| | <<<Back 1 day (to 2012/02/16) | 2012/02/17 |
| Robin_Watts | henrys: I can (and have) built Pauls code. | 00:09.00 |
| | tor8: So, can you cut/paste text from Acrobat Reader? | 00:09.38 |
| | IF we can't, then it's a no brainer to tell them it's a broken file. | 00:09.49 |
| | oh, you've replied already. Fair enough :) | 00:11.13 |
| sebras | tor8 Robin_Watts: some more text on sebras/doc and a fix or two on sebras/master. | 00:53.53 |
| | nn. | 00:53.56 |
| Robin_Watts | nn | 01:00.39 |
| | tor8: The ios project; does that interpret to display list then render the display list? Or does it just interpret to render. | 09:59.12 |
| tor8 | interpret to render, to save memory | 09:59.25 |
| Robin_Watts | Random idea; can we have a 'tee' device in mupdf? | 10:02.05 |
| | So we can interpret to render and make a display list at the same time? | 10:02.20 |
| tor8 | yeah, that's probably a good idea | 10:02.48 |
| | did you manage to get sebras mt test running with the freetype locking changes? | 10:03.25 |
| Robin_Watts | I haven't had a chance. I've been fighting gs fires. | 10:04.02 |
| sebras | Robin_Watts: when would you want both a pixmap _and_ a display list for a given page? | 10:36.21 |
| chrisl | sebras: an example would be to get a quick first view of a page (by direct rendering), but then have (more) rapid zooming by running the display list, rather than reinterpreting the page from the PDF. | 10:40.49 |
| sebras | chrisl: of course, I see. | 10:41.41 |
| chrisl | It's an idea I toyed with in my previous job, but never had time to pursue it | 10:42.20 |
| Robin_Watts | New Mac OS X due for the summer. Let's hope they can make it better than Lion. | 10:55.08 |
| tor8 | Robin_Watts: with promotion like "See how innovations from iPad inspire new features for the Mac." I highly doubt it | 11:08.31 |
| Robin_Watts | tor8: Yeah. | 11:08.42 |
| | If I'd wanted an ipad, I'd have bought an ipad! | 11:08.55 |
| | I have the customers modified source for mupdf-0.7. | 11:09.18 |
| tor8 | Robin_Watts: also, note the default option in http://images.apple.com/macosx/mountain-lion/images/security_settings.jpg | 11:09.22 |
| Robin_Watts | So... all of those settings are wrong :) | 11:09.53 |
| chrisl | Robin_Watts: can you tell me the set of commits I need to add to the 9.05 release in order to do the builds for Gemma, please? | 11:15.14 |
| Robin_Watts | chrisl: I'd just build her a snapshot. | 11:15.34 |
| | but if you want a list, I'll make one. | 11:15.53 |
| chrisl | I can't build her a snapshot because that will include all the GPL code, I either need to apply patches to the 9.05 commercial release, or create a new commercial release from a snapshot | 11:16.46 |
| Robin_Watts | d00354a,21579b0,43b14b2,d375ea1 | 11:17.35 |
| | Those are the only ones I know of. | 11:18.12 |
| chrisl | Cheers - I've got some e-mails I need to catch up on, then I'll do the builds for her | 11:18.26 |
| Robin_Watts | Thanks. | 11:18.31 |
| chrisl | Robin_Watts: we still expect the rangecheck in showpage, yes? | 12:29.17 |
| Robin_Watts | For what file? | 12:29.28 |
| chrisl | RBR1104559.pdf | 12:29.43 |
| Robin_Watts | At 400dpi, yes. | 12:29.47 |
| | Do you not get a warning message out before the rangecheck ? | 12:30.03 |
| chrisl | Yes. | 12:30.23 |
| Robin_Watts | At 400dpi the tif file would be 5Gig. | 12:30.27 |
| | and tiff internal pointers are limited to 2gig. | 12:30.46 |
| chrisl | Ah, okay, not much we can do about that! | 12:31.28 |
| | Robin_Watts: these non-encodeable pixels, other than throw a warning, how do we currently handle it, do you know? | 12:38.00 |
| Robin_Watts | We throw a warning and put 0's in, AIUI. | 12:38.14 |
