 |
| kens | AlecTaylor : Yes, that's what we've been saying :-) | 16:29.35 |
| AlecTaylor | :( | 16:29.43 |
| Robin_Watts | AlecTaylor: You didn't want to pick a simple project for your thesis, right? | 16:29.44 |
| AlecTaylor | My thesis uses something else, haven't started it yet (hasn't been approved yet). | 16:30.07 |
| | This was a one-semester undergrad research project | 16:30.19 |
| | Semester ends in 7 days | 16:30.58 |
| | Also have 2 exams for other subjects within these 7 days | 16:31.22 |
| kens | Well you can weite up the research | 16:32.07 |
| | write | 16:32.12 |
| AlecTaylor | Yeah, that's what I'll do instead. | 16:32.40 |
| | This seems wayyyyy too extensive for something I can do in 7 days, even if I wasn't working on anything else | 16:32.58 |
| kens | Yeah, not a hope of doing it in 7 days | 16:33.12 |
| Robin_Watts | indeed not. | 16:33.18 |
| kens | You could do an 80/20 probably | 16:33.21 |
| AlecTaylor | wah | 16:33.35 |
| Robin_Watts | AlecTaylor: If you have reliably detected headers/footers/page numbers, then that's a big job done. | 16:33.58 |
| kens | I think that's still information Tor and I would be intrested in getting | 16:34.31 |
| Robin_Watts | You could easily write a 'further work' section describing how you'd like to have a tool to 'reinsert' that information back into the original pdf. | 16:34.49 |
| AlecTaylor | All well, you guys want my research prototype which creates an XML from the XML generated by pdfextract, with header/footer tags with proper page numbers, implemented in only stdlib (with the header-only RapidXML library for reading in the XML)? | 16:34.55 |
| AlecTaylor | is giving it the poppler project, but no reason can't give it to you guys too | 16:35.13 |
| Robin_Watts | AlecTaylor: I think we'd all be interested in seeing it, yes, thanks. | 16:35.23 |
| kens | I'd certainly be interested in it | 16:35.24 |
| AlecTaylor | Sure, I'll send over a patch soon | 16:35.35 |
| Robin_Watts | In fact, when you write up your paper, give us a link :) | 16:35.39 |
| kens | At the least we can add the information to the XML outptu from MuPDF and Ghostscript | 16:35.41 |
| AlecTaylor | Will do :) | 16:35.46 |
| | I might add in some fuzzy stuff, are OCR errors prominent enough to warrant it? | 16:36.53 |
| kens | My experience is OCR is too good to need it today | 16:37.19 |
| Robin_Watts | 'fuzzy' ? | 16:37.51 |
| | OCR is at least as good at spelling as your average youtube commentard :) | 16:38.09 |
| AlecTaylor | Robin_Watts: So I could employ Levenstein Distance with a barrier of 1 or so | 16:38.13 |
| | On spell-checkers, Google has gotten terrible. It overly uses its result statistics to spell-check | 16:38.46 |
| | But sure, if OCR is fine at the moment, I'll leave it be | 16:39.44 |
| | Is there an article I can reference so that I can legitamately skip it? | 16:39.58 |
| kens | None that I know of. My only real experience of OCR is 30 years out of date | 16:40.33 |
| | Back tehn it was an intersting problem | 16:40.51 |
| Robin_Watts | AlecTaylor: I don't see that you need to cite a reference; stating that it's a problem that you are ignoring should be enough. | 16:41.15 |
| | The original document scanner is far better placed to do such fixes. | 16:41.37 |
| AlecTaylor | So can I safely say "levenstein distance is utilsed at scan stage so I won't do it here" | 16:42.02 |
| | or somethign along those lines? | 16:42.07 |
| kens | I don't know what techniques OCR people use today. | 16:42.37 |
| | I'd be inclined to say that it 'could be used as a further refinement, if experience demomstrates a problem' | 16:42.57 |
