| <<<Back 1 day (to 2011/09/25) | 2011/09/26 |
cryptopsy | alexcher: scrample? | 01:35.09 |
alexcher | cryptopsy: scramble | 02:01.05 |
| cryptopsy: for instance, use 2 rounds of rot13. | 02:02.08 |
cryptopsy | oh, i can't do that since it is scanned | 04:20.34 |
kens | cryptopsy if you can't share your document, or make one which causes a similar problem, there's little we can do to fix your problem. I think you are using an old version of Ghostscript (8.71) so you might like to try a more recent version. I have fixed a couple of bugs which could be described similarly to your problem. | 06:58.21 |
tkamppeter | chrisl, hi | 07:09.42 |
chrisl | tkamppeter: hi | 07:10.16 |
tkamppeter | chrisl, have you already seen bug http://bugs.ghostscript.com/show_bug.cgi?id=692539, Apple-generated EPS files make GS 9.0x crash. | 07:10.54 |
kens | I've seen it, but they work OK on Windows and Linux 32-bit. | 07:11.13 |
| Could be OS-specific or 64-bit ? | 07:11.25 |
chrisl | tkamppeter: yes, I saw it, but I have my hands full with another problem right now. Have you any idea where the problem might be? | 07:11.31 |
tkamppeter | kens, no 64-bit system for testing around? I can reproduce it on my 64-bit machine. | 07:12.12 |
| chrisl, kens, we have Final Freeze for Oneiric on Thursday, I am grateful if we could sort it out by then. | 07:13.03 |
kurtzhong | could i use ps2pdf14 convert a pdf file to later version? | 07:14.30 |
kens | tkamppeter, I don't have a 64-bit system I'm afraid. | 07:15.06 |
tkamppeter | kens, chrisl, a symbolic stacktrace is here: https://bugs.launchpad.net/ubuntu/+source/ghostscript/+bug/857948/+attachment/2450894/+files/Stacktrace.txt | 07:15.19 |
kens | kurtzhong pdfwrite will not use features later than PDF 1.4 | 07:15.19 |
| You can still create a version stamped as a later version, but not by using the script. | 07:15.44 |
| THis can be useful if you intend to supply pdfmarks which are only valid in a later version of PDF. | 07:16.04 |
chrisl | tkamppeter: well, that appears to be crashing in lcms, and we have fixed several problems in lcms, so I would suggest you test it with the lcms we supply. | 07:16.29 |
kens | tkamppeter, yes I looked at that and i... what chrisl said | 07:16.42 |
| It might explain why I cannot reproduce it. | 07:17.07 |
kurtzhong | then could i convert a newer pdf file to version 1.4 using ghostscript? | 07:18.32 |
kens | kurtzhong yes, but you will lose something in the process. | 07:18.51 |
| It may be as innocuous as removing linearization, or not using compressed xrefs, or it may mean some objects are turned into images | 07:19.31 |
tkamppeter | chrisl, kens, thanks. | 07:21.50 |
kurtzhong | thanks, ken | 07:23.07 |
tkamppeter | kens, chrisl, is it possible to switch GS to libcms2? | 07:27.30 |
chrisl | No | 07:29.56 |
tkamppeter | chrisl, why not? | 07:39.28 |
chrisl | Because it has bugs which mean it doesn't work right with Ghostscript | 07:40.05 |
tkamppeter | chrisl, OK. | 07:40.19 |
chrisl | These are taking more time to resolve than the bugs we found in lcms1 because, since lcms2 is still being worked on, we want to work with Marti to get these fixed in his distribution, rather than just our own. | 07:41.33 |
sebras | cryptopsy: just an idea -- is there any specific page of your document that can be shared? if so maybe it is possible to extract a subset of the file and hope that the problem persists with only that single document. | 07:52.35 |
chrisl | tkamppeter: I just did a quick test: if I (on Maverick) build our development GS to link with the system lcms, those files crash, and if I build in *our* lcms version, they all work just fine. So clearly, the problem is a bug in standard lcms. | 09:13.26 |