| | and rangecheck. | 12:38.53 |
| chrisl | I wondered if (for the short term) we could get away with using the nearest value we can represent - like a poor man's palette | 12:39.34 |
| Robin_Watts | I fear that's destined to fail; if the file prints all the separation 1 stuff first, and fills up the encoding, then prints separation 2, then there will be no 'close' colors left for sep 2. | 12:40.41 |
| chrisl | Well, I wasn't proposing it as a solution, just a "get the file out" workaround | 12:41.59 |
| Robin_Watts | Personally, I'd rather see a complete failure rather than an "oops, we printed 4000 copies of something that's missing half the colors" | 12:49.05 |
| chrisl | Well, I argued the same thing for missing CIDFonts and no one agreed! | 12:51.03 |
| Robin_Watts | Consistency is our hallmark, except when it's not. | 12:57.13 |
| chrisl | :-) obviously, immediate error out would have to remain an option..... | 12:57.49 |
| Robin_Watts | lunch | 13:19.27 |
| | tor8: ping ? | 16:09.39 |
| Robin_Watts | puts his hand in his tea. | 16:09.56 |
| sebras | Robin_watts: I'm going for korean barbecue at tor8's place so I believe he won't be here until tomorrow... | 16:11.19 |
| Robin_Watts | ah, fair enough. | 16:11.39 |
| | I'm doing this 'load images compressed and decode on demand' thing, and wanted to run some of the changes past him. | 16:12.06 |
| | I'll keep plodding and he can have the whole horror of it revealed to him later ;) | 16:12.25 |
| sebras | Robin_Watts: right. for jpeg only? | 16:12.26 |
| Robin_Watts | For all images (except JPX) is the way it's falling out. | 16:12.45 |
| sebras | as long as you have a good set of tests and it works I think he'll be happy. :) | 16:12.47 |
| Robin_Watts | I'm introducing an fz_image type that has a 'get_pixmap' function in it. | 16:13.17 |
| | so the existing code will change to passing images around and just getting the pixmap when it needs it. | 16:14.04 |
| | I've updated the PDF stream handling code so that it can stop before it does the last filter step, and records the step that's required into a structure. | 16:15.07 |
| | So we load the LZW or Flate or JPEG or Fax data and keep it compressed. | 16:15.33 |
| | And the get_pixmap function will decode it on demand. | 16:15.43 |
| sebras | why is JPX exempt? | 16:16.27 |
| Robin_Watts | because JPX has special handling already. | 16:16.38 |
| sebras | rushes off... | 16:16.50 |
| Robin_Watts | cu later. | 16:16.55 |
| henrys | Robin_Watts:did we ever agree on a way to query for a high level device? | 16:43.59 |
| Robin_Watts | Urm... | 16:44.21 |
| henrys | I ask you because I thought it was something possible with the lightweight parameter stuff you did. | 16:45.18 |
| Robin_Watts | curses mvrhel2 for splitting the headers up in the solution. | 16:45.29 |
| | Look in gxdso.h - that's what I'd do if I could find the damn thing. | 16:45.48 |
| henrys | don't curse mvrhel2 we need him. | 16:46.18 |
| Robin_Watts | gs/base/ gsdevsop.h | 16:48.06 |
| henrys | thanks I'll have a look | 16:48.48 |
| Robin_Watts | So, no, doesn't look like there is anything to say "are you a high level device" | 16:49.46 |
| | What exactly constitutes a high level device ? | 16:49.54 |
| chrisl | I thought Ken added a generic way to commonicate high level capabilities from the device to the interpreter | 16:51.03 |
| henrys | he he - something that copies the palette in an indexed color space when it is set. | 16:51.07 |
| | the first time. | 16:51.27 |
| Robin_Watts | henrys: Right, so I suspect you want to add a gxdso_handles_indexed_palette or something. | 16:51.55 |
| | (i.e. you want a gxdso to detect that it handles that specific thing rather than a nebulous 'is a high level device') | 16:52.20 |