| Robin_Watts | AlecTaylor: Yeah, don't tie yourself to a single possible algorithm. | 16:43.41 |
| AlecTaylor | k | 16:44.06 |
| | All well, gotta head off to grab a few hours sleep soon | 16:44.22 |
| | (3:44AM here) | 16:44.27 |
| | Just gotta figure out this file upload problem for a conference I am organising (which is in 7 days) | 16:44.43 |
| | Robin_Watts, kens: Porting MuPDF to javascript: possible/impossible? | 16:47.06 |
| | (just for a viewer) | 16:47.13 |
| kens | Re-writing in JavaScript ? Impossible (or rather takes too long | 16:47.41 |
| Robin_Watts | AlecTaylor: You wouldn't rewrite, you'd write from scratch. | 16:48.42 |
| AlecTaylor | Would it be a waste of time? - Is it what Google did? | 16:49.13 |
| Robin_Watts | Someone else has done it. | 16:49.19 |
| | It's not what google did. | 16:49.26 |
| | Yes, it would be a HUGE waste of time (unless you're interested in doing it to drive javascript development) | 16:49.47 |
| AlecTaylor | Nope, just interested | 16:50.43 |
| AlecTaylor | was thinking to port the Qt Poppler viewer stuff to Wt | 16:51.01 |
| mvrhel_laptop | good morning | 16:52.57 |
| Robin_Watts | Morning | 16:53.05 |
| kens | Hi mvrhel_laptop | 16:53.49 |
| | Would you mind looking at bug #692680 ? | 16:54.14 |
| | Just to read it! | 16:54.24 |
| mvrhel_laptop | sure | 16:56.07 |
| kens | Thanks, if I'm gone when Ray arrives, could you ask him to read it too please ? | 16:56.35 |
| mvrhel_laptop | yes | 16:56.50 |
| kens | At the moment it looks like Alex's suggestion is the only concrete possibility for a solution, but you and Ray were talking about high level colour so I'd like you to bear this problem in mind ;-) | 16:57.17 |
| mvrhel_laptop | ok. making my way through the comments | 16:58.08 |
| kens | Yeah, its a bit complicated. | 16:58.23 |
| | Feel free to ask if its not clear | 16:58.44 |
| | Cool technology: | 16:59.18 |
| | http://www.reghardware.com/2011/11/16/3d_holographic_displays_become_reality/ | 16:59.18 |
| mvrhel_laptop | wow this eps file sounds terrible | 16:59.49 |
| kens | I understand why its doign it, but I don't like it | 17:00.02 |
| AlecTaylor | I like that green flashing lazer | 17:04.30 |
| | makes me hungry for fake 3d fish | 17:04.37 |
| kens | I want my real 3D telly now please. | 17:05.00 |
| AlecTaylor | You already have one | 17:06.00 |
| kens | proper 3D display | 17:06.09 |
| | fake ones don't work for me | 17:06.16 |
| AlecTaylor | You have a 2D TV? - Amazing! | 17:09.03 |
| | You're such a square :P | 17:09.11 |
| kens | display.... | 17:09.22 |
| | ray_laptop : I'm about to go off, but can you review and think about bug #692680 please. | 17:12.47 |
| | Not for a solution, just to consider in terms of future work relating to colour | 17:13.02 |
| ray_laptop | kens I saw your comment and the gs-bugs email with | 17:13.11 |
| kens | Thanks Ray | 17:13.21 |
| ray_laptop | kens: I agree that it looks AWFUL | 17:13.31 |
| kens | ray_laptop : well on reflection I understand what's done and why, but its nasty. | 17:13.50 |
| ray_laptop | kens: yes, and they didn't even 'bind' the procset :-( | 17:14.11 |
| kens | Actually, even if they did it wouldn't help | 17:14.25 |
| | Because we actually need to do what Acrobat does. Send the Indexed sample through both colour spaces, to get 6 colour values for each image sample | 17:14.48 |
| | Then set up /DeviceN [/Black /Spot1 /Spot2 /None /None /None] /DeviceRGB | 17:15.21 |
| ray_laptop | oh, yuck! | 17:15.45 |
| mvrhel_laptop | wow | 17:15.50 |
| kens | So if the inks are rpesent we ignore teh last three values, if the inks aren't present then the tint transform to RGB does {pop pop pop} to remove the spot inks and elave the RGB | 17:15.54 |
