| <<<Back 1 day (to 2012/03/19) | 2012/03/20 |
jeagle_ | Has anyone used ghostscript and had an issue where the image of the pdf was viewable but the text is missing? XD | 01:39.32 |
| Hello! | 01:39.33 |
| Is anyone alive in here? :P | 01:41.37 |
henrys | jeagle_:does your program display properly with the free adobe pdf reader? If so report a bug at bugs.ghostscript.com, thanks for using ghostscript. | 02:08.30 |
kens | chrisl could you look at this on the Stack Overflow site please ? | 08:32.45 |
| http://stackoverflow.com/questions/9782774/upgrading-ghostscript-9-02-to-9-05 | 08:32.45 |
chrisl | kens: RTFM? http://www.ghostscript.com/doc/9.05/Make.htm#MacOSX - see the "NOTE" | 08:39.03 |
kens | Hee hee | 08:39.28 |
chrisl | kens: that is a guess, but it looks like the same problem..... | 08:39.39 |
kens | Will you post that sa a response then or want me to do it ? | 08:39.39 |
chrisl | I'll do it.... | 08:39.46 |
kens | Thanks | 08:39.51 |
| I've already answered three questions and I've had enough for today | 08:40.14 |
chrisl | I've still had enough from yesterday..... :-( | 08:40.32 |
kens | OK just got Avadhut's email to read now, been putting it off till last | 08:42.27 |
chrisl | Leaving that to Marcos..... | 08:45.11 |
kens | Yes me too. At least its only one dumb question this time | 08:45.26 |
| Now I have a free user problem to look at on comp.lan.postscript, then I can try and get some real work done. | 08:45.56 |
chrisl | We had a minor panic about the Brother problem last night.... chucking the ps2write output straight at the printer caused *very* strange problems. But I added the PJL and PS preample that the CUPS filter adds, and it work - thanks heavens! | 08:47.15 |
| kens: I take it your talking about the font subsetting thing on comp.lang.postscript? | 08:51.19 |
kens | Yes that's the one | 09:00.28 |
chrisl | OKay, there's also one about colour transforms, so I'll pen a reply to that. | 09:00.55 |
kens | Is the the grey-RGB one ? | 09:01.12 |
| I meant to look at that and forgot | 09:01.21 |
chrisl | Yes. He doesn't mention which version of GS their using, which makes it hard to answer..... | 09:01.51 |
kens | Right, I was going to run a test and see what happened, but it slipped my mind | 09:02.10 |
chrisl | I'm not feeling generous enough to run a test! | 09:02.38 |
tkamppeter | chrisl, thanks again for your great work on the PS interpreter bugs of Brother printers. Now it is finally solved and I can issue the fixed cups-filters in time for Precise beta2. | 09:08.44 |
kens | Great news tkamppeter, thanks for letting us know | 09:08.59 |
tkamppeter | kens, also thanks for your help on this. | 09:09.15 |
kens | No problem, happy to oblige, thanks should go to the tetsers for their work too fo course. | 09:09.43 |
chrisl | tkamppeter: I did have a minor panic last night when I saw Bruce's reply - I'm glad it turned out to be what I thought! | 09:11.13 |
kens | OK finally got around to buying an upgrade to Acrobat X | 10:05.59 |
|Frederik | Not sure if I'm in the right channel here... Every time I print a certain document (one page of a PDF file) on a HP LaserJet 4250 with Debian Wheezy, it prints this error: http://pastebin.com/FpAHWNRK . Could this be a ghostscript problem? | 10:32.12 |
kens | Well CUPS is converting the PDF to PostScript using Ghostscritp I expect. It 'looks like' the problem may be a pritner bug, but we would need to see the PostScript file to be sure. | 10:33.05 |
|Frederik | Is there a way to get hold of the created ps file? | 10:40.19 |
kens | If you are using CUPS, yes. There's a document somewhere describing it | 10:40.38 |
chrisl | https://wiki.ubuntu.com/DebuggingPrintingProblems | 10:40.54 |
kens | Thanks chrisl | 10:41.00 |
chrisl | goes to worship at the alter of the caffeine gods..... | 10:46.28 |
|Frederik | thanks | 10:47.29 |
| I placed the file here: https://artipc10.vub.ac.be/~frederik/d00095-001 | 10:47.38 |
kens | That's a PDF file | 10:48.22 |
| Not a PostScript file | 10:48.27 |
| Is that really what CUPS is sending to the printer ? | 10:49.04 |
| The data sheet for that series pritner does not mention direct PDF printing, so I don't think that you are sending the PDF file to the printer. | 10:51.38 |
| Oh actually it does say it supports PDF 1.3 | 10:52.07 |
| Of course yours is a PDF 1.5 file | 10:52.15 |
| If you are certain that the PDF file you have sent is what is being sent to the printer, then Ghostscript is not involved. The PDF file was produced by cairo apparently. | 10:54.12 |
kens | gets bitten by the GS macro frenzy again | 10:57.26 |
|Frederik | that was definitely the file which was in /var/spool/cups. Printer model is HP LaserJet 4050 Series | 11:09.05 |
kens | Well, it sounds like you are trying to send a PDF 1.5 file to a prtiner which only supports PDF 1.3. THe error message is surprising, and unhelpful, if that is the case but I would not expect that file to work if that is true. | 11:10.06 |
| I'm not enough of a CUPS expert to help you any further, and Ghostscript does not seem to be involved in any way, so I don't think we cna help you any further here. | 11:10.41 |
| If it turns out that PostScritp is being sent to the printer, feel free to come back and poke us again. | 11:11.02 |
|Frederik | OK, I'll ask around in #cups then | 11:11.31 |
| thanks for the help | 11:11.41 |
kens | Oh, you could convert the PDF to a PostScript file using the Ghostscript 'ps2write' device, and see if sending that to the pritner directly works any better | 11:11.43 |
| You are quite welcome | 11:11.49 |
chrisl | |Frederik: there are a couple of LJ4050 drivers that are (supposedly) Postscript: "HP LaserJet 4050 Series Postscript" and ""HP LaserJet 4050 Series Foomatic/Postscript" | 11:14.37 |
| Trying each of those might help your problem..... | 11:14.52 |
kens | Oops, I looked up the 4250.... | 11:15.33 |
| The 4050 doesn't support PDF so it look slike there must be *some* conversion to PostScript going on | 11:16.33 |
chrisl | Hmm, maybe that document is out of date since cups became a PDF workflow - should check with tkamppeter about that..... | 11:18.43 |
kens | That could be the problem. | 11:19.05 |
Robin_Watts | chrisl: We have to register with google for the android market. | 11:31.57 |
| Rather than registering under robin.watts@artifex.com, I was considering registering as something like android@artifex.com | 11:32.22 |
| Do we have the ability to set up such names? | 11:32.35 |
chrisl | Robin_Watts: No, you'll need to get marcosw to do it | 11:32.52 |
Robin_Watts | I meant 'we' as in 'artifex' rather than specifically you or I, but I guess you answered it :) | 11:33.17 |
| IIRC mailboxes cost us, but aliases don't. | 11:33.28 |
chrisl | Yes, sorry (mind is somewhere else). That's right, although we have a spare mail box available - it might be useful to assign one for this type of thing, and then use aliases for specific uses. | 11:34.39 |
tkamppeter | kens, chrisl, some minutes ago I released the first native PDF printer PPD file for Linux/CUPS ever. |Frederik will have most probably a setup where the PDF is converted to PS before being sent to the p[rinter. | 13:02.12 |
Robin_Watts | paulgardiner: ping | 13:03.47 |
|Frederik | tkamppeter: this is the ppd file on my system for that printer: http://artipc10.vub.ac.be/~frederik/printer.ppd | 13:06.02 |
tkamppeter | kens, chrisl, |Frederik, the PDF is converted to PostScript by CUPS, on Ubuntu Precise (12.04) actually by Ghostscript with the "ps2write" device, in older Ubuntu most probably by Poppler. One can see in error_log (in debug mode) what gets actually used. | 13:06.34 |
| kens, |Frederik, the d... files in /var/spool/cups are what comes from the app, not yet filtered by CUPS. | 13:08.01 |
| |Frederik, kens, the PPD is the original HP PPD file for these prnters. With this CUPS knows that it needs to send PostScript and therefore it turns PDF into PostScript. | 13:09.22 |
kens | just back from lunch | 13:10.18 |
|Frederik | here is cups error_log for that job which fails: http://pastebin.com/0hQ4EGJc (yes, I made an annoying mistake, I called the printer queue hp_LaserJet_4250, but it really is a 4050) | 13:10.56 |
kens | OK tkamppeter can you help |Frederik get a PostScript file for us to look at ? | 13:10.57 |
| Well I see it invoked GS to convert to PostScript | 13:12.00 |
| It made a temp file, I wonder if there's a way to capture that ? | 13:12.44 |
tkamppeter | |Frederik, kens, first, follow the instructions under "CUPS error_log" on https://wiki.ubuntu.com/DebuggingPrintingProblems. With this you can find out which filters get actually used. | 13:12.49 |
| There is a cpdftocps in older Ubuntu/Debian versions, this converts PDF to PostScript using Poppler. | 13:13.24 |
| In Precise there is pdftops which actually uses Ghostscript. | 13:13.47 |
kens | tkamppeter, looking at the error log, there is PostScript which is being produced by GS (I can tell from the PS comments) | 13:14.05 |
| SO I believe it is using GS | 13:14.14 |
tkamppeter | |Frederik, kens, to get the resulting PostScript, you have to clone the print queue into one which prints into a file: | 13:14.47 |
| cupsctl FileDevice=yes | 13:15.00 |
kens | Seems reasonable | 13:15.10 |
|Frederik | (I'm using cups-filters 1.0.2 in Debian Wheezy, should I try cups-filters 1.0.5 from SID?) | 13:15.28 |
tkamppeter | lpadmin -p test -E -v file:/tmp/printout -P /etc/cups/ppd/printer.ppd | 13:15.30 |
| |Frederik, this can help, there are lots of bug fixes in cups-filters. Today I released 1.0.6, more bug fixes, native PDF printer support, ... Is already in Ubuntu Precise, but, unfortunately, the Debian maintainer is sick today, so in Debian it comes tomorrow at the earliest. | 13:17.09 |
| |Frederik, kens, print into the queue 'test' and find the PS in the file /tmp/printout. | 13:18.02 |
|Frederik | Here is the printout file: https://artipc10.vub.ac.be/~frederik/printout | 13:40.53 |