| henrys | yes that would be best. | 16:52.30 |
| Robin_Watts | Dead easy to do - let me know if you want me to walk you through it. | 16:53.01 |
| henrys | indexed_palette_const seems reasonable. | 16:53.22 |
| | I'll give it a go and ask you if I need help | 16:54.07 |
| | chrisl:one concern with not recommending the ufst is all the cjk typefaces that come with it. I am not sure it shouldn't be recommended in this case, but we'll see what happens. | 16:57.27 |
| chrisl | henrys: the main thing was I didn't want it forgotten that we had that option. | 16:58.11 |
| | s/was/want | 16:58.16 |
| Robin_Watts | The 'standard' sets of latin fonts that come with postscript/PDF... are there equivalent CJKV sets ? | 16:58.23 |
| chrisl | No | 16:58.34 |
| | henrys: Ken's device interrogation code uses gs_getdeviceparams() see commit d36e6d74ce15d2dbe2efb61970145dca8b57d5db for example | 16:58.53 |
| Robin_Watts | So there is no tickbox equivalent to Courier,Times, etc but for the far east? | 16:59.31 |
| chrisl | No, not really. There are some very common fonts, but nothing set in stone | 17:00.38 |
| Robin_Watts | OK, so we can't be criticised for not supplying a 'standard set' in the same way we supply a standard set of latin ones. | 17:02.31 |
| | Every email from your Japanese customer renders in Windows Mail in a way that reminds me of: http://xkcd.com/1015/ | 17:03.34 |
| henrys | chrisl:my concern is the customer will dismiss our solution becuase they assume (properly) we don't have those fonts. | 17:03.40 |
| Robin_Watts | s/Windows Mail/Thunderbird/ | 17:04.46 |
| chrisl | henrys: well, I'm sure you've made that point to Miles and Scott, hopefully, if they feel that might happen, they'll mention UFST | 17:05.50 |
| | Robin_Watts: the new customer? It's odd, my Thuderbird is set to display in "plain text", but what I'm seeing in the window isn't pain text..... | 17:06.41 |
| Robin_Watts | His email display the same regardless of whether I have it set to Plain text or not. | 17:08.04 |
| | I think it's just detecting the fact there are CJKV glyphs and using a fallback font that looks REALLY horrible. | 17:08.43 |
| chrisl | Yes, you're probably right - but it *really* shouldn't do that :-( | 17:09.09 |
| Robin_Watts | "support ing" | 17:09.10 |
| henrys | chrisl:no I haven't I will now - I just started mulling over it. I've been buried doing other stuff and didn't think much about it. | 17:09.14 |
| chrisl | henrys: note that using the ACT fonts would need *some* extra development work - maybe a week, since the way cust 532's code does it is not entirely portable. | 17:11.08 |
| | Robin_Watts: tbh, compared to some of the Japanese dealings I've had in the past, I'm relieved at how good their English is! | 17:12.43 |
| Robin_Watts | It wasn't the english I was complaining about (it's SOOO much better than my Japanese). It was the fact that the fallback font is horribly designed so it looks like there is a whole extra space there. | 17:13.35 |
| henrys | chrisl:what else are you going to do over spring vacation? | 17:14.55 |
| chrisl | henrys: you mean, with my other hand on my other keyboard? How about cold fusion ;-) | 17:15.28 |
| henrys | ;-) | 17:15.51 |
| chrisl | It is on my to-do list, it's just unfortunate that it's one of a couple of things that isn't a simple cut'n'paste from 532's code | 17:16.39 |
| Robin_Watts | And I thought a "one handed computer user" was a euphemism... | 17:17.41 |
| chrisl | That's why I thought I'd better throw in the "other keyboard" part - didn't want people getting the wrong idea.... | 17:18.15 |