tkamppeter | chrisl, how can I find out which bugs of standard lcms are fixed in your custom lcms? | 09:38.01 |
chrisl | comparing the source would be the only way. As Marti had ceased work on lcms1 by the time we found the problems, I don't think we officially reported the problems. | 09:41.00 |
| tkamppeter: ^^ | 09:41.10 |
| tkamppeter: also, I'm at all sure whether the changes we've had to make would prevent our lcms1 source being used to build a library package. Can't you just use ours for Ghostscript this time? FWIW, for this very reason, we'll be removing the option to share lcms for 9.05. | 09:45.01 |
Robin_Watts_ | tkamppeter: lcms1 does a poor job of checking for allocation failure. | 09:57.18 |
tkamppeter | WDYT is better for the GS in Ubuntu, switch to lcms2 or try to fix the lcms1 issue? | 10:00.37 |
Robin_Watts_ | tkampeter: neither would be sensible. | 10:00.56 |
| The sensible course is to build our lcms v1 in. | 10:01.14 |
kens | nods | 10:01.24 |
Robin_Watts_ | It's the only sensible, tested method. | 10:01.33 |
| It's not a huge library, and we are working to get to a stable state where shared libs can be used safely with lcms2 as soon as possible. | 10:02.04 |
| To dogmatically stick with a shared lib at the moment will give you problems. It's just not a smart thing to do in this instance. | 10:02.33 |
kens | henrys study should be nice and warm when he gets up.... | 11:12.08 |
Robin_Watts_ | Well, according to marcosw's latest figures, plank is now faster than pamcmyk4 on average. | 13:47.41 |
| Without mvhrel2's fast halftoning code. | 13:47.53 |
kens | Is that expected ? | 13:49.18 |
Robin_Watts_ | ish. | 13:49.35 |
| I believe the result. | 13:49.43 |
kens | Not what you had hoped for though ? | 13:49.59 |
Robin_Watts_ | No, I'm pleased. | 13:50.24 |
kens | Well, good news tehn :-) | 13:50.36 |
Robin_Watts_ | I could also believe that there is something in pamcmyk4 that if fixed made it faster though :) | 13:50.43 |
Fandekasp | hi | 14:32.21 |
ghostbot | hi | 14:32.21 |
Fandekasp | I have a question about mupdf. If it's possible to search into it, why can't I copy the text from the page ? | 14:32.51 |
kens | I guess because ithat isn't implemented. | 14:34.20 |
| tor8 ? | 14:34.24 |
tor8 | Fandekasp have you tried selecting a region of text? | 14:34.49 |
kens | There is a text extraction module for MuPDF | 14:34.55 |
kens | has never tried copying text with MuPDF | 14:35.10 |
Fandekasp | tor8: yes | 14:35.22 |
tor8 | that follows the unix way of putting the selected text into x11:s selection buffer, which can be pasted like any other text using middle click | 14:35.26 |
Fandekasp | What it does for me is scrolling pages | 14:35.37 |
tor8 | or on windows, it just copies it into the clipboard | 14:35.38 |
kens | Maybe its not text ? | 14:35.40 |
tor8 | right click | 14:35.44 |
| or, well, right drag | 14:35.50 |
kens | Perhaps the text is actually an imaghe, or vectors linework | 14:35.59 |
kens | is typing particularly badly today | 14:36.15 |
Fandekasp | no doesn't really change. After, my trackpad settings can be bad enough that it wouldn't do a proper right clic | 14:36.52 |
| but I tried alt+clic without success as well | 14:37.07 |
| it's a pdf generated from a latex document, the text should be text | 14:37.32 |
tor8 | right dragging on a page in mupdf inverts the selected region and copies what text there is in the region to the clipboard | 14:37.42 |
kens | Fandekasp : can you share the PDF file ? | 14:38.24 |
henrys | Robin_Watts:I expected the same speed, odd. | 14:39.01 |