| ray_laptop | kens: I understand now. | 17:16.37 |
| henrys | ray_laptop:it looks like marcos dumped 692674 on you without history investigation - that customer did get poor service with their problem, will you have a chance to look at it soon? Or should we send it back to marcos for history? | 17:16.51 |
| ray_laptop | henrys: I've started looking at it (on peeves) | 17:17.12 |
| | henrys: I'll post something today. I think Marcos just gave it to me today | 17:17.32 |
| | henrys: and I agree that they sort of got ignored, but then, they aren't my favorite customer either | 17:18.04 |
| AlecTaylor | Damn .htaccess file, it's working now | 17:18.24 |
| | WOOT | 17:18.25 |
| kens | Did they get ignored, they opened it one day and started bleating 24 hours later | 17:18.28 |
| henrys | ray_laptop:I didn't follow it carefully they whined in the bug report. | 17:19.12 |
| ray_laptop | henrys: these were the guys that had me work on trimming down the initialization time of gs because they were timing 5,000 simple jobs and starting gs each time | 17:19.16 |
| henrys | oh same folks, sigh | 17:19.38 |
| mvrhel_laptop | I wonder what the *huge* difference is | 17:19.44 |
| ray_laptop | henrys: Miles had them pay extra for that. | 17:19.47 |
| | I'll post my comments to the bug once I do the timings on peeves. | 17:20.06 |
| | (I'll do profile as well) | 17:20.28 |
| kens | TBH I think the answer is 'there are bg fixes in newer releases, sometimes these cause performance penalties for correct behaviour. You cna have fast or good, your choice' ;-) | 17:20.33 |
| ray_laptop | kens: may be -- but first I want to get the facts | 17:20.52 |
| kens | Yeah, I was thinking of looking at it, but got diverted, and then realised it wasn't pdfwrite | 17:21.23 |
| Robin_Watts | kens: performance differences between 8.x and 9 is probably color management - but ray_laptop is right that we'd like to be sure of that. | 17:21.28 |
| mvrhel_laptop | sure. blame it on the color... :) | 17:21.56 |
| ray_laptop | mvrhel_laptop: if it is the color, I'll assign the bug to you :-) | 17:22.15 |
| kens | It could be almost anything, including the fact that it uses FreeType, which is what fixed hteir *original* problem. | 17:22.28 |
| | Bear in mind that Chris has done some perfomrmance fixes in that area since the release of 9.04 | 17:22.54 |
| ray_laptop | kens: iirc, FT is slightly faster than the AFS (Artifex Font Scaler) | 17:23.13 |
| kens | But that's what made me think of 'yes,. its right now, but slower' as an answer | 17:23.13 |
| | ray_laptop : There was something that made it slower, Chris fixed it. | 17:23.27 |
| ray_laptop | kens: don't worry, I will also time HEAD | 17:23.33 |
| kens | I think it was something about freeing memory | 17:23.39 |
| chrisl | Yeh, confusion about what Freetype does and when. | 17:24.30 |
| kens | Anyway, time for me to go, goodnight all | 17:24.34 |
| AlecTaylor | Alright, well thanks very much Robin_Watts and kens. I will be sure to include the additional relevent information in my research paper, and proceed without reprocessing the PDF file, but just showing the XML as PoC | 17:24.35 |
| chrisl | ray_laptop: there was a FAPI/Freetype performance regression in 9.04, which I fixed, and actually fixed what I was trying to do when I introduced the problem - in theory, it should be slightly faster than 9.02, but I suspect it won't be measurable | 17:26.46 |
| ray_laptop | chrisl: we'll see. I'll let you know if the profile shows that it is font related in the HEAD rev. | 17:27.36 |
| chrisl | ray_laptop: okay, thanks., | 17:27.50 |