| henrys | chrisl:I think it should have pretty high priority. | 17:20.37 |
| | marcosw1:did you want me to field the question to ken? | 17:20.56 |
| chrisl | henrys: okay, I'll try to get it done between next week and the week after. | 17:21.18 |
| marcosw | henrys: I haven't read my email yet this morning, so give me a sec. | 17:21.34 |
| henrys | so when should I use gs_getdeviceparams vs. gxdso? | 17:21.54 |
| chrisl | henrys: My understanding (which may be wrong!) is that the gs_getdeviceparams() method is a fairly simple "can the device do...." yes/no. More complex things would need gxdso | 17:23.47 |
| henrys | marcosw:I looked a bit and there is a bug iin there somewhere but the customer (as is his way) has sent us the wrong output and a jumbled up message. | 17:23.50 |
| Robin_Watts | AIUI, gs_getdeviceparams makes a dictionary of ALL the device params and returns it. | 17:25.05 |
| henrys | marcosw:much rather you sort through that and talk to him. I'm looking at the bug "I actually see" | 17:25.05 |
| Robin_Watts | which we then look in for an answer. | 17:25.23 |
| | gxdso allows us to query a specific thing. | 17:25.36 |
| | so gxdso is lighter weight. | 17:25.55 |
| henrys | okay I'll probably use gxdso. | 17:25.58 |
| chrisl | But then everything for pdfwrite and ps2write and txtwrite is using gs_getdeviceparams(). | 17:26.33 |
| Robin_Watts | Adding a gxdso handler to those is trivial. | 17:28.35 |
| | Especially one that just has to say "YES!" | 17:28.44 |
| | and do we really want to be exporting device params (which are postscripty thing, right?) into PCL ? | 17:29.21 |
| chrisl | But why add a new way to do something we already have a well used method for? | 17:29.33 |
| Robin_Watts | because the change to the code to add gxdso is smaller than the change to the code to do it the old way? :) | 17:30.29 |
| chrisl | So: "here's how we interrogate a device to determine it's capabilities - well, except for this one thing, which we do totally differently"? | 17:31.29 |
| | henrys: I take it this is to do with how PCL (or HPGL?) uses indexed color spaces confusing pdfwrite? Does changing the color space ID actually hurt performance? | 17:35.31 |
| Robin_Watts | Random idea, possibly rubbish... | 17:38.10 |
| | If I remember correctly the problem here was that PCL passed the indexed space across to pdfwrite, then added a new colour to the end, and passed it again. | 17:38.42 |
| | Then added a new color and passed it again, etc. | 17:38.50 |
| chrisl | I'm assuming that's what henrys is looking at, yes - maybe even removed colors, too? | 17:39.40 |
| Robin_Watts | Could we maybe NOT update the colorspace id if we're only appending to it? And have pdfwrite always recopy the palette? | 17:39.58 |
| henrys | there are many cases in pcl where we change the palette after calling setcolorspace. | 17:40.40 |
| chrisl | Robin_Watts: That bloats the resulting PDF | 17:41.07 |
| marcosw | henrys: I'll deal with the "PCL6.exe Output Print Issues" email. | 17:41.15 |
| Robin_Watts | chrisl: I was hoping that pdfwrite would only keep 1 copy of the palette with the same id. | 17:41.36 |
| chrisl | I don't *think* that's how it works just now, but it might be possible to do that. I still think changing the color space *after* it's been set is just a bug...... | 17:42.36 |
| marcosw | tor8: you forwarded an email with the subject "using mupdf in our android app" to support. Was this for information or did you need me to respond to the potential customer? | 17:42.52 |
| Robin_Watts | henrys: Does PCL need to do that? | 17:42.53 |
| marcosw | henrys: I'm finally ready for the support meeting. It was a busy week :-) | 17:43.08 |