kens | OK well it looks fine. | 13:43.15 |
| Will take some tiem to check | 13:43.20 |
| PS file works OK on Ghostscript | 13:58.25 |
| Works OK on distiller | 13:59.59 |
| Jaws throws an error on the page size. | 14:01.00 |
Robin_Watts | Morning tor8. Feeling better? | 14:01.46 |
tor8 | Robin_Watts: barely... didn't get much sleep last night either | 14:03.19 |
Robin_Watts | tor8: ew. | 14:04.32 |
| What remains to be done before release of mupdf 1.0 ? | 14:04.44 |
tor8 | heartburn episodes, all through the night :( | 14:04.47 |
| iOS app, need to review the latest version of the android app, update the readme and that's all I can think of | 14:07.15 |
| Robin_Watts: you leaving tomorrow? | 14:08.59 |
Robin_Watts | I am. | 14:09.15 |
| So, is there anything I can usefully be doing ? | 14:09.29 |
| I may pull in zlib-1.2.6 | 14:09.59 |
tor8 | rewrite the README and make a news blurb for the mupdf.com page :) | 14:10.06 |
Robin_Watts | OK. | 14:10.13 |
tor8 | I doubt you want to touch the iOS app, and I don't think I'll have energy enough for doing a lot | 14:10.36 |
Robin_Watts | I really don't want to be doing the iOS app :) | 14:11.19 |
| I'm going to get marcos to set up an alias for android@artifex.com and then I'll register with google with that address. | 14:11.51 |
tor8 | are all paulgrahams android app changes merged into master? | 14:11.54 |
Robin_Watts | tor8: Yes. | 14:11.58 |
tor8 | great. | 14:13.18 |
kens | Oh great the AIX link error is in openjpeg... | 14:28.35 |
tor8 | Robin_Watts: have you tried the new link button in the android app? It's not very reliable on my tablet. | 14:36.37 |
Robin_Watts | tor8: I haven't. | 14:36.56 |
| Tweaked README pushed. | 14:37.41 |
tor8 | turning on the link highlight doesn't take effect until new pages are redrawn | 14:38.35 |
| rotating the device forgets the state of the button | 14:38.43 |
| and the highlights seem to be incorrectly placed on some pages | 14:39.02 |
| so I think I'll just turn them off for the 1.0 release and we can sort them out in a later update | 14:39.22 |
Robin_Watts | ok. | 14:39.31 |
| The first 2 of those should be easy to solve (I hope) | 14:39.49 |
tor8 | yeah, it's the last one I worry about | 14:40.08 |
| I'm also not too happy with the three-state button in practice | 14:40.19 |
| it was worth trying out, but it's kind of awkward in use | 14:40.44 |
paulgardiner | tor8: is that the link highlight? I thought I checked whether it redrew, although I remember thinking I didn't check it on the hardware assisted tablet | 14:44.42 |
tor8 | paulgardiner: yes. | 14:44.54 |
paulgardiner | Damn! I meant to ask Robin to test it on the tablet, but forgot. :-( | 14:45.22 |
| Aagh! And I never thought about retaining the state on rotation. | 14:46.09 |
| I could probably fix that fairly easily, but the incorrectly placed highlights are more worrying. Best turn it off, as you suggest. Sorry about that. | 14:47.23 |
tor8 | one of the early pages in pdfref13.pdf has the highlights in weird places | 14:47.49 |
paulgardiner | tor8: I thought you asked for a three-state button. I must have misunderstood. Or did you mean the look of it? My artistic tallents are a little lacking. | 14:48.54 |
tor8 | paulgardiner: Robin asked for the three-state button :) | 14:49.13 |
| and I thought we should try it out | 14:49.23 |
paulgardiner | Ah. | 14:49.24 |
tor8 | but I'm not sure I like how it behaves. the artistic talent thing can be solved, that's not the issue :) | 14:49.48 |
paulgardiner | Have you tested the search with it's progress bar? | 14:49.49 |
kens | tkamppeter, is there a way to get the PDF job converted to PostScript exactly as it is on Linux, but with -dCompressPgaes=false -dCompressFonts=false on the command line? | 14:50.45 |
tor8 | paulgardiner: yeah, that's a nice touch! | 14:51.47 |
| I love how it counts down the % when you search backwards | 14:51.59 |
paulgardiner | Ah good. Glad that seems to be working ok. | 14:52.11 |
tor8 | would it be hard to add a timeout so it doesn't appear until 200ms have passed or so? | 14:52.23 |
| when you get hits on successive pages it flickers needlessly, would be a nice touch to not do that | 14:52.40 |
paulgardiner | tor8: yes that shouldn't be too hard I'd have thought, and I too didn't like the flicker. | 14:53.28 |
tor8 | paulgardiner: page 4 of pdfref13.pdf has the highlights in what looks to me very wrong places. maybe it's got the highlights for the wrong page? | 14:54.42 |
paulgardiner | Although it would be difficult to predetermine that it wont just flicker after 200ms. Something makes me think that would be less of a problem, but I'm not sure of the logic behind it | 14:54.46 |
tor8 | yeah, we can't eliminate it without looking into the future :) | 14:55.16 |
| but at least we can solve the most annoying case | 14:55.27 |
paulgardiner | Any of this worth trying to fix before the release? | 14:55.47 |
Robin_Watts | Given that the ios app doesn't have the button at all, it's not unreasonable to leave it out of the android app for consistency. | 14:56.33 |
paulgardiner | Yes true. | 14:57.00 |
tor8 | paulgardiner: that's up to you. we had hoped to do the release before Robin Watts left tomorrow, but then I was sick all day yesterday and not feeling too well today either so it may slip a day | 14:57.31 |
| but I think if we want to push it out soon, we can disable the links temporarily for the release and then work on getting them in good shape on both apps later | 14:58.13 |
paulgardiner | I guess there's nothing stopping independent releases of the apps at a later date. | 14:58.50 |
tor8 | there is something else that bothered me with the links. when following one, I instinctively hit the 'back' hardware button and it took me out of the mupdf app where I sort of expected to go back to the page I was at before following the link :) | 14:59.06 |
paulgardiner | I'll have a quick look at delaying the search progress dialog though. | 14:59.11 |
sebras | tor8: there are some XXX in the headers I should sort out before 1.0. | 15:00.18 |
tor8 | WARNING: Attribute minSdkVersion in AndroidManifest.xml (8) is lower than the project target API level (11) | 15:00.34 |
Robin_Watts | sebras: really? | 15:00.52 |
sebras | Robin_Watts: link_dest I belive. | 15:01.01 |
Robin_Watts | Ah, yes. | 15:01.12 |
sebras | believe. | 15:01.13 |
tor8 | link_dests | 15:01.14 |
sebras | at least I spelled "I" correctly... ;) | 15:01.37 |
paulgardiner | tor8: yes, that would be nice for "back" to undo the link jump. Could be slightly messy in andriod though. Would require the app to take over the back button, I think, which risks breaking other things. Shouldn't be rocket science, but would want a good period of testing after. | 15:02.26 |
tor8 | paulgardiner: yeah, not something for 1.0 :) | 15:02.48 |
| that warning though, anything we should worry about? | 15:02.56 |
paulgardiner | warning? | 15:03.10 |
| oh | 15:03.15 |
Robin_Watts | I think we only moved the target API level up to see if that would help with the 'x-large' screens, right ? | 15:03.35 |
tor8 | Robin_Watts: wasn't it to turn on HW acceleration on 3.0 devices? | 15:04.01 |
Robin_Watts | tor8: Wasn't x-large screens what turned on HW acceleration? | 15:04.50 |
| (goldfish memory strikes again) | 15:05.04 |
| http://ghostscript.com/~robin/BLURB | 15:05.20 |
paulgardiner | Yes, I think that ok. I think the warning is alerting us to the possibility that we may be using features from above 11, while claiming we can run on 8, but that should be ok, because the only non-8 fearture is a line in the xml | 15:05.33 |
| If we were using a level 11 method, we should be putting a test around it, but I'd hope the xml line was ok | 15:06.12 |
| "features from above 8" I meant | 15:06.31 |
Robin_Watts | paulgardiner: Right, so we could drop back to level 8, and the xml would still work, right ? | 15:06.33 |
paulgardiner | That's what I believe | 15:06.46 |
Robin_Watts | Can anyone spot anything I've missed from the BLURB ? | 15:06.50 |
tor8 | I changed the 11 to 8 and that still builds and runs with the same performance | 15:06.57 |
paulgardiner | tor8: that's interesting. That's a test I never thought of. | 15:07.25 |
tor8 | I now get the little "Stretch/Zoom to fill screen" chooser | 15:07.27 |
| so that may have been the reason | 15:07.37 |
paulgardiner | I was expecting the use of hardware acceleration to be dependent on declaring 11, but sounds like that isn;t the case | 15:08.17 |
tor8 | oh, and the lists (file chooser, outline) look different. with 11 we get the MuPDF icon on the top but not so with 8. | 15:08.37 |
Robin_Watts | OK, so stick with 11 and ignore the warning ? | 15:09.18 |
tor8 | I haven't tested it on my phone with 11 | 15:09.56 |
paulgardiner | Runs on my Galaxy S2 and on a HTC Desire | 15:10.25 |
| with 11 that is | 15:10.34 |
sebras | hm... isn't 11 android 3.0 honeycomb? | 15:11.10 |
| http://developer.android.com/resources/dashboard/platform-versions.html | 15:11.25 |
tor8 | yeah, with 11 we get the icon in the list headers. looks a bit odd to be honest. | 15:11.30 |
sebras | I believe there are a lot of devices out there with 2.3 gingerbread on them that you may not want to exclude... | 15:11.56 |
tor8 | bah. there are *two* places you have to set the target. | 15:12.42 |
| if you set android-8 in project.properties it fails to build | 15:12.54 |
| if you set android-8 in AndroidManifest.xml it works as I described | 15:13.09 |
paulgardiner | sebras: with minSDKVersion set to 8, we should still run ok on 2.3 | 15:13.22 |
tor8 | if project.properties version is 8, it barfs on the hardwareAccelerated declaration | 15:13.33 |
Robin_Watts | sebras: I have a 2.3 device here (my phone) | 15:13.39 |
| sebras: I've done the BLURB and the README, so I'll have a quick pass at the link_dest stuff now. | 15:15.30 |
sebras | Robin_Watts: ok. the reason I haven't been so active on the link_dest is that I really think that we could make that interface easier on the clients. | 15:17.13 |