| | ray_laptop: and just in case, the -dDisableFAPI command option still works if you want back-to-back FAPI/AFS numbers. | 17:29.08 |
| ray_laptop | chrisl: thanks for the hint | 17:30.05 |
| henrys | chrisl:so how can shelly run the clusters but not have an account? | 17:31.50 |
| chrisl | henrys: I have no idea - does git have its own permissions? | 17:32.27 |
| Robin_Watts | chrisl: He does have an ssh account. | 17:32.41 |
| henrys | and he has gs-priv | 17:33.05 |
| | he's in the group | 17:33.11 |
| | my guess is he's trying to check into http read only. | 17:33.24 |
| chrisl | Okay, let me talk to him, and see if it's a settings problem...... | 17:33.28 |
| Robin_Watts | If he's using "git push" then he doesn't need to have an account. | 17:33.31 |
| | sorry "git cluster". | 17:33.41 |
| henrys | we want him to be able to commit and I don't see why he can't. | 17:34.33 |
| Robin_Watts | Me either. Tell him to log in here, and we'll talk him through it. | 17:36.34 |
| ray_laptop | if he cloned from the http, then his remote.origin may be wrong | 17:36.45 |
| chrisl | If the worst comes to the worst, he can come down here, and I'll hit his computer with a mallet until it works. | 17:36.52 |
| henrys | ray_laptop:that's what I suspect | 17:36.55 |
| ray_laptop | have him check his git config -l | 17:37.01 |
| Robin_Watts | git remote -v | 17:37.16 |
| | chrisl: Is there a reason he doesn't log in here ? | 17:37.46 |
| ray_laptop | Robin_Watts: yeah, that has the info too :-) | 17:37.47 |
| chrisl | Robin_Watts: timing, mostly. | 17:38.16 |
| ray_laptop | with so many people here with strange nicks, how can you tell he's not | 17:38.27 |
| Robin_Watts | He'd have to be up at VERY odd hours for someone not to be here :) | 17:38.44 |
| henrys | he's been on before - IRC may not be compatible with his work situation. | 17:39.10 |
| ray_laptop | there's alway webchat | 17:39.30 |
| | s/alway/always/ | 17:39.41 |
| Robin_Watts | henrys: Presumably he's not working for us while *at* work.. | 17:39.47 |
| ray_laptop | Robin_Watts: why not ? | 17:40.04 |
| | ;-) | 17:40.10 |
| henrys | that's what I'd do, why waste my free time ;-) | 17:40.11 |
| chrisl | Robin_Watts: I'd probably best not comment on that...... | 17:40.39 |
| AlecTaylor | :P | 17:41.35 |
| chrisl | henrys: I've dropped a mail to Shelly, saying what we think - he'll give me a call if it's not what we think | 17:47.01 |
| henrys | chrisl:please do invite him to irc next time you talk, I assume I don't need to answer his last email since youare are talking to him. | 17:47.08 |
| | oops should have said that sooner. | 17:47.25 |
| chrisl | henrys: I do mention IRC whenever we talk - I'll keep doing so | 17:48.36 |
| henrys | thanks | 17:49.01 |
| chrisl | henrys: it would be good if you could reply to him wrt to the billing - seeing us right for the bounty fix that was wrong | 17:50.48 |
| henrys | chrisl:will do. | 17:51.37 |
| chrisl | Thanks - that's the proverbial "above my pay grade" ;-) | 17:52.01 |
| henrys | for those of us that have been kicked upstairs before we break something ;-) | 17:52.59 |
| chrisl | refuses to comment ;-) | 17:53.49 |
| henrys | I must warn you guys before the meeting I've cut my hair so try to recognize me. I've had some confusing encounters... | 17:54.42 |
| chrisl | Wow! Was it weighing you down too much? | 17:55.11 |
| Robin_Watts | henrys: What would you have got if you'd won the bet? | 17:56.11 |
| henrys | as you age it gets thinner and you've got cut it back so it'll get thicker. | 17:56.45 |
| | ;-) | 17:56.49 |
| Robin_Watts | henrys: I'm not sure that works... by that token I'd have *really* thick hair. | 17:57.26 |
| | :) | 17:57.39 |