| henrys | Robin_Watts:it would be awkward at best to arrange it otherwise. | 17:43.41 |
| | oh yea support | 17:43.57 |
| Robin_Watts | Can it not hold off on doing the setcolorspace until it has to? (i.e have a dirty flag and check that at every place before using the color) | 17:44.06 |
| Robin_Watts | shuts up. | 17:44.13 |
| chrisl | Robin_Watts: I have a feeling we argued with Henry before about this, and failed to browbeat him ;-) | 17:45.21 |
| henrys | I'm afraid chrisl is right. | 17:47.02 |
| marcosw | I think the first bug that needs to be discussed is 851, which I'll reassign. | 17:47.20 |
| henrys | marcosw I'm finding the report | 17:47.31 |
| | start with 692848 | 17:48.13 |
| | oh okay 851 sorry didn't see that. | 17:49.01 |
| | assign that to me it's first stop should be luratech. | 17:49.40 |
| marcosw | done | 17:50.10 |
| henrys | do you want to verify 852? | 17:50.58 |
| | I think alexcher jumped into support duty briefly for these problems. | 17:51.20 |
| marcosw | I presume Alex's analysis is correct. He states that Evince and AR9 render the file correctly and we don't. I don't think I could do much more (I could test with mu but am not sure how that would help). | 17:52.25 |
| henrys | I"m inclined to give it Robin_Watts who in turn might give to mvrhel2 | 17:53.04 |
| | are you driving? | 17:53.28 |
| marcosw | yeah, I was just rereading the bug. Why Robin_Watts instead of michael directly? | 17:54.05 |
| henrys | either is fine - it seems like a graphics library thing to me. | 17:54.49 |
| marcosw | okay, I'll reassign it to Robin_Watts | 17:55.04 |
| Robin_Watts | Well, it was alex that put the code in that's going wrong... | 17:55.31 |
| henrys | I'm assuming that has to be reverted and reanalysis is needed. | 17:55.58 |
| Robin_Watts | Ok. | 17:56.08 |
| marcosw | the next interesting bug is 856. Do you have any comments? | 17:56.34 |
| henrys | seems properly assigned we could up the priority | 17:57.47 |
| | P2? | 17:57.51 |
| marcosw | okay, seems likely to be easily fixable. | 17:58.10 |
| | we should talk about 859, a P1 enhancement. | 17:59.02 |
| henrys | 859 ;0( | 17:59.09 |
| | that has to be done with an option and it shouldn't be the default behavior | 18:00.11 |
| | better than acrobat ;-| | 18:00.36 |
| marcosw | I think the customer would be fine with that, particularly if pdf_info.ps returned that the file was a portfolio PDF. | 18:01.43 |
| | right now it just says it's 1 page at 5 x 7. | 18:01.58 |
| | alexcher: do you want comment on bug 692859? | 18:02.54 |
| henrys | so we should ask alexcher to make this optional behavior | 18:02.58 |
| | ? | 18:02.59 |
| marcosw | sure, I can add a comment to that affect. | 18:04.30 |
| henrys | 860 miles is dealing with. | 18:04.40 |
| marcosw | right, I didn't see they had opened a bug as well as sent email to support. You can ignore my email to you regarding this. | 18:05.22 |
| Robin_Watts | 862 is me. | 18:05.35 |
| henrys | I'll add a little something to 860. | 18:05.43 |
| marcosw | I've reassigned 862 to Robin_Watts | 18:06.12 |
| henrys | oh he's broken the music. | 18:07.12 |
| marcosw | nice when the bug report does the bisect for me :-) | 18:07.40 |
| | ^report^reporter | 18:07.48 |
| henrys | yeah those guys are good. | 18:08.09 |
| | that's it right? | 18:08.14 |
| marcosw | I think that's it. I do have question for you about tracking down a ghostspcl issue. | 18:08.34 |
| Robin_Watts | "broken the music" ? | 18:08.44 |
| chrisl | lillipond is a music typesetting app | 18:09.13 |
| Robin_Watts | Ah. | 18:09.20 |
| chrisl | Lilypond, I mean | 18:09.25 |
| marcosw | I recently brought up a new cluster node (tigerdirect was selling an x6 system for $300). It's running ubuntu 10.04 but every Ghostpcl command fails: | 18:10.27 |