Fandekasp | kens sure: http://mirror.ctan.org/graphics/pstricks/base/doc/pstricks-doc.pdf | 14:39.23 |
tor8 | kens: I believe it's a UI problem more than an actual problem with the text? | 14:40.48 |
| Fandekasp, kens: I can copy the text in that file just fine | 14:41.03 |
Robin_Watts_ | henrys: On some we're faster, on others we're slower. | 14:41.09 |
| but it averages out. | 14:41.13 |
Fandekasp | tor8 if I try really hard (my tracpad really sucks), I can make a black rectangle, but when I leave my fingers, nothing of what was selected by the rectangle has been copied | 14:41.13 |
henrys | tor8:I know you are busy with important projects but the xps bugs are P2's. | 14:41.20 |
kens | tor8 beat me to it :-( | 14:41.21 |
tor8 | Fandekasp: are you using a macbook? | 14:41.33 |
Fandekasp | tor8: yes | 14:41.39 |
tor8 | macosx? | 14:41.46 |
Fandekasp | with sabayon on it | 14:41.46 |
| gentoo | 14:42.11 |
tor8 | right. if it behaves anything like the osx trackpad, two-finger click followed by lifting one finger and drag with the remaining should do it | 14:42.54 |
kens | Yes, right-click select Ctrl-C works for me to copy the text | 14:43.06 |
tor8 | kens: you don't actually need ctrl-c, it's one of those unix-isms left over in the windows port | 14:43.30 |
kens | Oh, well :-) | 14:44.46 |
| Still works though.... | 14:45.00 |
Fandekasp | doesn't work. I'll try to map the right click to another key, good to know that mupdf can copy finely ! | 14:46.06 |
tor8 | Fandekasp: then buy a proper mouse, like this one http://h10010.www1.hp.com/wwpc/us/en/sm/WF06c/A10-51210-69998-329254-69998-403892-403894-403895.html | 14:46.22 |
| Fandekasp, or you can edit the source and make it select with a left click instead ;) | 14:46.58 |
Fandekasp | tor8: I won't buy a mouse just for mupdf :P I almost never need to use the tracpad | 14:47.01 |
| ahaha I already have a patched version of mupdf, why not :) | 14:47.25 |
tor8 | I'm surprised people can use linux without a proper three button mouse... I guess it's all down to gnome/kde mimicking windows style UI:s | 14:47.55 |
Robin_Watts_ | realises that he didn't get any answers to his holiday mail, because he sent it to the wrong address :( | 14:48.33 |
Fandekasp | :/ | 14:48.34 |
| no I have everything in tilda/tmux, and just use the gui for chromium, still using keymaps. btw my wm is xfce4 | 14:49.16 |
| Can't believe you'd think linux is just gui xD | 14:49.45 |
tor8 | up until about a year ago I used evilwm + xterm + acme and was happy with that | 14:50.22 |
| then X.org botched something and since then window borders don't get redrawn properly | 14:51.00 |
Fandekasp | maybe you f*** something in your kernel conf ? | 14:52.00 |
tor8 | no, it's a graphic repaint/dirty flag/clip region bug. | 14:52.27 |
Fandekasp | if you try recompiling with a kernel seed, that might be fixed | 14:52.28 |
tor8 | which I guess nobody has discovered or care about since very few people use twm or evilwm | 14:52.32 |
Fandekasp | lol | 14:52.44 |
| yes never tried it yet :) | 14:52.55 |
tor8 | borders get drawn, but not flagged dirty when obscured so they're not repainted when windows move | 14:52.56 |
Fandekasp | I got a similar problem with e17, but don't remember how I fixed it | 14:53.27 |
tor8 | there's a similar bug in the os x version of X11, if you use xterm with a scroll bar, the line separating it from the main window isn't repainted properly either. that also uses window borders. | 14:53.45 |
| gtk+ and Qt don't use window borders, so I blame them again ;) | 14:54.01 |
kens | Robin_Watts_ : early March could be a problem for me, I'm trying to find out. | 14:55.16 |