| henrys | according to 23andme I have the genes for male pattern baldness but it hasn't taken hold yet. | 17:59.30 |
| ray_laptop | henrys: have you told Miles and Scott ? (it would be fun to see their reaction) | 17:59.32 |
| henrys | no I haven't told them. | 17:59.59 |
| | I've had a ponytail for 15 years, I didn't go down easily ;-) | 18:01.25 |
| mvrhel_laptop | wow | 18:02.00 |
| | I had told my kids about your long hair. Luckily Alden had seen you before otherwise they won't believe me when they see you in Miami | 18:02.37 |
| henrys | it's still longer than Robin_Watts' ;-) | 18:03.23 |
| Robin_Watts | mvrhel_laptop: Tell them an alligator got to it. | 18:04.10 |
| henrys | my kids were shocked, they don't have memory of me with short hair. | 18:05.12 |
| mvrhel_laptop | You should be faster in your runs now | 18:05.52 |
| henrys | anyway speaking of Florida it's be great to have a get together with everybody, I'm glad Miles invited spouses, families and all. | 18:06.15 |
| Robin_Watts | mvrhel_laptop: You're forgetting the Samson effect. | 18:06.20 |
| mvrhel_laptop | good point | 18:06.27 |
| henrys | the stats so far indicate a slowdown but we'll see. | 18:06.43 |
| AlecTaylor | lolol | 18:07.21 |
| henrys | chrisl:I notice the jbig2 stuff is going to get complicated if he has sjbig2.c changes - have we looked at git externals yet? This business of having 2 repos for jbig2 is a problem. | 18:14.14 |
| chrisl | henrys: I thought we wanted to avoid externals | 18:15.09 |
| Robin_Watts | god yes, we want to avoid externals. | 18:17.13 |
| | not even sure it's possible with git. | 18:17.18 |
| chrisl | git submodule | 18:17.34 |
| henrys | well I thought git externals were better than svn externals and tor was going to ease us into it. But I haven't read about git externals myself. | 18:17.39 |
| ray_laptop | my kids were shocked when I showed them pictures of me with a beard | 18:20.58 |
| | henrys: did you donate your hair ? | 18:21.23 |
| | my beard wasn't long enough to donate ;-) | 18:21.55 |
| AlecTaylor | needs a haircut, his hair is almost down to his lobes! | 18:22.01 |
| AlecTaylor | shaved for the first time in 2.5 weeks a few days ago | 18:22.17 |
| henrys | ray_laptop:yes locks of love | 18:22.18 |
| chrisl | Robin_Watts, henrys: git submodules pulls in a specific revision of the sub-project - so we'd have the same issue of not being able to commit to it. | 18:22.28 |
| ray_laptop | henrys: good going !! | 18:22.30 |
| henrys | chrisl:so what to do about jbig2 continue to maintain it in 2 repos? | 18:24.19 |
| Robin_Watts | Why can't we make the ghostpdl repo the 'one true repo' for jbig2? We can do releases in turn with the ghostscript releases. | 18:25.30 |
| chrisl | henrys: if I'm honest, my inclination would be to make the ghostscript jbig2dec the "canonical" repos, and kill the other one | 18:25.35 |
| henrys | chrisl:the mail I've sent so far suggests an API change requires updates in both repos and regular fixes go to the standalone jbig2, then you'll grab that as you please. | 18:25.38 |
| | I tried that and tor didn't like it. | 18:25.58 |
| | but I guess he can be outvoted - I didn't know you guys would buy into that also. | 18:27.00 |
| tor8 | henrys, chrisl: that'l be painful for mupdf thirdparty libraries, but as long as we provide release tar balls of jbig2dec we're not in a worse spot than with zlib | 18:27.34 |
| chrisl | tor8: where do you get the jbig2dec source for the third part libs? | 18:28.13 |
| tor8 | at the moment I have a (private, on my machine) git repo which uses submodules to pull in all third party libs | 18:28.22 |
| | from the jbig2dec git on casper | 18:28.34 |