| | marcos@feet:[187]% ./pcl6 -sOutputFile=test.ppm -sDEVICE=ppmraw -r75 -dNOPAUSE -dBATCH /home/marcos/cluster/tests/pcl/jason_top.pcl | 18:10.28 |
| | Warning interpreter exited with error code -1 | 18:10.28 |
| | Flushing to end of job | 18:10.28 |
| | marcos@feet:[188]% | 18:10.28 |
| | I moved the pcl6 binary from the new node to an existing one and it works fine, so it's apparently a shared library issue or something else weird, but I'm not sure how to debug this. Any thoughts? (BTW, ghostscript and ghostxps work fine). | 18:11.30 |
| chrisl | I can't imagine GhostPCL is using a shared lib that Ghostscript isn't...... | 18:12.32 |
| henrys | make pcl-debug and run with -ZIi | 18:13.18 |
| | or you don't think it is even getting past loading libraries? | 18:14.03 |
| marcosw | No, it generates some pages of (apparently) correct output and then fails. | 18:14.25 |
| henrys | then -ZIi# is best. | 18:14.41 |
| tor8 | marcosw: the latter | 18:14.42 |
| henrys | marcosw:if you want report a bug and put my key on the box. | 18:15.17 |
| marcosw | tor8: will do. Do we have documentation on how to do this? | 18:15.33 |
| chrisl | marcosw: one other thought - try building without the x11 and/or display devices (I forget which goes into the Unix pcl build). The X11 libraries pull in a *load* of extra shared libs that we otherwise don't need. | 18:16.16 |
| marcosw | henrys: [Printer Reset] | 18:16.18 |
| | ESC & l | 18:16.18 |
| | (ESC & l) 2 A | 18:16.18 |
| | [Page Size] | 18:16.18 |
| | D | 18:16.18 |
| | ocuments for this printer should have 1/4" margins on all sides. | 18:16.18 |
| | Warning interpreter exited with error code -1 | 18:16.18 |
| | Flushing to end of job | 18:16.19 |
| | [i][file pos=1020168] | 18:16.19 |
| | [i][file pos=1020230] | 18:16.20 |
| henrys | Robin_Watts:the regresion 692862 is from lilypond they render high quality music scores | 18:16.57 |
| Robin_Watts | right. | 18:17.07 |
| henrys | oh sorry didn't see chrisl's response. | 18:17.49 |
| marcosw | chrisl: I can try that. | 18:19.13 |
| henrys | marcosw:did you run with # also.? | 18:19.26 |
| marcosw | yes, -ZIi# (that was capital-i, not lower case L :-) | 18:19.59 |
| henrys | marcosw:very odd probably need to debug. | 18:20.31 |
| marcosw | henrys: Okay, I'll work on getting you access. | 18:21.20 |
| henrys | maybe my jason file is different than yours - there should be a couple command before the page size and no printer reset. | 18:22.24 |
| marcosw | the thing that bugs me is that I just set up your newly formatted henrysx6 as a cluster node, so all the steps were fresh in my mind (as well as being well documented in the readme), so I don't see how I can have screwed up the installation. at the same time the operating systems, Ubuntu 10.04, is the same as most of the other cluster nodes, so that doesn't seem likely. And we have several x6 cluster node cpus (and the /proc/cpuinfo is the same for thi | 18:24.33 |
| | henrys: It doesn't have to be jason, other files fail too. | 18:24.46 |
| | here's the output from owl.pcl: | 18:25.29 |
| | marcos@feet:[218]% main/debugobj/pcl6 -ZIi# -sOutputFile=test.ppm -dMaxBitmap=30000000 -sDEVICE=ppmraw -r75 -dNOPAUSE -dBATCH -K1000000 -dClusterJob /home/marcos/cluster/tests/pcl/owl2.pcl | 18:25.35 |
| | [i][file pos=0] | 18:25.36 |
| | ESC E | 18:25.36 |
| | [Printer Reset] | 18:25.36 |
| | ESC & l | 18:25.36 |
| | oops | 18:26.33 |
| henrys | never seen irc error Excess Flood before | 18:26.37 |
| chrisl | Holy Moses! | 18:27.12 |
| henrys | ;-) | 18:27.31 |
| chrisl | Time for me to go - enjoy your weekends! | 18:28.58 |