Robin_Watts | sebras: Me too, but it's not trivial. | 15:17.30 |
| I may add a note saying that this interface is still subject to change. | 15:17.59 |
sebras | couldn't we have a enum at the top of link_dest and then a union struct for each type of link as outlined in pdfref? | 15:18.09 |
| adding such a note is probably worthwhile yes. :) | 15:18.32 |
Robin_Watts | The reason I didn't scribble stuff for it already was that we hoped to rework it slightly before 1.0, but I ran into a problem with it so didn't do it. | 15:18.47 |
| IIRC it was to do with when the data was available. | 15:19.17 |
sebras | the data? | 15:20.18 |
| oh you mean each of the points in the rect? | 15:20.38 |
| x y z, etc..? | 15:20.43 |
| paulgardiner: right, I was more worried about sticking with 11. then I can not enjoy MuPDF on my own phone! :) | 15:22.19 |
paulgardiner | Sorry. Didn't explain well. If we set the target to 11 and the min to 8, we should also be ok. | 15:23.15 |
tor8 | sebras: yes, the link destination coordinates need to be rotated and translated to match the mediabox orientation | 15:23.30 |
kens | tkamppeter, are you there ? | 15:23.35 |
tor8 | which needs looking up stuff in the page tree, which we currently do with pdf_load_page, but pdf_load_page is a bit too heavy | 15:24.28 |
Robin_Watts | sebras: what tor8 said :) | 15:24.55 |
sebras | tor8: right, that makes sense. but isn't this a problem with the current implementation? i.e. without havin enum kind at the top of link_dests? | 15:25.14 |
tor8 | no, but it's something that still needs to be solved | 15:25.28 |
Robin_Watts | sebras: It would be a problem however we solved it. | 15:25.46 |
sebras | right. | 15:26.16 |
| and the rotation should be handled before 1.0 I suppose. | 15:26.30 |
| anyway, since I hoped that we'd (I?) would change the interface I have not documented it. | 15:27.17 |
| I can put some effort on it tonight if you want to do more important matters Robin. | 15:27.38 |
Robin_Watts | Doing the rotation is similarly problematic, I thought. | 15:27.58 |
| I think we should mark the text device interface as subject to change too. | 15:28.18 |
tor8 | Robin_Watts: actually, it shouldn't be too hard to do | 15:28.37 |
Robin_Watts | tor8: oh, ok. | 15:28.47 |
tor8 | we just need to split part of pdf_load_page into pdf_page_transform (an internal function) | 15:29.10 |
| we have already loaded the page tree, so it should be a fairly light weight operation | 15:29.34 |
Robin_Watts | sebras: I don't believe that link_dests should exactly ape the PDF ones. | 15:30.23 |
| i.e. our representation should be capable of representing all the PDF ones, without tying us to just them. | 15:30.49 |
sebras | Robin_Watts: why? does xps have more variants? | 15:30.50 |
tor8 | we could probably drop the rotate and mediabox fields from the pdf_page struct and replace with a matrix that encompasses the same stuff | 15:30.51 |
Robin_Watts | it's not just xps, it's 'any format we want to add in future'. | 15:31.08 |
| tor8: I can imagine that some clients would prefer to be able to read a rotate value rather than having to retrieve it from a transform. | 15:31.53 |
sebras | Robin_Watts: yes, I know, but that only means you want to add another enum and another struct...? | 15:31.57 |
tor8 | Robin_Watts: the rotate field is internal | 15:32.23 |
Robin_Watts | oh, well then, I retract my objection. | 15:32.39 |
tor8 | one of the more recent changes (the one that broke the text lines to blocks thing yesterday, incidentally) was to hide the rotate and mediabox origin offset and 72dpi vs 96dpi of pdf/xps stuff from clients | 15:33.31 |
| so that from their point of view, all pages have the origin at the top left, y going down, at 72 dpi | 15:33.51 |
henrys | chrisl:does shelly have access to private attachments? I assume he does. | 15:35.56 |
kens | chrisl is away | 15:38.32 |
| But yes I think he can get private attachments | 15:38.45 |
henrys | kens:thanks | 15:39.01 |
kens | fetches coffee for meeting | 15:39.33 |
| Hi Ray | 15:46.28 |
henrys | kens:so you aren't a tea drinker? | 15:46.56 |
kens | Nah, can't stand the stuff | 15:47.05 |
| I need coffee, lots of coffee.... | 15:47.18 |
chrisl | henrys: I believe Shelly does have access to private attachments, but I'll double check with him. | 15:48.01 |
henrys | chrisl:I believe I understand the illusive 692816 now. PCL never actually closes the output file - it just releases the device. | 15:50.23 |
| odd the problem only happens occassionally. | 15:50.43 |
chrisl | henrys: surely releasing the device should also close the file? | 15:51.32 |
henrys | no you do have to call closedevice() | 15:51.49 |
chrisl | Oh, I see what you mean. Yes, it's strange we don't see problems more often..... | 15:53.06 |
henrys | you can see it with trace and any file the OutputFile never gets a close. | 15:53.11 |
chrisl | Is it a big change to address it? | 15:53.43 |
henrys | no problem I got it. I just recall you being interested in that one. | 15:54.13 |
chrisl | Well, we were sort of right - the IO buffers weren't being flushed properly........ | 15:54.54 |
kens | chrisl did you see we have another 'ps2write otuptu doesn't work with my printer' problem ? | 15:55.04 |
danilo_b | hello. someone from mupdf around here? :) | 15:55.23 |
kens | several | 15:56.01 |
danilo_b | i'd like to use mupdf for latex-previews. is there a way to autoreload a pdf file when it changes? | 15:56.05 |
chrisl | kens: I saw the beginnings of it, but I didn't see anything after tkamppeter's explanation about the cups spool files...... | 15:56.07 |
danilo_b | or can i trigger the reloading from the shell? | 15:56.15 |
tor8 | danilo_b: hit the 'r' button or send a sighup | 15:56.23 |
kens | chrisl the PS file works for me on GS and Distiller, but not Jaws (an error on setpagedevcie) | 15:56.28 |
danilo_b | tor8: sighup will do :) | 15:56.30 |
kens | chrisl but apparently fails on an HP pritner (4050) | 15:56.52 |
chrisl | kens: do you know why we're doing a setpagedevice? | 15:57.22 |
kens | One reason for definite; the CUPS spooler inserts a page size request | 15:57.43 |
| But it still fails if I remove that | 15:57.51 |
| I'm just decoding the ascii85 compressed file :-( | 15:58.05 |
danilo_b | tor8: apparently when i send mupdf a sighup signal, it only reloads after gaining the x focus. is there a way to get around this? | 15:59.03 |
tor8 | danilo_b: not sure, sebras implemented the sighup reloading | 15:59.29 |
henrys | and the meeting starts, not much from me this week but will we have the mupdf release this week before Robin_Watts leaves? | 16:00.13 |
Robin_Watts | henrys: We've been discussing that. We are very close. | 16:00.32 |
| tor8 wants to tweak fz_link_dests slightly. | 16:00.48 |
| the ios app needs to be double checked, and tor8 is just disabling something in the android app so it'll match the ios app. | 16:01.21 |
mvrhel_laptop | meeting time? | 16:01.38 |
henrys | I guess I'm a little concerned posting a release and having both robin_watts and tor8 (soon) on vacation. I wonder if it shouldn't wait until all the vacations end. | 16:01.42 |
tor8 | henrys: I was sick yesterday, and still recovering :( I've got the android app in good shape, still need to update the ios app for the latest text device changes. | 16:01.52 |
Robin_Watts | So, I believe we should ship this week, but maybe not quite before I leave. | 16:01.56 |
henrys | mvrhel_laptop:I started it, yes. | 16:02.12 |
mvrhel_laptop | Robin is going to have his laptop on the yak | 16:02.14 |
Robin_Watts | Yaks don't offer magsafe connectors though :( | 16:02.30 |
henrys | I think they have usb ports | 16:02.42 |
chrisl | I no *not* want to know where you stick the connector....... | 16:03.03 |
henrys | but anyway why don't we just hold off the release until tor8 returns so everyone is here if there is some calamnity? | 16:03.21 |
Robin_Watts | henrys: Given it's a 1.0, we could post it as a release candidate ? | 16:03.44 |
tor8 | henrys: there's one week of holiday overlap, during the easter so I'm not super worried. a release candidate and final release when we're both back works too, I guess. | 16:04.10 |
danilo_b | tor8: ok, i guess i'll add a feature request :) | 16:04.28 |
henrys | okay a candidate sounds good. | 16:04.35 |
tor8 | danilo_b: try adding a winrepaint call just after the winreloadfile call in the main loop in x11_main.c if that works? | 16:04.59 |
| that gives sebras time to change the link destinations to something he likes :) | 16:05.31 |
henrys | the other issue for the meeting I had was about the bug aging list. It seemed like alexcher could have bisected one of his bugs (regresssion) | 16:05.35 |
kens | chrisl I found the PS file has a *load* of setpagedevice calls which are specific to the HP printer. I presume they are injected by the CUPS spooler | 16:05.43 |
chrisl | kens: yes, I noticed a shed load of HP specific cruft in there..... | 16:06.05 |
kens | I htink that's what Jaws is complaining about | 16:06.17 |
ray_laptop | aren't the setpagedeive calls in stopped contexts ? | 16:06.39 |
kens | No | 16:06.46 |
| Well, some of them are | 16:06.57 |
tkamppeter | kens, chrisl, assuming we are under Precise (or Debian unstable) then PostScript is generated by the following command line: "gs -q -dNOPAUSE -dBATCH -dSAFER -sDEVICE=ps2write -sOUTPUTFILE=%stdout -dLanguageLevel=2 (or 3) -dDEVICEWIDTHPOINTS=XXX -dDEVICEHEIGHTPOINTS=YYY -c 'save pop' -f file.pdf | 16:07.11 |
henrys | other than that I really didn't have anything this week. | 16:07.40 |
| anybody else? | 16:07.44 |
chrisl | henrys: that bug you mentioned Alex could have bisected is a FAPI problem, and the backchannel shows the FAPI errors - I posted on the bug. | 16:07.46 |
kens | I don't believe that the DEVICEWIDTH and so on will have any effect on ps2write | 16:07.51 |
danilo_b | tor8: ok, i'll try. | 16:07.52 |