| | there are many ways to have multiple repos for one project with git | 18:29.14 |
| | submodules is just one way | 18:29.20 |
| | android uses the 'repo' tool | 18:29.26 |
| henrys | chrisl:so can I tell shelly to just check into gs? | 18:29.29 |
| chrisl | henrys: for now, I think so. Let's see if we can hammer out a solution keeping the separate repositories - if we do, we'll up date it from the gs code | 18:30.56 |
| tor8 | henrys: if we want to start using submodules I can set that up, but it means all of us have to run a few more git invocations to keep things in sync | 18:31.04 |
| | chrisl: I wonder if git cherrypick will work for that, though I think the different paths will pose a problem | 18:31.57 |
| chrisl | tor8: could we use some triggers so that commits into the gs/jbig2dec get mirrored to the jbig2dec repos and vice versa? | 18:32.22 |
| henrys | tor8:maybe something for the meeting - I wouldn't expect a warm reception to new git training. | 18:32.38 |
| tor8 | chrisl: I think it'd be relatively easy to make a script pull and update the jbig2dec sources in gs if we use the external jbig2dec git as a master | 18:32.56 |
| | henrys: yeah. conceptually git submodules work great, but they aren't trivial to use :( | 18:33.27 |
| | you need to run a few extra commands to keep the subrepositories updated | 18:33.47 |
| chrisl | Yeh, it was the other way (gs->jbig2dec) I was wondering about | 18:34.01 |
| Robin_Watts | tor8: Why not have a script that automatically recommits any changes to the ghostscript version into the other repo ? | 18:34.01 |
| ray_laptop | since jbig2dec doesn't get updated frequently, it might work | 18:34.14 |
| henrys | tor8:the immediate issue is shelly has a fix that requires a change to gs (sjbig2.c) and jbig2 - somehow this must be syncronized. | 18:34.21 |
| ray_laptop | henrys: but jbig2 _is_ in the gs repos | 18:34.56 |
| henrys | ray_laptop:it is in 2 places. | 18:35.16 |
| ray_laptop | is the problem the shared lib staying in sync ? | 18:35.39 |
| tor8 | henrys: that's not a problem if we use git submodules (since we'd update the submodule version in the same commit as sjbig2.c) | 18:35.44 |
| ray_laptop | votes for disallowing shared lib support in gs builds ;-) | 18:36.59 |
| tor8 | henrys: one approach is to make the jbig2dec.git auto-generated by filtering out all non-jbig2dec related stuff from the main repo | 18:37.08 |
| tor8 | agrees with ray! I hate shared libraries. | 18:37.24 |
| henrys | fair enough - the simplest thing is to make gs canonical and do jbig2 tarballs but that solution seems to have slipped away. | 18:37.25 |
| Robin_Watts | What tor8 just said. | 18:37.36 |
| tor8 | henrys: let's start with that, and I'll work on a script to to create a read-only jbig2dec subset git | 18:37.52 |
| Robin_Watts | (about auto-generating from the main repo). | 18:37.54 |
| henrys | tor8:I'm sure chrisl will appreciate that. | 18:38.13 |
| tor8 | it's very similar to what I did when converting the svn to git repos | 18:38.16 |
| | henrys: put it on the tech agenda so you can remind me later | 18:39.11 |
| chrisl | tor8: if it's doable that'd be great, but if it's a problem I don't mind adding updating of the jbig2dec repos to the Ghostscript release process - we just need to make sure people don't generally commit to it. | 18:39.55 |
| tor8 | chrisl: we can add hooks to disable commits, or just change the file permissions. | 18:40.30 |
| henrys | tor8:will do about the agenda for now shelly will commit to gs - and any bug that affects mupdf you'll get one way or another. | 18:40.43 |
| tor8 | the issue is filtering the whole gs history to create a jbig2dec repo is rather cpu expensive, so we probably don't want to do that on every commit | 18:41.05 |