| henrys | you too chrisl | 18:29.13 |
| chrisl_away | Thanks! | 18:29.18 |
| marcosw | chrisl_away and others in the UK and Sweden: Monday is a holiday for the US, so see you on Tuesday. | 18:29.52 |
| chrisl_away | marcosw: in that case, you can enjoy it 50 per cent more ;-) | 18:30.36 |
| marcosw | henrys: besides dynamic libraries what else might the pcl6 code load during execution? Fonts, color profiles, etc.? I'm going to try and build a statically linked binary next. | 18:34.00 |
| henrys | what in pcl isn't static ... just X stuff really right? | 18:40.43 |
| | what does ldd pcl6 say on your box? | 18:41.06 |
| | valgrind could also help here. | 18:41.39 |
| marcosw | it's not dynamic library issue. Using chrisl_away's suggestion I built without the X11 devs and then statically linked it. It fails in the same way (i.e. on feet it generates a "Warning interpreter exited with error code -1" and on x6 it runs fine)/ | 18:42.31 |
| Robin_Watts | marcosw: It's not something stupid like the clist temp files filling a disc partition or something ? | 18:43.47 |
| henrys | strace and valgrind would be my next choices. | 18:44.52 |
| Robin_Watts | actually that command shoudn't use clist | 18:45.06 |
| marcosw | I assume it's something stupid :-) but that's not it, plenty of disk space (only one partition, except for swap). Also it fails with even a large -dMaxBitmap and a small -r value) | 18:45.15 |
| | I wish ghostscript failed as well, then I could blame it on hardware... | 18:45.51 |
| | I'm installing valgrind, but it's taking ~12 minutes to download, ubuntu's servers must be busy. | 18:46.22 |
| henrys | strace should say something interesting too. | 18:47.24 |
| marcosw | I'd paste the strace results but that would get me kicked off again :-) | 18:47.54 |
| henrys | you can email them to me. | 18:48.38 |
| | problems like this are annoying. I assume you tried touch test.ppm (permissions) | 18:49.48 |
| marcosw | strace showed something, ghostpdl tries to open ("/home/marcos/artifex/ghostpdl"), which fails on the feet | 18:51.09 |
| | open("/home/marcos/artifex/ghostpdl", O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC) = -1 ENOENT (No such file or directory) | 18:51.13 |
| | I'm guessing an environemnt variable is to blame... | 18:51.33 |
| | yup: PCLFONTSOURCE=/home/marcos/artifex/ghostpdl/urwfonts | 18:51.53 |
| henrys | we certainly need a better error message. | 18:52.20 |
| marcosw | All good now, thanks for everyone's help. The strace suggestion was the answer. | 18:53.05 |
| | does beg the question as to why Ghostpdl was trying to load fonts... | 18:53.36 |
| henrys | pcl always loads fonts. | 18:54.40 |
| | if they are there. | 18:54.45 |
| | if you are going to get all explicit with an env variable it assumes you have fonts. | 18:55.11 |
| | but it does need a better message I'll make a bug. | 18:55.30 |
| marcosw | the really odd thing was that on large files like jason it generates many pages of correct output before getting an error and quitting. Does that make any sense? | 18:58.07 |
| tor8 | marcosw: Robin is better equipped to answer how to build on android, but you can point him to the readme in the android directory first since that has a detailed guide how to do it | 19:04.49 |
| Robin_Watts | tor8: I've obviously missed stuff here. | 19:05.18 |
| | tor8: I have *just* got load on demand working for images in pdfs. | 19:05.48 |
| tor8 | Robin_Watts: marcosw asked about the potential customer / gpl violator whose email I forwarded to support | 19:07.31 |
| henrys | marcosw:I might be incorrect that it fails with the env variable. Without an env it will fail if it reaches a font that is not embedded or a stick/arc font - the latter are built in. | 19:08.31 |