ray_laptop | most PPD's that I've seen have all of the 'feature' setting in stopped, but somebody may have tweaked the PPD and didn't know to do that | 16:08.00 |
tkamppeter | kend, chrisl, on Brother in addition "-dNoT3CCITT -dEncodeMonoImages=false" is added right before the "-c". | 16:08.06 |
alexcher | henrys: what bug? | 16:08.09 |
tkamppeter | kens ^^ | 16:08.12 |
kens | tkamppeter I believe this is an HP printer | 16:08.16 |
Robin_Watts | henrys: What's the company policy on getting outside people under NDA so they can see private attachments ? | 16:08.17 |
| zeniko (of SumatraPDF) would like access to our private attachments/tests_private. | 16:08.50 |
henrys | Robin_Watts:there really isn't one exactly - sort of a comfort level we have to reach. | 16:08.54 |
ray_laptop | policy ? we don't need no stinking policies ;-) | 16:09.00 |
tkamppeter | kens, chrisl, simply add "-dCompressPgaes=false -dCompressFonts=false" somewhere before "-c". | 16:09.17 |
ray_laptop | i.e. we have to like them ;-) | 16:09.19 |
henrys | i.e. I see that shelly has had a long standing relationship with kens and chrisl ... | 16:09.34 |
chrisl | tkamppeter: there's more happening than that GS command line, because there is a whole bunch of HP specific stuff in the job. | 16:10.15 |
kens | tkamppeter I can do that, but I would really like to see the file as produced under linux that way, if I make it here, it won't be the same. | 16:10.17 |
| chrisl tkamppeter if I remove the HP-soecific stuff, teh nit also works on Jaws. | 16:10.39 |
| SO it looks like another pritner-specific bug. | 16:10.50 |
marcosw_ | alexcher: any progress on http://bugs.ghostscript.com/show_bug.cgi?id=692886? The customer hasn't asked but I'd rather give them a solution before they do.. | 16:10.59 |
tkamppeter | |Frederik, can you try the thingy with cloning the queue which I have shown here earlier? | 16:11.01 |
henrys | chrisl:okay than why is it assigned alexcher? | 16:11.08 |
kens | ray_laptop : is there any problem with freeing memory which is beign GC tracked ? What will happen to the pointers if they are being tracked in other structures ? THey are set to NULL ? | 16:11.55 |
kens | has just discovered that many types of resource are not freed by pdfwrite, it relies on teh GC to free them. | 16:12.30 |
chrisl | henrys: I commented on bug, but I didn't want to take this one if alexcher wanted to look into it a bit more - there is something slightly odd about the matrix we end up using. | 16:12.53 |
alexcher | marcosw_: We've collected a lot od opinions and I agree now that the customer's request is quite reasonable. | 16:12.54 |
tkamppeter | kens, chrisl, the printer-specific PS code are the code snippets of the options in the PPD, inserted by the pstops CUPS filter. | 16:12.56 |
Robin_Watts | kens: If you free a block, you should be sure to set all pointers to it to be NULL, otherwise the enumeration stuff will think it's still valid, and it will step into it under GC builds and crash. | 16:12.57 |
ray_laptop | kens: having pointers to a structure you free is BAD | 16:13.03 |
alexcher | marcosw_: A fox will be committed soon. | 16:13.11 |
marcosw_ | alexcher: thanks for the update. | 16:13.21 |
kens | : What I mean is, if I free an object pointed at by one of my structures, what happens if somethgin else is poiting at it too ? | 16:13.33 |
Robin_Watts | kens: Boom! | 16:13.43 |
kens | OK so that's a problem. | 16:13.49 |
ray_laptop | kens: then the GC will crash | 16:13.53 |
kens | I don't think I can tell if anything else owns pointers to it as well. | 16:14.06 |
| THat will take a lot longer to check | 16:14.19 |
ray_laptop | kens: that's why we reference count things | 16:14.20 |
Robin_Watts | kens: Then that's a design flaw :( | 16:14.27 |
kens | ray_laptop : reference coutned how ? | 16:14.33 |
ray_laptop | kens: what structure is it ? | 16:14.39 |
kens | ray_laptop : many of the pdfwrite internal structures | 16:14.49 |
| But in this case, its the pdfont structure and the BaseFont string it looks at | 16:15.05 |
Robin_Watts | If everything that holds a pointer to a structure holds a reference to it, then it's easy; you drop the references, and when it hits zero, you free. | 16:15.14 |
tkamppeter | kens, chrisl, so I need to make a cups-filter 1.0.7 perhaps which has a special debug mode in which Ghostscript sends uncompressed PS? | 16:15.33 |
kens | Robin_Watts : yes I understand, but is that how the GS one is implemented ? :-) | 16:15.34 |
henrys | mvrhel_laptop:are you ready to start the ht before clist stuff or are you still buried in other stuff? | 16:15.38 |
kens | tkamppeter if you could do that it would certainly be useful. | 16:15.49 |
Robin_Watts | kens: Yes, I believe so. How else could it be done? >8*) | 16:15.57 |
kens | Robin_Watts : I've given up second-guessing the GS memory manager(s) | 16:16.13 |
ray_laptop | kens: to add ref counting add a 'rc' element and increment it every time someone sets a pointer to that struct, and decrement it when done (as you clear the pointer to it). the rc_decrement going to 0 will free it | 16:16.22 |
kens | ray_laptop : but take the case of the BaseFont string. | 16:16.44 |
| It 'looks' like we inherit it from the font. | 16:16.52 |
danilo_b | tor8: apparently it already works in the dev version. wasn't there yet in arch's 0.9 build :) | 16:16.55 |
kens | But sometiems we subset, and so we add a rpefix to it. | 16:17.01 |
henrys | alexcher:chrisl and I are talking about 692850 | 16:17.17 |
ray_laptop | kens: "inherit" ? | 16:17.19 |
kens | In that case we use gs_resize_string, which gets us a new string. | 16:17.19 |
Robin_Watts | Oh, something else for the meeting; while I'm away, I've told Radu Lazar to mail support@artifex.com or appear in here to ask for help. | 16:17.27 |
kens | ray_laptop : looks like we copy the pointer form the original font | 16:17.32 |
Robin_Watts | he's from the MuPDF customer, so if he appears, please don't just ignore him :) | 16:17.42 |
kens | But I'm unclear if this is precisely true | 16:17.46 |
| ray_laptop : Its one of those 'twisty maze of passages all alike problems | 16:18.00 |
| I see that gs_resize_string frees the original block of memory | 16:18.21 |
ray_laptop | kens: if you want something that you manage, then you need to copy it into your own struct (string) that you can then free | 16:18.23 |
kens | ray_laptop : Maybe this is waht is happening. | 16:18.37 |
henrys | Robin_Watts:okay, marcosw_? | 16:19.09 |
ray_laptop | well, if resize_string is freeing something that the BaseFont is still pointing to, that is a problem | 16:19.11 |
kens | Since the 'original' string is freed by resize_string I'm sort of guessing that this must be a copy of the original font name | 16:19.12 |
Robin_Watts | Oh, yes, another thing for the meeting... | 16:19.26 |
| We need to register with google for the android marketplace. | 16:19.39 |
kens | So OK. If the string is mine, and I free it, and set the struct pointer to NULL, I''m OK ? | 16:19.42 |
marcosw_ | henrys: okay. | 16:19.48 |
alexcher | henrys: re 692850, it's a rendering problem. I've identified the problem but don't yet know how to proceed. | 16:19.51 |
ray_laptop | kens: yes (in theory) | 16:19.56 |
Robin_Watts | Rather than me register, I was going to suggest we make an android@artifex.com address and then register with that ? | 16:19.58 |
| That can just forward to me (or to me and anyone else interested). | 16:20.10 |
henrys | alexcher:so it works with the artifex scaler and not freetype? | 16:20.22 |
ray_laptop | Robin_Watts: you can establish it as an alias for yourself | 16:20.25 |
Robin_Watts | ray_laptop: I'm not sure *I* can, can I ? | 16:20.54 |
alexcher | henrys: I need to re-check this. | 16:21.01 |
ray_laptop | or, we can make a 'list' (like support and tech) so several people see the email | 16:21.15 |
Robin_Watts | ray_laptop: I was thinking it was something that marcosw would need to do ? | 16:21.22 |
tkamppeter | kens, chrisl, only problem is that today I cannot get it uploaded to Debian, as the Debian maintainer is sick. | 16:21.23 |
henrys | Robin_Watts:I thought we agreed tor8 was going to manage the android post so he'll be the alias. | 16:21.45 |
ray_laptop | Robin_Watts: henry or I can do it -- not sure ablout marcos | 16:21.48 |
marcosw_ | Robin_Watts: I can do it. How should it forward to initially? | 16:21.56 |
Robin_Watts | henrys: fine, whoever. | 16:22.00 |
chrisl | alexcher: I commented on 692850 - I *think* the artifex scaler code is clamping the values as I mentioned in my comment | 16:22.05 |
ray_laptop | henrys: if we do it as a list, then it makes it easy | 16:22.16 |
Robin_Watts | marcosw_: Well, if we can do it as a list, then to tor and me? | 16:22.17 |
henrys | I would alias it to tor8 | 16:22.26 |
kens | tkamppeter we cna probably do yet another bisect of the file and try to see what it is the HP pritner doesn't like. It will take time though | 16:22.56 |
henrys | or anyone that wants to be on the list... | 16:22.58 |
marcosw_ | how is tor8 different than tor? | 16:23.23 |
kens | |Frederik : would you have time to run some tests on your printer over the next few days ? | 16:23.23 |
tor8 | marcosw_: not much, both mentions ping on my irc. | 16:24.06 |
paulgardiner | Robin_Watts, tor8: the last commit on master of http://git.ghostscript.com/?p=user/paulg/mupdf.git delays the progress dialog appearance, although fixing that just shows up the page rendering progress dialog doing the same annoying thing. I'll see if the same trick applies there too. | 16:24.21 |
tor8 | and someone else is using the 'tor' nick on freenode :( | 16:24.24 |
kens | ray_laptop : thanks, I'll give it a go tomorrow. I need to look at how named resources are freed and see if the fonts are freed there (spent a large part of the day gettgin that far). If they are not I will need to add code to free the fonts. Either way I will ad specific code to free the BaeFont string , and then see if anything goes bag :-) | 16:24.47 |