| chrisl | henrys: is this a jbig2dec API change, or just a change in how GS uses it? | 18:41.35 |
| | tor8: once a week is probably (more than!) enough | 18:41.59 |
| tor8 | chrisl: indeed! | 18:42.28 |
| chrisl | And if even that uses too much CPU time - once a month...... | 18:43.19 |
| Robin_Watts | tor8: Once we've generated it, can we not do incremental updates? | 18:43.38 |
| | i.e. use a hook on the golden repo on casper so that whenever we commit to ghostpdl in the jbig2dec dir, it recommits to the jbig2dec standalone ? | 18:44.23 |
| tor8 | Robin_Watts: I'll have to look into it, but I think it should be possible. | 18:44.32 |
| Robin_Watts | That way we get it instantly in sync, for low cost. | 18:44.39 |
| tor8 | git filter-branch is what I was thinking of using | 18:44.51 |
| | git filter-branch --subdirectory-filter gs/jbig2dec --prune-empty ...revs... | 18:45.25 |
| henrys | chrisl:now I can't find the patch, hang on. | 18:45.51 |
| chrisl | henrys: it's really for tor8's benefit - if the jbig2dec API changes, mupdf will need to be aware of it, too. | 18:46.44 |
| falko_ | hi is it possible to create a rpt port to ghosscript and if i print i print to a printer, create a psfile and after that start a executable (that moves the ps file around) ? | 18:48.18 |
| | at the moment i have a script that checks every 3 sec for a new file ,.. does the printing to the printer and teh moving around .. but that is not a very nice solution | 18:49.32 |
| henrys | chrisl:apparently the sjbig2.c is reference in his email as being in the patch but it isn't there? | 18:49.49 |
| chrisl | henrys: he did ask how we wanted it - I may have misinterpreted what he said. | 18:51.02 |
| henrys | figure it out later if he commits it's easy enough to back out. | 18:55.30 |
| ray_laptop | mvrhel_laptop: the color stuff is DEFINITELY in the top of the profile for the performance (904 vs 871) :-( | 18:58.34 |
| mvrhel_laptop | :( | 18:59.00 |
| | ray_laptop: if you want to dump this on me, that is fine | 18:59.19 |
| ray_laptop | mvrhel_laptop: cmsSample3DGrid is particluarly heavy. | 18:59.34 |
| | mvrhel_laptop: I think the issue is that 871 runs with 'simple' color and 90x effectively always runs with -dUseCIEColor | 19:00.12 |
| chrisl | henrys: Okay, getting Shelly able to commit should resolve the API question from a GS point of view, and I'll make sure to tell him to inform tor8 if the API really has changed. | 19:00.56 |
| mvrhel_laptop | ray_laptop: yes. I guess what I need to do is to get the "dumb" CMM in place | 19:01.18 |
| | for those who don't want color management | 19:01.28 |
| ray_laptop | mvrhel_laptop: we'd need a 'dumb' cms that totally ignored the ICC profiles | 19:01.33 |
| mvrhel_laptop | hehe | 19:01.38 |
| | that is on my todo list. and I did start it | 19:02.04 |
| henrys | what is the effect of just using the dump profiles for now? | 19:02.17 |
| ray_laptop | mvrhel_laptop: but the complication is that we can only do that for the 'default' colorspaces -- If Lab or ICC comes in, we still need to handle it | 19:02.30 |
| mvrhel_laptop | henrys: they will still use lcms | 19:02.52 |
| | ray_laptop: yes | 19:03.00 |
| | I need to think a bit about this | 19:03.07 |
| henrys | mvrhel_laptop:it may be unique to pcl but many jobs run significantly faster with the dump profiles. | 19:03.29 |
| mvrhel_laptop | interesting. | 19:03.39 |
| | I do have an idea though how to do this | 19:03.46 |
| | basically we have an interface between the graphics library and the CMM and we know when the profile was a substituted one for a defaultRGB etc | 19:04.24 |