| | if there are no fonts. | 19:08.42 |
| Robin_Watts | tor8: We may need to rework the store slightly in this case. | 19:10.49 |
| | actually, you're having a social evening, right? we can discuss this on monday. | 19:11.35 |
| | What does pdf_info.ps do if you call it on a portfolio pdf ? | 19:15.57 |
| | Just gives you the answers for the cover page? | 19:16.15 |
| | Does it also say "This is a portfolio pdf" or something, so that people know to try the new option? | 19:16.52 |
| henrys | alexcher are you around to discuss this? | 19:37.25 |
| Robin_Watts | Making people use an option to get the contents of a portfolio PDF seems entirely fair, but it would be nice to give them some hint that they should try that. | 19:38.47 |
| henrys | that does seem reasonable to me also. | 19:39.17 |
| | I'll send email to alexcher to read the logs. | 19:44.04 |
| marcosw_ | henrys and Robin_Watts: I considered suggested that pdf_info.ps return "this is a portfolio pdf" but decided that would require a portfolio file to be run twice. seems better to just have pdf_info.ps return a "this is not a portfolio pdf" and the info for the file if the read as portfolio pdf option is given. | 19:59.29 |
| Robin_Watts | I am not familiar with the output of pdf_info.ps, so I'm speaking from a position of ignorance here. | 20:00.46 |
| | but I have to imagine that it would be easier for most people to adapt their code to run gs a second time on the specific portfolio entry they are interested in, than to suddenly have to cope with lots more lines of output. | 20:01.55 |
| | Perhaps what we need is for pdf_info.ps to say: "NumberOfPortFolioEntries=%d" (or whatever syntax is in keeping). | 20:02.43 |
| | That way they can run once and get (almost) the same results as they currently do (we can swallow the 0 case if we want). | 20:03.33 |
| | And if portfolio entries *are* present, the user needs to be told how many there are to know how many to enumerate though. | 20:04.03 |
| rayjj | Hi, guys. Orignally pdf_info.ps only gave minimal info by default, but feedback led us to have it give out more. I agree that dumping info for all portfolios seems the most user friendly. | 20:05.14 |
| | I was thinking that rather than saying just "Page 1" we could say "Page M File N MediaBox ..." | 20:06.53 |
| Robin_Watts | File N Page M MediaBox ... ? | 20:08.25 |
| rayjj | Robin_Watts: whatever. | 20:08.47 |
| Robin_Watts | That way people that grep the output for lines starting 'Page %d' will still work. | 20:08.56 |
| rayjj | If we want to make it more easily parsed, we could always keep the File N Page M but have a way for the user to be able to tell if this was a portfolio or not | 20:10.11 |
| | Maybe File 0 for non-portfolio | 20:10.25 |
| | or File N of M Page P | 20:10.58 |
| | but I'll leave it up to whoever does it (alex? ken?) I've sort of got a full plate right now, even though I've usually been the pdf_info guy -- I'm glad to hand it off | 20:12.15 |
| Robin_Watts | I was thinking we'd have "Page 1" for the pdf itself, then "File 1 Page 1" but yes, any reasonably sane scheme should be fine. | 20:13.17 |
| rayjj | but I think the default (when run with gs -- toolbin/pdf_info.ps my.pdf) should give the info for all pages of all files | 20:16.08 |
| Robin_Watts | OK. I have mupdf loading images on demand (and with the ability to request subsampled ones in the case of JPEGs). I'm still passing pixmaps through the device interface though, so the images are still being requested at interpret time. I'll change that on monday. | 20:47.35 |
| | Forward 1 day (to 2012/02/18)>>> | |
IRC Logs