sebras | danilo_b: oh, I may have missed something in the reload-implementation. if your testing is successful let me know and I'll make sure we include a fix for this. | 16:25.01 |
henrys | tor8:you should go with "odin" or something like that. | 16:25.14 |
kens | Surely Thor ? Its nearly the same | 16:25.27 |
sebras | tor8: goodie, because I'm not convinced that link_dests is simple enough for clients as it is. | 16:25.47 |
danilo_b | sebras: apparently it works in current dev release. i had 0.9 installed. | 16:26.18 |
chrisl | marcosw: ping | 16:26.18 |
tor8 | henrys, kens: I used to go with ccxvii, but nobody know that was me... | 16:26.24 |
danilo_b | sebras: makes it a great addon for vim/latex :) | 16:26.34 |
tor8 | knew* | 16:26.34 |
henrys | so I'm going to start studying for the next meeting with paulgardiner for which I'm woefully unprepared. | 16:26.36 |
chrisl | tor8: have you got "tor" registered on freenode? | 16:27.10 |
mvrhel_laptop | oops sorry I missed your earlier question henrys | 16:27.16 |
sebras | danilo_b: excellent! then no patches are required. :) | 16:27.21 |
Robin_Watts | sebras: The intention of the link_dest stuff was that it should be simpler than PDFs stuff. | 16:27.43 |
mvrhel_laptop | I am trying to wrap up this tiffsep planar stuff | 16:27.44 |
henrys | mvrhel_laptop:yeah I was going to ping you later, but ... | 16:27.46 |
marcosw_ | tor8 and Robin_Watts: there is now an android@artifex.com email address that forwards to the both of you. I've sent a test email. | 16:27.49 |
mvrhel_laptop | I did have a question for everyone about this | 16:27.58 |
danilo_b | is it possible that the jpeg-8d build dependency is missing in the documentation / README? | 16:28.11 |
henrys | mvrhel_laptop:shoot | 16:28.21 |
ray_laptop | marcosw added the android@artifex.com with Robin_Watts and tor -- anybody else want on the list ? | 16:28.29 |
danilo_b | and is jpeg-8d a new library? i can't find any package for arch linux, neither official nor user-contributed in AUR, which is rarely the case for libraries :) | 16:28.53 |
mvrhel_laptop | so with the tiff planar solution, we are going to be currently limited to 32 spot colors due to the hard code of 32 pointers in gxgetbit.h line 44 | 16:28.57 |
henrys | marcosw:can we put paul on it also? | 16:29.05 |
sebras | Robin_Watts: yes, but I'm not so sure that it is... I guess what bugs me the most about it is that too many tests are required. I'll try to get something together and get back to you and tor8. | 16:29.07 |
Robin_Watts | Specifically the library fills out as many of the t/l/b/r things as it can. | 16:29.16 |
sebras | Robin_Watts: yes, I know. but still... | 16:29.34 |
Robin_Watts | So a 'dumb' client can just treat t/l/b/r as a bbox and it'll work most of the time. | 16:29.36 |
kens | mvrhel_laptop : 32 spots soudns like enough to me :-) | 16:29.44 |
mvrhel_laptop | Me too | 16:29.50 |
marcosw_ | sure, what's paul's email address? | 16:29.52 |
tor8 | Robin_Watts, sebras: we could support two levels of access -- a dumb one (page & rect) and a full level of access to all the PDF icky ones | 16:30.02 |
henrys | marcosw_:I'll send it to you so it isn't picked up here. | 16:30.27 |
chrisl | mvrhel_laptop: 32 spot colors, or 32 total plates? | 16:30.30 |
Robin_Watts | but feel free to come up with something better ; sumatraPDF have already made it plain they want more stuff in there. | 16:30.30 |
mvrhel_laptop | 32 total plates | 16:30.38 |
ray_laptop | mvrhel_laptop: it seems like it should be set to GS_CLIENT_COLOR_MAX_COMPONENTS] | 16:30.41 |
tor8 | or we could say we're happy enough with the dumb one and let sumatrapdf (and others) implement the full thing themselves? | 16:31.02 |
mvrhel_laptop | yes | 16:31.05 |
Robin_Watts | henrys: paulgardiner may want to use a different alias on his domain - don't know how he filters his mail... | 16:31.06 |
mvrhel_laptop | ray_laptop | 16:31.10 |
henrys | Robin_Watts:oh okay. | 16:31.28 |
ray_laptop | mvrhel_laptop: | 16:31.28 |
mvrhel_laptop | that was supposed to be | 16:31.37 |
henrys | Robin_Watts:btw what day do you leave? | 16:31.37 |
chrisl | mvrhel_laptop: I've seen a couple of files with 36 plates (dratted n-channel!!), but only a couple...... | 16:31.38 |
mvrhel_laptop | yes ray_laptop | 16:31.43 |
Robin_Watts | henrys: tomorrow pm. | 16:31.48 |
henrys | have a great time, it does sound like a great adventure. | 16:32.09 |
ray_laptop | the current default GS_CLIENT_COLOR_MAX_COMPONENTS is 14 | 16:32.21 |
sebras | danilo_b: you can find the latest version of libjpeg here: http://www.ijg.org/ | 16:32.23 |
mvrhel_laptop | if ok, I will set it to GS_CLIENT_COLOR_MAX_COMPONENTS then people can go ahead configure to their needs | 16:32.30 |
kens | Soudns good to me | 16:32.38 |
| Do you need to check it does not exceed 32 though ? | 16:32.51 |
chrisl | marcosw_: Do you still have access to the IBM partner AIX VM? | 16:32.51 |
mvrhel_laptop | and not worry about having to set things in more than 1 place when they run into chrisl's file | 16:32.57 |
| no, the 32 will be changed to GS_CLIENT_COLOR_MAX_COMPONENTS | 16:33.10 |
| more than 32 should be ok I believe | 16:33.29 |
kens | Ah! | 16:33.31 |
mvrhel_laptop | then | 16:33.35 |
marcosw_ | chrisl: I need to re-request access. It expires after a few days. | 16:33.39 |
kens | mvrhel_laptop : also sounds good | 16:33.47 |
mvrhel_laptop | ok thanks for the input | 16:33.55 |
| I am also updating psdcmyk to work like this | 16:34.07 |
| then I can run a cluster test | 16:34.13 |
| to check for issues | 16:34.16 |
Robin_Watts | henrys: Yes. I can cope with anything except dirty hotels. and cold hotels. and no wifi. and no power. | 16:34.17 |
| and no hot water. and no aircon. | 16:34.25 |
| Why the HELL am I going again? :) | 16:34.32 |
paulgardiner | marcosw_: artifex_android@glidos.net would be good thanks. | 16:34.33 |
mvrhel_laptop | Robin_Watts: have fun! | 16:34.45 |
| I imagine it is cold this time of year there | 16:34.58 |
Robin_Watts | mvrhel_laptop: Does that 32 include the CMYK ? | 16:35.00 |
chrisl | marcosw_: we have a build problem reported on AIX, and I can't fathom what could be causing the problem...... #692938 | 16:35.07 |
mvrhel_laptop | Robin_Watts: yes | 16:35.08 |
Robin_Watts | In Kathmandu it's 33 degrees (C). | 16:35.20 |
mvrhel_laptop | wow | 16:35.35 |
Robin_Watts | In Bhutan it's 13 or so during the day, 2 at night. | 16:35.39 |
| -2 at night, sorry. | 16:35.48 |
mvrhel_laptop | so you need to pack a bit of everything | 16:36.03 |
henrys | Robin_Watts washed away in glacier melt.... | 16:36.26 |
Robin_Watts | mvrhel_laptop: Yes, it's a problem for Helen. Hair Straightner/Hair Curler/Hair Volumiser... | 16:37.50 |
marcosw_ | chrisl: you can either assign it to me or I can request an AIX virtual machine and send you the login information (or we can get you an AIX developer account and associate it with Artifex's membership, but I think Miles has to get involved for that). | 16:37.56 |
mvrhel_laptop | hehe | 16:38.07 |
henrys | marcosw:I'll add paulgardiner address after we figure out how he wants it set up. | 16:38.11 |
danilo_b | sebras: ah, there's the problem. arch linux includes libjpeg-turbo by default. | 16:38.17 |
Robin_Watts | henrys: paulgardiner said; up about 20 lines :) | 16:38.38 |
marcosw_ | henrys: I believe paulgardiner responded with a preferred address ^^ | 16:38.45 |
mvrhel_laptop | Robin_Watts: you mean you don't need all of those things | 16:38.49 |
marcosw_ | which I've already added to the alias. | 16:38.59 |
Robin_Watts | mvrhel_laptop: Saved by male pattern baldness :) | 16:39.08 |
chrisl | marcosw_: if you don't mind doing the request and passing on the details - given that it's a build problem. | 16:39.19 |
Robin_Watts | So tor8: I'll register that address with the android market/pay the $25 etc. | 16:39.44 |
henrys | oh sorry indeed I missed it entirely. | 16:39.47 |
marcosw_ | will do, it takes a half day or so. | 16:39.53 |
chrisl | thanks | 16:40.03 |
ray_laptop | mvrhel_laptop: are you ripping out the compressed color encoding ? | 16:42.09 |
henrys | can we start the next meeting with paulgardiner? tor8, Robin_Watts? | 16:42.33 |
| et al | 16:42.41 |
Robin_Watts | I'm ready. | 16:42.46 |
paulgardiner | Me too | 16:42.59 |
marcosw_ | I have to run, my Tuesday afternoon meeting is reschedule for 10:00 this week. Robin_Watts have a good time, I'm envious :-) | 16:43.24 |
henrys | I still have gotten no word back from raph at google but I'm going to ping him again today. | 16:43.25 |
Robin_Watts | marcosw: Thanks. | 16:43.35 |
mvrhel_laptop | ray_laptop: no. there is a #if that one can currently set to use the planar tiffsep or the old style tiffsep the changes are not that great that I decided to keep both for now | 16:44.18 |
| there may be cases where we want the compressed encoding | 16:44.48 |
| printing a swatch book with 100s of colors comes to mind | 16:45.19 |
henrys | tor8:did you have a chance to read paulgardiner's status? | 16:45.44 |
tkamppeter | kens, chrisl, |Frederik, I will release cups-filters 1.0.7 upstream today with the new debug mode and hope that it gets into Debian unstable tomorrow. | 16:46.46 |
kens | OK thanks Tuill | 16:46.55 |
ray_laptop | mvrhel_laptop: I see -- of course that would require a re-compile after changing GS_CLIENT_COLOR_MAX_COMPONENTS as well as the #if in tiffsep | 16:46.57 |
mvrhel_laptop | yes | 16:47.06 |
ray_laptop | mvrhel_laptop: are you changing the MAX to 32 ? | 16:47.32 |
mvrhel_laptop | no, I was not planning on that | 16:47.43 |
chrisl | kens: did you get the problem job fully decompressed? | 16:47.49 |
henrys | paulgardiner:interesting chrome didn't use the native widgets. I wonder if it is planned as a future enhancement. | 16:48.12 |
ray_laptop | mvrhel_laptop: didn't one of the files from Gemma have 17 or 18 colors ? | 16:48.14 |
kens | chrisl yes | 16:48.28 |