Robin_Watts_ | kens: Thanks. | 14:55.43 |
kens | Forget that, we are away in February, but would still prefer to avoid very early March as that's my daughter's birthday. not | 14:56.41 |
| Umm excise the not from that last sentence. | 14:56.50 |
Robin_Watts_ | There is only 1 flight to Bhutan every other day, and there is a religious festival then, so apparently the flights are filling up. | 14:57.20 |
tor8 | Robin_Watts: I'm planning a vacation around the same dates as you, so early march fits me great | 14:57.22 |
Robin_Watts_ | tor8: Ah, strength in numbers :) | 14:57.53 |
| henrys: What's the Artifex policy on things like this? Do we clear dates with Miles, or do we all just book and then fight it out later ? | 14:58.24 |
henrys | Robin_Watts:I'm good with your dates. | 14:58.26 |
| Miles pretty much handles that. | 14:59.01 |
tor8 | henrys: so we should all have rolling vacations so there's no single date we can all go, and there won't be any meetings! ;) | 14:59.45 |
henrys | my job is to remind him each quarter we don't actually need to have a meeting ;-) | 14:59.46 |
| This is probably the only area I have been completely unable to budge Miles. Sorry. | 15:01.31 |
Robin_Watts_ | felt the last meeting was very productive. | 15:01.49 |
henrys | I think we should have meetings, but 4 is excessive for euro folks. | 15:04.25 |
Robin_Watts_ | ponders; are parts of the different test suites repeated? | 15:13.10 |
| We should md5 the test files, and remove (or at least disable) the duplicates. | 15:13.30 |
chrisl | I think there are some of the QL tests in compare_files, possibly others. I don't think there are enough to worry about time on the cluster, but it can be a little annoying doing a bmpcmp | 15:14.43 |
henrys | marcosw's has checked that in the past. | 15:15.24 |
chrisl | Possibly he's looked at it since I noticed..... | 15:16.41 |
Robin_Watts_ | Well, there are various C302.bin files, for example. | 15:17.22 |
chrisl | Robin_Watts_: actually, something I was thinking about was that it might be good to be able to tell bmpcmp to test without only one device. | 15:17.50 |
Robin_Watts_ | I have a plan :) | 15:18.01 |
kens | A cunning one no doubt | 15:18.26 |
Robin_Watts_ | The plan is that you should be able to give the cluster a 'filter'. | 15:18.29 |
| So something like: clusterpush.pl gs -filter=ppmraw.300.0 | 15:19.06 |
chrisl | That would be good - it's just, a change at the interpreter end doesn't generally need tested with lots of devices, at least, not every test run | 15:20.54 |
Robin_Watts_ | And the cluster would filter the 'jobs' line. | 15:21.09 |
| It would mean we could test just a given device, or a different resolution, or banding, or even a subset of the files fairly easily. | 15:21.37 |
kens | Night all. | 16:00.41 |
Robin_Watts_ | alexcher, mvrhel2, ray_laptop, chrisl: Would any of you have a problem with the March meeting avoiding the dates I mention in my mail? | 16:06.28 |
chrisl | Robin_Watts_: no problem for me - I don't plan that far ahead..... ;-) | 16:07.25 |
mvrhel2 | I am out the week of March 5th as I am going to be an NC State University doing a short course based on my book but it looks like Miles wants the meeting to be near the end of Feb which works fine with me | 16:10.47 |
| Robin_Watts: sounds like you better get your vacation ordered | 16:15.02 |
Robin_Watts_ | ooh. euromeet. | 16:16.32 |
mvrhel2 | just added that to the calendar. I except a guided tour.... | 16:16.53 |
Robin_Watts_ | Anyone who can make it to the village can come on a tour of our brewery. | 16:17.17 |