| | this is needed for PDFwrite | 19:04.29 |
| ray_laptop | mvrhel_laptop: if the "quick and dirty" cmm recognized the input ICC profiles enough to recognize that it is RGB, Gray or CMYK then the 'link' profile it returns can be a suitable dummy that the 'quick and dirty' color conversion code can recognize and shortcut | 19:04.36 |
| henrys | I think there are many cases in the code where you don't get "pure" colors with the fancy profiles. | 19:04.44 |
| mvrhel_laptop | yes exactly | 19:04.45 |
| | let me look at adding that as an option | 19:05.07 |
| ray_laptop | mvrhel_laptop: OK. But let's decide when that fits with your other priorities | 19:05.36 |
| Robin_Watts | I'm about to risk breaking the cluster. | 19:07.22 |
| | Anyone have any jobs they want to get in desperately? | 19:07.32 |
| ray_laptop | Robin_Watts: not me | 19:07.39 |
| mvrhel_laptop | well I am doing a run right now | 19:07.49 |
| | oh it just finished | 19:07.55 |
| Robin_Watts | This will only affect new jobs. | 19:08.01 |
| mvrhel_laptop | I am done. Going to do a commit now | 19:08.23 |
| Robin_Watts | ok, I've swapped to a modified run.pl that should at least get the build right on windows cluster machines. | 19:08.58 |
| mvrhel_laptop | actually I may run one more real quick | 19:09.03 |
| Robin_Watts | Go for it. | 19:09.09 |
| | In theory with no windows cluster machines you should see no differences. | 19:09.19 |
| mvrhel_laptop | ok | 19:09.27 |
| henrys | Robin_Watts:oh are you getting the windows node going? | 19:10.20 |
| Robin_Watts | henrys: trying to. | 19:10.29 |
| tor8 | chrisl, kens: more ammo if you don't like the _t suffix ... it's reserved by POSIX so we shouldn't be using it in the first place! | 19:17.38 |
| ray_laptop | you mean like tin64_t ? | 19:19.23 |
| tor8 | yeah | 19:19.36 |
| ray_laptop | int64_t that is :-/ | 19:19.39 |
| tor8 | posix makes the claim on *_t in ANY header | 19:20.11 |
| ray_laptop | that seems a bit excessive | 19:20.47 |
| | what happened to C standard | 19:21.07 |
| tor8 | yeah, but _t is an ugly wart so I don't complain :) | 19:21.15 |
| | the C standard, it got let out in the wild... :/ | 19:21.45 |
| | looking at the list of name spaces that posix lays claim on, it's pretty excessive even without *_t! | 19:22.42 |
| Robin_Watts | ok, we have a windows cluster node... and it's pinging the cluster, let's try doing a build. | 19:22.56 |
| | My cygwin installation appears to be having problems. Gah. | 19:25.38 |
| ray_laptop | btw, the HEAD rev is virtually the same on this test as 904, so FT isn't bolliixing anything up | 19:25.42 |
| henrys | bbiab | 19:29.07 |
| mvrhel_laptop | bbiaw | 19:45.50 |
| Robin_Watts | mvrhel_laptop: I've taken my node offline, so your job should run without a problem. | 19:48.35 |
| mvrhel | Robin_Watts: you around? | 20:51.25 |
| Robin_Watts | I am. | 20:51.31 |
| mvrhel | so I was just testing 692512 to see if we could close this | 20:51.55 |
| | and I see that we get horizontal lines at 600dpi | 20:52.09 |
| | I thought we had this fixed | 20:52.18 |
| Robin_Watts | so did I :( | 20:52.46 |
| mvrhel | the bug was originally for vertical lines | 20:52.50 |
| | I have a single page version of this file if you want to take a look at it | 20:53.06 |
| Robin_Watts | Assign it to me, and I'll have a look when I uncluster myself. | 20:53.09 |
| mvrhel | ok. | 20:53.15 |
| | bbiaw | 20:55.36 |
| Robin_Watts | Damn. Looks like I really am going to need to check out another copy of tests/tests_private. | 21:08.01 |
| | dinnertime. | 21:08.21 |
| mvrhel_laptop | stormy here. we may lose power | 23:43.32 |
| | Forward 1 day (to 2011/11/17)>>> | |
IRC Logs