mvrhel_laptop | ray_laptop: not sure. she can recompile with whatever she needs I figure | 16:48.48 |
paulgardiner | henrys: I'm not sure it would be an enhancement. The inline style of filling in text widgets looks quite impressive. | 16:49.20 |
ray_laptop | mvrhel_laptop: They are one of the companies that keeps requesting binaries. We probably need to push back on that | 16:49.23 |
chrisl | mvrhel_laptop: they don't compile their own GS........ | 16:49.23 |
mvrhel_laptop | ray_laptop: right now I am making the changes to psdcmyk so that I can see what is broken from the cluster push | 16:49.26 |
| ray_laptop and chrisl: ick | 16:49.35 |
| they should do it themselves | 16:49.48 |
chrisl | kens: okay good. I hope the HP error handler doesn't erase the current page, or we'll be struggling to diagnose this...... | 16:49.49 |
henrys | paulgardiner:yes but presumably you can't get the look and feel of each platform right? Or am I missing something? | 16:50.07 |
chrisl | mvrhel_laptop: I'll refer them to you to deal with next time, then ;-) | 16:50.09 |
ray_laptop | mvrhel_laptop: I think this time we will require them to do it. | 16:50.11 |
paulgardiner | henrys: and I'm not aware of any reduced functionality. You can still select copy and paste | 16:50.13 |
mvrhel_laptop | hehe | 16:50.16 |
| I am hoping with this fix we wont hear from them for a while | 16:50.41 |
ray_laptop | marcos can tell them his boss won't let him provide binaries any more ;-) | 16:50.43 |
paulgardiner | henrys: yes that's true. | 16:50.55 |
ray_laptop | henrys can issue a management directive :-) | 16:51.01 |
mvrhel_laptop | they are located in England. Maybe chrisl should visit them.... | 16:51.12 |
paulgardiner | henrys: but having chosen PDF for forms, it might be desirable for the text to always look like part of the document. | 16:51.30 |
chrisl | mvrhel_laptop: only if you send me a genuine US baseball bat to take with me! | 16:51.41 |
mvrhel_laptop | hehe | 16:51.46 |
henrys | paulgardiner:yeah I think you are right about that. | 16:51.52 |
Robin_Watts | henrys: Native widgets gives you the exact look and feel for a given OS. Inline editing gives the exact styles/fonts/layout. It can also come close to the feel (if not the look) as the selection/editing behaviour can be matched well. | 16:51.54 |
ray_laptop | paulgardiner: henrys: it's a lot of work to try and get inline text widgets to match each platform | 16:52.14 |
paulgardiner | henrys: I haven't established beyond all doubt that native widgets aren't used. It just seems unlikely. | 16:52.16 |
ray_laptop | paulgardiner: you are basing that on the AA appearance as you type ? | 16:53.03 |
paulgardiner | ray_laptop: that's the bit I don't understand: if you are using PDF for forms, I'd have thought you wanted the PDF to determine the look and feel | 16:53.05 |
Robin_Watts | As an integrator coming to a project, what I'd like to hear is "You can use Native Widgets - that's easy. Or you can do some more work and get inline editing." | 16:53.05 |
| and that's exactly what paulgardiner is proposing, I think. | 16:53.24 |
paulgardiner | ray_laptop: the way the antialiasing changes on the caret as it moves was what I considered the main evidence that native widgets weren't involved. | 16:54.04 |
ray_laptop | what about Acrobat (the other gorilla) ? have you looked at what they do ? | 16:54.26 |
henrys | looks at the code and falls to the ground - java - of course I should have know that but the reality of it all is sinking in. | 16:54.57 |
ray_laptop | henrys: you're looking at chrome ? | 16:55.39 |
henrys | I'm looking at paulgardiner's code. | 16:55.57 |
ray_laptop | which code ? | 16:56.07 |
paulgardiner | Supporting native widgets looks to be the easier option for our implementation, but I can imagine that inline might be desired by some customers | 16:56.16 |
henrys | ray_laptop:the link in his status report. | 16:56.35 |
ray_laptop | henrys: thanks -- I hadn't loaded that yet. | 16:57.07 |
tor8 | henrys, paulgardiner: sorry, was passed out on the sofa from exhaustion. yes, I read the status email. | 16:57.08 |
Robin_Watts | henrys: I see no java. | 16:57.23 |
| http://git.ghostscript.com/?p=user/paulg/mupdf.git;a=commitdiff;h=03ecbc9d28da66f1267f45619d9e12c855c5116c | 16:57.28 |
| That's what we're looking at, right ? | 16:57.35 |
tor8 | my main fear with going non-native text widgets is not emulating the keybindings and copy&paste, it's the input method editors | 16:57.56 |
| the former is mainly a lot of work and testing, the latter is a lot more difficult | 16:58.15 |
henrys | upon more thought it does seem native widgets would not "with the pdf philosophy" in some cases. | 16:58.26 |
Robin_Watts | "input method editors" ? | 16:58.49 |
henrys | Robin_Watts:I am looking at MuPDFActivity.java | 16:59.12 |
tor8 | Robin_Watts: japanese text input | 16:59.14 |
| Robin_Watts: or even as simple as the yellow highlighting you get with macosx dead keys | 16:59.47 |
Robin_Watts | henrys: Oh, right. That's the android app, rather than the forms work. | 17:00.05 |
paulgardiner | I have no great desire to implement the inline mode. It looks difficult. I was just thinking it was best to think through it at the API level a bit, so we can keep the option opne | 17:00.16 |
Robin_Watts | tor8: I see. | 17:00.46 |
paulgardiner | Robin_Watts, tor8: I've just pushed another change to the android app, so that the page rendering progress dialogs are delayed. | 17:00.58 |
henrys | Robin_Watts:oh okay I am confused I followed that link and assumed there was some form stuff woring on droid. sorry. | 17:01.02 |
Robin_Watts | But that's the point. We offer 'native widgets' as the 'fast' route, and offer 'inline editing' as the long and winding road where the complexities are on the implementer. | 17:01.37 |
| henrys: No, at the moment, I believe paulgardiners work has been on research and sketching out the API. | 17:02.03 |
| I read through the email earlier and had various comments on it. I phoned paul, and I think they boiled down to 2. | 17:02.36 |
tor8 | paulgardiner: looks better, but the search highlights don't consistently appear now :( | 17:03.09 |
ray_laptop | we should probably have a notice that 'the support for inline (other than native) widgets is sort of allowed for, but has never been implemented, so may not be complete' | 17:03.12 |
tor8 | paulgardiner: sometimes it loads the new page with the results, but no blue highlights | 17:03.29 |
Robin_Watts | 1) I think there *should* be a callback (but it's a callback that just says 'you need to redraw the document at some point'). The app can catch that and then use that to know when to ask for rectangles to redraw. | 17:04.01 |
paulgardiner | tor8: on the tablet? | 17:04.02 |
tor8 | paulgardiner: seems related to whether the progress display pops up or not | 17:04.03 |
| on the tablet, yes | 17:04.06 |
ray_laptop | chrisl: are you available for a private chat about cust 532 (I don't want to interleave the forms API and other discussion) | 17:04.23 |
henrys | Robin_Watts, paulgardiner:I'm trying to follow the link in the email to the "partial version of the forms api" | 17:04.24 |
tor8 | yeah, if I get the progress dial it doesn't highlight the search results | 17:04.28 |
Robin_Watts | henrys: Just use the link I pasted above. | 17:04.36 |
henrys | okay | 17:04.53 |
chrisl | ray_laptop: sure - IRC or phone? | 17:04.58 |
ray_laptop | chrisl: IRC. look for the tab | 17:06.01 |
Robin_Watts | 2) In Pauls email he talks about 'Factor 3' - problems to do with the fact that there are multiple threads flying around the place. | 17:06.11 |
paulgardiner | henrys: yeah, sorry that link wasn't very useful. | 17:06.14 |
henrys | ray_laptop:I think we are close to finished and miles is going to drag me into a 532 meeting today so I'd like to be up to date. | 17:06.14 |
Robin_Watts | the specific example he gives of a problem, is not, I believe a problem. | 17:06.36 |
ray_laptop | henrys: OK, I can wait | 17:06.48 |
Robin_Watts | I don't think there should ever be confusion at the app level about whether we are dragging to select, or dragging to move the page. | 17:07.45 |
tor8 | Robin_Watts: re native vs inline text editing -- my main opposition is from concern that in keeping options open for the latter, we'll make the project bigger in scope and add a lot of work for something that realistically speaking probably won't ever happen. | 17:08.01 |
Robin_Watts | (In acrobat, you change the pointer tool etc). | 17:08.03 |
tor8 | though there is a point in that mupdf text rendering differs quite a *lot* from windows native text rendering | 17:08.19 |
danilo_b | sebras: is there a way for mupdf to make it switch to a specific page from the outside? | 17:08.41 |
tor8 | danilo_b: not unless you fancy sending fake keypress events | 17:09.01 |
ray_laptop | tor8: I would expect that once a widget loses focus we would use fz to re-render correctly | 17:09.03 |
danilo_b | tor8: ok, thanks. | 17:09.14 |
tor8 | ray_laptop: yes, which will look a bit jarring when all of a sudden it loses the color fringing and becomes blurry :) | 17:09.34 |
Robin_Watts | tor8: If we were faced with a decision between simple and small, or extensible and huge, I'd agree. But I don't think the overhead is massive (or even necessarily anything - we may just spec the entry points and not implement them to start with) | 17:09.41 |
ray_laptop | tor8: I think I've seen that happen with Acrobat form filling (but I wasn't looking for it) | 17:10.36 |
tor8 | Robin_Watts: yes, I think we (or paul, rather) will find out sooner or later how much work is involved. I just want to make it clear that if it turns out to be a lot of extra work, he should not hesitate to reconsider the decision. | 17:10.42 |
| ray_laptop: yeah. it happens with apple preview form filling a well. | 17:11.25 |