| Powered by a Steam engine. | 16:17.27 |
henrys | liked the "chin-up bar" on xkcd, can't stop thinking about it. | 16:18.30 |
mvrhel2 | Robin_Watts: that sounds cool | 16:20.07 |
henrys | sabrina will be pretty excited with a euro trip | 16:21.04 |
| she wants to visit the area where Doc Martin is filmed I guess that's pretty far from Robin_Watts | 16:23.31 |
Robin_Watts_ | That's Ilfracombe, right ? | 16:23.53 |
henrys | port isaac? | 16:24.26 |
Robin_Watts_ | OK, small village in cornwall - I wasn't far off. | 16:26.16 |
| In UK terms that's a long way away. | 16:26.22 |
| In US terms it's probably as far as scott drives for lunch :) | 16:26.39 |
| I just got pspcl6 running under ubuntu on the Company I board. | 16:56.14 |
| A vanilla gcc build took 12:43 to process pdf_reference17.pdf at 600dpi. Using their special sauce compiler, it took 8:40ish. | 16:56.53 |
henrys | which device? | 16:57.41 |
Robin_Watts_ | henrys: see private message | 17:00.05 |
henrys | no I meant the ghostscript device - are you halftoning for example? | 17:00.49 |
Robin_Watts_ | oh. ppmraw | 17:01.17 |
| to /dev/null. Just wanted the simplest first attempt. | 17:01.39 |
| have to fetch helen. back in a bit. | 17:09.41 |
henrys | 1:20 on my mac as I recall that job is susceptible to slowness if memset is slow, blanking the page. | 17:13.24 |
ray_laptop | henrys: to compare with Robin_Watts's times you probably want to use BAND_LIST_STORAGE=memory. | 17:23.44 |
| which reminds me to make the change to make this dynamic, so that if it is built with 'file' we can still use 'memory' | 17:24.37 |
| anybody object to the parameter -sBandListStorage={memory | file} | 17:25.47 |
henrys | no objection here. Will you fix the other languages also or should I do that? | 17:29.17 |
| moving my build to memory bands would make the difference worse. | 17:30.26 |
| so something needs work in Robin_Watts's build. | 17:31.01 |
| ray_laptop:my wife's bday is cinco de mayo also ... | 17:32.46 |
ray_laptop | henrys: it should work with all languages (as long as the parameter is passed to the device). Does plmain need special exceptions for -s params ? (as opposed to -d) ? | 17:32.48 |
henrys | ray_laptop:no should be good. | 17:33.20 |
ray_laptop | henrys: interesting coincidence. I'll have to tell Karen | 17:33.24 |
henrys | you have to change the makefiles. | 17:33.50 |
ray_laptop | henrys: yes, if we want the BAND_LIST_STORAGE= to define the default, and 'memory' is given, we'd have to still include 'file'. Right now if 'file' is given, then we still always include 'memory' (for pattern clist). The cahnge should be common to all languages, however (in lib.mak) | 17:36.09 |
| I did this for cust 532 where they have an optional disk. They are happy with it. | 17:37.06 |
henrys | ah right you are for some reason I though we were selecting the clist device in the language makefiles, but we aren't - lib.mak only. | 17:45.18 |
ray_laptop | plmain _will_ need special handling for the -s params related to ICC profile stuff (these are NOT device parameters, and with PS are 'userparams'). plmain needs to call the gsicc_ functions with the strings | 17:57.58 |
| on my i5 laptop (2.4GHz) the PDF 1.7 Ref manual takes 126 sec. | 18:01.33 |
| with BAND_LIST_STORAGE=memory I see 117 sec | 18:07.17 |
mvrhel2 | good grief. bug 692537 is starting to appear to be an overprint problem. | 18:15.19 |
| I think all my remaining customer issues are overprint problems | 18:15.47 |
henrys | hmph lots of regression machines down | 18:24.32 |
| pcl lowres regression test seems suspiciously fast at 4 minutes or so. | 18:35.38 |