paulgardiner | Robin_Watts: I totally agree. The intention was no more than to keep the inline option in the back of my mind, and keep it open while it doesn't cost anything. | 17:12.12 |
| That was more aimed at Tor | 17:13.21 |
henrys | tor8:can you give the api a careful read (if you haven't already), I imagine Robin_Watts has read it already and let's plan out next week, Robin_Watts won't be here. | 17:13.25 |
| I don't think there is more to "meeting" about is there? | 17:13.52 |
| can we do same time next week? | 17:14.20 |
tor8 | paulgardiner: did you consider tabbing to jump between form widgets? | 17:14.23 |
| or is that handled behind the scenes by the library, and returned in fz_get_focused_widget? | 17:15.00 |
paulgardiner | tor8: I did a little. It might be done by passing through arrow keys. | 17:15.07 |
| And then rerequest the focussed widget | 17:15.33 |
| henrys: my plan now - if it makes sense to everyone else of course - is to try to get some small part of it working. e.g., check boxes. | 17:16.42 |
Robin_Watts | paulgardiner: To do 'tab' or 'shift-tab' to move forwards/backwards between widgets, presumably the app would have to 'set' the current value of the native widget into the library, then say "focus next widget" (or "focus previous widget") ? | 17:17.09 |
henrys | paulgardiner:I would be much more comfortable with a small prototype, and if Miles asks me where the money is going I can show it to him ;-) | 17:17.30 |
paulgardiner | Robin_Watts: Ah yes, you would have to write back the contents of the native widget before moving. Hadn't thought of that. | 17:18.07 |
tor8 | this all reminds me we really need to rewrite the x11 and win32 viewer app... | 17:18.12 |
| unless paul wants to prototype it on the android app (which is needlessly painful, IMO) | 17:18.56 |
paulgardiner | henrys: yes exactly. | 17:19.11 |
Robin_Watts | I was thinking that the android app would be the place to prototype it :) | 17:19.24 |
paulgardiner | tor8: I wondered about trying to knock up something quickly with the MFC under windows | 17:19.45 |
henrys | I'd vote for windows. | 17:19.54 |
Robin_Watts | It's not me that's coding it, so I say go with whatever you are happiest with :) | 17:20.14 |
paulgardiner | but it might be possible to work with the existing win32 viewer | 17:20.20 |
tor8 | Robin_Watts: paulgardiner: cooking up a minimal mupdf viewer on win32 with no bells and whistles that will let you use native win32 controls for the form prototype shouldn't be too much work | 17:20.26 |
| the existing win32 viewer is a bit of a nasty hack | 17:20.35 |
| it's just been organically grown together with the x11 viewer in an unholy hybrid mess | 17:21.02 |
Robin_Watts | The existing win32/x11 viewers are both (AIUI) wrappers around pdfapp.c | 17:21.12 |
henrys | I really do think windows would be the most useful though. | 17:21.27 |
paulgardiner | tor8: sounds sensible to me. Had occurred to me, if for no other reason than I don't really know non-MFC windows programming very well. | 17:21.28 |
tor8 | pdfapp.c is the ugliest part of it all though :) | 17:21.32 |
Robin_Watts | Can you mix MFC with raw win32 stuff ? | 17:22.04 |
paulgardiner | Yep | 17:22.12 |
| MFC is just a wrapper | 17:22.20 |
tor8 | paulgardiner: the win_main.c app uses raw win32 controls. I think if you dig through the git history you can also find the old mozilla plugin that uses scrollbars etc. | 17:22.21 |
| isn't MFC a bunch of version-dll-hell-dependent C++? | 17:22.45 |
Robin_Watts | right, but I thought MFC was an 'all or nothing' thing? Everything had to be in MFC, or nothing could be. | 17:22.50 |
paulgardiner | tor8: but you can statically link | 17:23.07 |
tor8 | paulgardiner: creating native win32 controls and text fields is actually not that hard | 17:23.22 |
henrys | what about hacking something into sumatrapdf? | 17:23.43 |
paulgardiner | Debugging support is good on Windows too. The difficulty of debugging throught the C is what puts me off Androd. | 17:24.06 |
Robin_Watts | I think for the proof of concept, doing anything that works is fine. If Paul is happiest with MFC, then go for it (just don't drag C++ into mupdf :) ) | 17:24.28 |
tor8 | I'd also be happier if I never have to touch C++ :) | 17:24.32 |
paulgardiner | :-) | 17:24.45 |
tor8 | so I'll update the iOS app, wait a few days for sebras to bash on the link dest stuff, then tag an RC1 of mupdf and while that one is waiting I think I could start on cleaning up the viewer apps for x11 and win32 | 17:26.39 |
tkamppeter | kens, chrisl, |Frederik, cups-filters 1.0.7 released. Takes "lpr -o psdebug ..." to produce uncompressed PS output. | 17:26.58 |
Robin_Watts | tor8: Anything obvious I've missed in http://ghostscript.com/~robin/BLURB | 17:27.14 |
chrisl | tkamppeter: that's great, thank you | 17:27.19 |
mvrhel_laptop | ok psdcmyk is working with planar | 17:27.37 |
| time to cause an explosion on the cluster | 17:27.45 |
tkamppeter | kens, chrisl, |Frederik, only upstream for now, Ubuntu Precise follows in a few minutes, Debian unstable I hope tomorrow. | 17:27.47 |
henrys | so doing it in sumatra would be too much learning and startup? With that we could really have a demo? | 17:27.57 |
tor8 | henrys: I think hacking sumatrapdf will have a high entry cost, it's a lot of C++ to grok | 17:28.03 |
Robin_Watts | psdcmyk ? I thought you were working on tifsep ? | 17:28.06 |
henrys | tor8:okay | 17:28.17 |
tor8 | they have their own wrapper classes around everything, so we'd need to dig through those to get at mupdf, and learn how their ui stuff is set up | 17:28.59 |
Robin_Watts | Damn google payments. | 17:29.12 |
tor8 | ? | 17:29.22 |
Robin_Watts | henrys: SumatraPDF can wrap something other than mupdf too. | 17:29.24 |
| It's tied the developer account to my google account. | 17:29.35 |
henrys | okay paulgardiner are you good with all this? | 17:29.35 |
paulgardiner | tor8: annoyingly, the missing search highlighting is definitely not reproducable on my S2 :-( | 17:29.46 |
tor8 | paulgardiner: how about link highlights? | 17:30.15 |
paulgardiner | It's likely to be a missing "invalidate". Probably an existing bug that we were getting away with | 17:30.19 |
tor8 | on my tablet, if I hit the link button, the current page never changes its appearance but if I swap pages with the link button in the highlight mode new pages that scroll in have the highlights | 17:30.50 |
| it could be related | 17:30.53 |
henrys | I'd like to move on to the ray_laptop and chrisl meeting if paulgardiner has enough to go on until next week. | 17:31.10 |
paulgardiner | tor8: same there. Less surprisingly in that case, because I remember thinking "Oo, that worked without having to add an invalidate call. I wonder if it'll work on the tablet?" | 17:31.18 |
| henrys: Yes thanks. | 17:31.32 |
| tor8: I have an idea what it might be. Hang on a min | 17:32.28 |
tor8 | Robin_Watts: BLURB looks fine, but we should probably make another one for the RC | 17:32.35 |
henrys | ray_laptop, chrisl:so what's up with 532? I am really very concerned that Miles is going to send a letter we don't want to be sent now and I'm trying to hold him back. That's where I am with 532. I think the letter places too many demands on our most significant customer. | 17:35.02 |
Robin_Watts | tor8: Updated BLURB. | 17:35.18 |
chrisl | henrys: erm, I thought you wanted to push back on them more? | 17:35.43 |
tor8 | Robin_Watts: ta. I'll publish that before the end of the week, soon as the iOS app is updated. | 17:36.31 |
henrys | ray_laptop, chrisl:I want to say reproduce it on the 532 simulator - he wants ghostscript unless they pay extra - in between - in no case do I want a letter sent to management. | 17:36.48 |
paulgardiner | tor8: I've just pushed another change. It's random swipe at the code, but it might do the trick. | 17:37.03 |
henrys | ray_laptop, chrisl:but I'll likely loose this one. | 17:37.20 |
ray_laptop | henrys: I agree that Miles should probably stay out of it. | 17:37.54 |
henrys | I've lost 2x you guys and miles really I'm really sucking air. | 17:38.08 |
ray_laptop | henrys: I haven't seen a recent 'draft' from Miles. | 17:38.54 |
chrisl | henrys: well, for what my opinion counts, I'm kind of in the middle - I'd like them to pay extra if they're going to work up to using all our PDLs, but cut them a *good* deal on them.... but that's another issue, really. | 17:38.55 |
tor8 | paulgardiner: sorry, doesn't work :( | 17:38.55 |
paulgardiner | :-( | 17:39.04 |
| Actually, no that probably shouldn't have worked. I have another idea. | 17:39.59 |
ray_laptop | henrys: I don't think there's any point in pushing the rest of our PDL's on them. More headache for us and little or no incremental revenue | 17:40.28 |
henrys | well let's figure out an engineering policy that works for all of us and I'll promote that to Miles. I don't think we are even arguing about prices anymore. | 17:40.30 |
tor8 | paulgardiner: if I land on a page with missing highlights, hide and show the buttons, then they appear | 17:41.11 |
henrys | ray_laptop:I am very concerned politically about doing that. Seems like a great way to alienate the folks that like us, well like us better than some other parts of the company. | 17:41.23 |
ray_laptop | henrys: I thought our policy is that we just advise on how to debug their target (stuff that doesn't show up on the simulator) | 17:41.30 |
paulgardiner | tor8: that's weird. | 17:42.06 |
ray_laptop | henrys: if we want extra revenue for having me on site helping debug on their printer, fine | 17:42.12 |
henrys | ray_laptop:geez chrisl has now 3 customer bugs and no progress on his most important project it's too much time. | 17:42.14 |
chrisl | ray_laptop: that's a difficult one - this current thread is really just "advising" them on debugging their target..... | 17:42.47 |
ray_laptop | henrys: having chris involved in this target crash doesn't make sense | 17:42.49 |
| IMHO chrisl should only be on the font bugs that can be seen on the simulator | 17:43.19 |