Robin_Watts_ | pcl highres is only 12. | 18:37.12 |
henrys | alright I guess it is okay. | 18:38.27 |
Robin_Watts_ | How many people do you need to have in a room before you have even odds that 2 of them will share a birthday? | 18:38.32 |
henrys | ah the birthday problem | 18:38.49 |
| http://en.wikipedia.org/wiki/Birthday_problem | 18:39.37 |
| so the odds are that somebody in the "artifex family" shares a birthday are quite high. | 18:41.28 |
| s/somebody/2 folks | 18:41.40 |
Robin_Watts_ | mvrhel2: Overprint is interesting for the planar case. Requires a getbits, | 18:42.34 |
mvrhel2 | looks like the color_component_map in the graphic state is getting out of sync after the device N color space is set up | 18:53.46 |
| when we set up the overprint information, it only finds Cyan | 18:54.15 |
Robin_Watts_ | So by default, bands go to file ? | 18:55.44 |
mvrhel2 | oh. looks like it is in gs_swapcolor_quick | 18:56.14 |
| hmm maybe not | 18:57.57 |
| darn | 18:57.59 |
| Robin_Watts: you mean for the clist? | 18:58.34 |
Robin_Watts_ | yes. | 18:59.00 |
henrys | Robin_Watts:for pcl/xps no for gs/ps/pdf yes | 18:59.15 |
mvrhel2 | oh. I was not aware of that for pcl/xps | 18:59.36 |
Robin_Watts_ | henrys: For ls ? | 18:59.49 |
henrys | see in the makefiles BAND_LIST_STORAGE=memory ... | 18:59.51 |
mvrhel2 | ah | 18:59.56 |
| ok | 18:59.58 |
henrys | memory | 19:00.17 |
| for pspcl6 | 19:00.27 |
Robin_Watts_ | OK. memory is better for me, as I only have a flash drive on this thing. | 19:00.46 |
henrys | do we have profiling tools for this beast? | 19:01.22 |
Robin_Watts_ | That's part of their special sauce bundle that I have on 30 day trial. | 19:01.44 |
| But Ubuntu should have oprofile too. | 19:02.05 |
henrys | I bet it is a slow memset on the embedded platform. | 19:04.10 |
| nothing interesting goes on with the manual it is mostly cached fonts... | 19:05.12 |
mvrhel2 | hmm. so the issue is how overprint is set up with the call in gs_pattern2_set_color. At that point in time there is no color map information, which we need if the shading color space is DeviceN | 19:09.19 |
| while it used to be stored in the graphic state, is was blown away by a restore | 19:09.34 |
| what a pain | 19:09.41 |
| so when we come in and set up the overprint settings we only get the index of zero which happens to be cyan | 19:10.45 |
| (since all the colorants are zero index due to how it is initialized) | 19:11.03 |
| I believe that gx_set_overprint_DeviceN should probably recheck the device and see what is valid | 19:12.27 |
| that should solve the issue. | 19:12.49 |
| lunch time now though | 19:12.52 |
henrys | so was this introduced with alex's shading change - to bracket it with save/restore? | 19:14.08 |
| Robin_Watts:Company 'M' says the sdk release is "in the process" | 19:24.46 |
mvrhel2 | henrys: I don't know. good question. when was that done? | 19:41.47 |
henrys | 9/6/2011 | 19:42.45 |
| which bug is this? | 19:43.03 |
mvrhel2 | Bug 692537 Marcos says the regression started with a commit that I did, so I don't think it could be alex's commit | 19:59.05 |
| ok. I may have a simple fix for this | 20:11.08 |
| ok. that fixed my simple file | 20:13.57 |
| let me see if the customer's file is fixed | 20:14.03 |
| yeah! that fixed their file too. now need to check clist mode | 20:15.56 |
| ok. time for a clusterpush | 20:16.58 |
| wow. why are there so many machines down? | 20:22.27 |
| off to run a few errands while this runs. | 20:23.21 |
| well, found out the piano tuner is coming in 10 minutes so no errand running. back to the salt mine | 20:29.35 |
| Forward 1 day (to 2011/09/27)>>> | |