chrisl | ray_laptop: this started life as a font bug, on the simulator | 17:44.01 |
ray_laptop | and we (you if you wish) can tell Len that chrisl is only available for this kind of thing | 17:44.20 |
| chrisl: this started as a font bug ? | 17:44.30 |
tor8 | paulgardiner: dinner calls, I'll be back later | 17:44.38 |
ray_laptop | chrisl: I thought the original dumps showed it was in the transparency code | 17:45.03 |
chrisl | ray_laptop: the apparent cause of the crash is my fix for filling a type 3 glyph with a transparent pattern | 17:45.07 |
ray_laptop | you mean that without your 'fix' it doesn't crash ? | 17:45.31 |
henrys | I was extremely happy that Len went off and worked on it, I was not expecting that at all I thought all hell would break loose. | 17:45.40 |
paulgardiner | tor8: I'll have to go now too. Might have to stick with the flivkering dialog box for now. :-( | 17:45.54 |
chrisl | ray_laptop: yes, exactly - although without my fix, the glyph doesn't render *at all* | 17:46.04 |
henrys | I'm sorry that should say "I was expecting that all hell would break loose" | 17:46.22 |
ray_laptop | chrisl: is the glyph being painted as a bitmap or as a path ? | 17:47.03 |
henrys | speak of the devil incoming mail | 17:47.05 |
chrisl | ray_laptop: it's a bitmap - it's being rendered from the cache. | 17:47.27 |
| ray_laptop: see http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=f63237e1 | 17:48.10 |
| same change applied to our code..... | 17:48.20 |
ray_laptop | chrisl: thanks. | 17:48.32 |
tkamppeter | ken, chrisl, |Frederik, cups-filters is now uploaded to Ubuntu Precise. You can find it on https://launchpad.net/ubuntu/+source/cups-filters. Perhaps it also works on Debian, at least you could download the source package and rebuild it on your Debian system. | 17:48.50 |
ray_laptop | it looks like Len is "owning" this, so we can just wait to see what he finds. | 17:48.57 |
| henrys: did you open a private chat for me (or did I hit a key) ? | 17:49.28 |
henrys | no I didn't | 17:49.40 |
chrisl | ray_laptop: so it passes our cluster tests, and works fine on my G4 Mac mini. The only other thing I was going to try was to valgrind it. | 17:49.42 |
tkamppeter | kens ^^ | 17:49.47 |
ray_laptop | chrisl: I can give you mods that let you build a standalone gs of their code (I did it for the color product, so it won't take long). That way you can valgrind their code base | 17:51.03 |
chrisl | ray_laptop: will that work on Linux? All the configure stuff has been removed in their source base :-( | 17:51.42 |
ray_laptop | chrisl: I was testing _041 but didn't have your fix | 17:51.47 |
| henrys: do you want chrisl to spend the time on that ? If so I'll get their code base building on linux | 17:52.27 |
henrys | chrisl's call. | 17:53.11 |
ray_laptop | henrys: I was asking you because you were concerned about the time is was taking from chrisl | 17:53.36 |
henrys | I wasn't clear that a specific patch had introduced the problem. | 17:54.11 |
| sort of changes my perspective on the issue. | 17:54.24 |
chrisl | henrys, ray_laptop: whether I used it for this problem or not, I think being able to valgrind (most of) their GS code base could well be useful... | 17:54.36 |
paulgardiner | tor8: for when you get back, I've just pushed another possible fix. This one really should work, but I bet it wont. | 17:54.49 |
chrisl | ray_laptop: the end result of that fix is that we go through the general imagemask code for these cases (glyph, filled with a transparent pattern). One of the reasons I opted to do it this way was because we'd both commented we'd seen cases going through there before, so I figured it was safe! | 17:55.15 |
ray_laptop | chrisl: OK. I'll ask Len for an update of their sim code (to make sure I match his patches) and then get it building on peeves. | 17:55.34 |
| oh oh. I think my laptop is about to crash .... | 17:56.55 |
| it's been doing it and it happens after a window hangs inexplicably | 17:57.19 |
chrisl | ray_laptop: cool, okay. FWIW, there is a very simple workaround - we can render Type 3 glyphs uncached - doesn't help if another font type goes through that code, though..... | 17:57.21 |
Robin_Watts | paulgardiner: Let me see if I can reproduce the problem here on the prime. | 17:58.18 |
chrisl | henrys: on the customer bugs I have right now: I'm currently working on upgrading our Freetype version, which fixes one customer problem. I've held off the tiffsep changes until mvrhel has finished his overhaul of that code. And the third bug also needs the Freetype update to even start on it. | 17:59.41 |
Robin_Watts | alexcher: How many changes have we got to our openjpeg code? (What version of openjpeg are we running?) | 18:00.13 |
chrisl | Robin_Watts: seems to 1.4.0 "plus"..... | 18:03.38 |
| s/to/to be | 18:03.47 |
Robin_Watts | Right. I was wondering how many changes were in the 'plus' ? | 18:04.00 |
| i.e. if we move to 1.5.0, how many of those go away ? | 18:04.14 |
| At least 1 potentially. | 18:04.27 |
alexcher | Robin_Watts: Not much. All changes are in the git history. | 18:05.49 |
mvrhel_laptop | bbiab | 18:06.23 |
alexcher | Robin_Watts: Most effort took the interface module, which is our code. | 18:07.13 |
Robin_Watts | alexcher: Right. I spotted the other day that our openjpeg had an extra arg to opj_decode, which if we move to 1.5 we can remove. | 18:08.09 |
alexcher | Robin_Watts: We are running v. 1.4.0 | 18:08.16 |
| Robin_Watts: So who of us will upgrade openjpeg ? | 18:10.06 |
kens | I have to be off, goodnight everyone | 18:10.39 |
Robin_Watts | alexcher: Not me (at least not immediately) - I haven't got time before I go. | 18:10.52 |
alexcher | Robin_Watts: OK, I'll take a look. | 18:11.26 |
tkamppeter | chrisl, kens, |Frederik, cups-filters 1.0.7 is now up on the Launchpad page I linked earlier. | 18:12.41 |
chrisl | tkamppeter: thanks. | 18:13.31 |
| |Frederik: I know kens asked earlier, but if I post some links to test files tomorrow (and possibly beyond), can you try them on your printer? Is IRC okay, or would you prefer e-mail? | 18:14.56 |
henrys | chrisl:I'm sorry I didn't understand the problem was not reproducible without your patch, that casts a new light on things. | 18:22.37 |
chrisl | henrys: np, it's all a bit convoluted. Truth is, I'm fairly sure that with the *right file* it would be reproducible without my patch, otherwise I'd give them an alternative solution to the original problem - just bad luck it's worked out this way :-( | 18:24.14 |
ray_laptop | chrisl: I was right. My machine crashed -- it takes a long time to restart when it goes down 'dirty' | 18:25.35 |
| chrisl: can you send the simple fix for Len to try on the target. Even if it doesn't fix every case, it may be good enough (good enough for them) | 18:26.27 |
| chrisl: if you want to step back from it, send the fix to me and I'll send it to Len | 18:26.56 |
| that way I can do the testing | 18:27.06 |
chrisl | ray_laptop: no (more) testing needed - it is a *very* simple change - but it means losing performance on Type 3 fonts..... | 18:28.03 |
ray_laptop | chrisl: I wouldn't even mention that to them ;-) | 18:30.06 |
chrisl | ray_laptop: as I said above, I am also concerned that with the right (wrong?) test job, we can hit the same problem in a non-Type 3 font..... | 18:31.58 |
ray_laptop | chrisl: since the problem is happening during rendering, I don't see how the glyph cache is relevant. That's why I want them to test it on the target | 18:33.04 |
| how the heck do I tell git to allow me to commit a makefile that has tab indents ? | 18:33.41 |
chrisl | ray_laptop: commit -n I think..... | 18:34.05 |
ray_laptop | chrisl: thanks. I'm looking at the help. I was trying TortoiseGit (gui tool) and didn't see the option | 18:34.46 |
chrisl | ray_laptop: the workaround I thought of was to change PDF's d1 operator to behave like d0, so PDF T3 fonts use setcharwidth instead of setcachedevice - if that seems okay to you, I'll send it to Len when I get back from playing squash (where I'm going momentarily) | 18:36.13 |
ray_laptop | chrisl: that's fine. Thanks. (and -n worked) | 18:37.11 |
chrisl | ray_laptop: okay, cool - on both counts. I'll send that over to Len in a couple of hours, then. | 18:37.40 |
ray_laptop | henrys: I finally remembered to commit the /PROFILE fix | 18:37.49 |
Robin_Watts | tor8: FWIW: I just applied pauls fixes and ran it on the transformer prime. | 18:38.21 |
ray_laptop | having a machine that crashes so often is no fun -- but not bad enough yet that I want to re-install the entire OS | 18:38.29 |
Robin_Watts | I've had at least 1 failure (without his latest fix) and adding that I've had no failures (and 20-30 successful runs) | 18:38.52 |
| (seraching for 'cascading' in pdf_reference17.pdf is a good example) | 18:39.18 |
ray_laptop | call me on the phone if anyone wants me back online. | 18:40.25 |
tor8 | paulgraham: that version works! | 19:48.05 |
henrys | who is paulgraham? do you mean paulgradiner? | 20:01.27 |
| oops paulgardiner | 20:01.46 |
|Frederik | chrisl: I can try, yes. I will try to follow here on IRC, buty maybe it's safer to use e-mail so I won't miss it. | 20:06.27 |
| chrisl: fhimpe@vub.ac.be | 20:06.35 |
tkamppeter | |Frederik, chrisl, please CC me on the e-mails: till dot kamppeter at gmail dot com. | 21:12.43 |
chrisl | tkamppeter: you just missed the first one - I'll cc you from now on..... | 21:13.48 |
|Frederik | chrisl_away, tkamppeter: thank you for your help in debugging this problem. Much appreciated! I will take a look at it tomorrow when I have access to the printer | 21:17.33 |
tkamppeter | |Frederik, can you paste the error message again which you got on the initial error page when you started reporting here? Thanks. Or mail it to me. In Pastebin it seems that it has already expired. | 21:33.28 |
tor8 | henrys: ahem. yes. paulgardiner indeed. | 21:59.16 |
| Forward 1 day (to 2012/03/21)>>> | |