| <<<Back 1 day (to 2013/08/13) | 2013/08/14 |
selim | hello | 00:25.45 |
ghostbot | hola | 00:25.45 |
selim | I found it what do you think : .IgnoreNumCopies <boolean> Some page description languages support a NumCopies parameter. This parameter instructs the device to ignore this, producing only one copy of the document on output. Note that some devices ignore NumCopies regardless because of limitation of the output format or the implementation. | 00:28.15 |
| anyone ? | 00:45.10 |
| j'ai essaye avec mais ca genere encore un seul PDF : : @C:\pdf.rsp -sOutputFile="C:\Documents and Settings\user\Bureau\%d.pdf" -d.IgnoreNumCopies=false -c save -f - | 00:51.03 |
| I tried to but it still generates a single PDF : @C:\pdf.rsp -sOutputFile="C:\Documents and Settings\user\Bureau\%d.pdf" -d.IgnoreNumCopies=false -c save -f - | 00:51.25 |
ray_laptop | well, at least I quashed bug 694514 :-/ | 01:26.52 |
mvrhel_laptop | ray_laptop: I think you were looking at the old bmcmp. | 02:29.35 |
| I do see plenty of issues still though | 02:29.48 |
| need to eat dinner now | 02:29.52 |
| then I will continue with this | 02:29.55 |
| ok. back to the salt mine | 03:09.33 |
| oh I see.... | 03:18.13 |
| doh | 03:18.14 |
mvrhel | trying another one. bbiab | 03:29.24 |
mvrhel_laptop | and that fixed it | 04:41.53 |
| ray_laptop: b8976 gives the fix for the transparency problem. | 04:44.20 |
| now to collect clist memory usage for the JEITA files | 04:44.40 |
| first though get all my run times computed with ray's awk program | 04:45.14 |
henrys | ah found the slowdown⦠introduced with the language fapi changes. | 04:56.41 |
mvrhel_laptop | ray_laptop: shoot do we have the J12 file? | 05:10.15 |
| weird. so with 8M of Bufferspace file renders slow. with 16M it is better than twice as fast and then at 32M is it back to being slow again | 05:22.21 |
| why would it do that | 05:22.34 |
henrys | does anyone know why ray_laptop's last commit changed clustercmd and clusterdatestamp? | 05:22.56 |
| strange. | 05:23.01 |
| mvrhel_laptop: how many bands do you get in each config? | 05:24.03 |
| it does sound odd | 05:24.20 |
| strcmp(pattern-clist, device) sigh ⦠oh well. | 05:25.41 |
mvrhel_laptop | with the bitcmyk device I dont see this. Only with the bitrgb device | 05:26.17 |
| let me check the # bands | 05:26.30 |
| hmm which -Z option will give me the band count | 05:30.34 |
| maybe L | 05:31.44 |
| nope. probably need to drag ray into this tomorrow | 05:32.51 |
| head is starting to hurt | 05:32.59 |
| after that transparency fix | 05:33.06 |
| done for night now | 05:33.09 |
henrys | oh sorry I was away, good night | 05:33.44 |
ray_laptop | all: I don't have any idea why the clister* files changed. I did what I usually do. | 06:03.06 |
kens | ray_laptop : they are (I htink) generated by the clusterpush script | 06:03.24 |
ray_laptop | anyone that can "undo" that is welcome to. I don't want to muck with it | 06:03.28 |
| kens: yeah, I know. And I do cluster push: toolbin/localcluster/gitpush.sh all the time. AFAIK I didn't do anything different this time | 06:04.25 |
kens | I noticed after I'd done a puch the other day that those files anded up in git gui as changed | 06:04.50 |
ray_laptop | kens: strange. Hope it doesn't screw anything up | 06:07.08 |
| I guess kens is getting coffee (or tea) | 06:08.12 |
| mvrhel: I saw your commit. Now to try and get into the customer's code | 06:08.45 |
| welcome back, kens | 06:10.06 |
kens | network interruption :( | 06:10.16 |
| BTW I don't like tea :-) | 06:10.48 |
| chrisl, follow link: | 06:50.13 |
| https://maps.google.com/maps?hl=en&ll=51.492159,-0.19092&spn=0.005291,0.013937&sll=51.492140,-0.193028&layer=c&cid=12502927659667388442&panoid=c9UMhWP_MWm9U0L48xEjYw&cbp=13,291.8,,0,18.86&gl=US&t=m&cbll=51.492132,-0.192862&z=17 | 06:50.13 |
| press double arrow at bottom of screen :-) | 06:50.22 |
chrisl | :-) I *knew* it was real! | 06:51.16 |
kens | Can't keep those GOogle spy mobiles out of anything.... | 06:51.38 |
chrisl | Mind you, those seats don't look very comfortable | 06:52.11 |
ray_laptop | Robin_Watts: do you have any idea about why the clustercmd and clusterdatestamp ended up getting "pushed"? I had done a git -a --amend that seemed to include them, and didn't realize they would be pushed as part of my patch | 07:03.28 |
| Robin_Watts: and is it a problem? and if so, what do we need to do? The cluster seems to be creaking right along, so maybe it isn't a problem | 07:04.49 |
| Robin_Watts: BTW, I am glad you are feeling better. | 07:06.16 |
IMAC | Hi, I have integrated MuPDF in my Android app and would like to know if MuPDF provides any PDF features like "Adding Notes", "Retaining Notes" and "Sharing Notes" or if I can add such features to my MuPDF reader present in my app? | 07:06.55 |
ray_laptop | Robin_Watts: at least it is better than having dengue. :-/ | 07:07.05 |
| I was too sick to fly home for 5 days :-( | 07:07.34 |
IMAC_ | Hi, I have integrated MuPDF in my Android app and would like to know if MuPDF provides any PDF features like "Adding Notes", "Retaining Notes" and "Sharing Notes" or if I can add such features to my MuPDF reader present in my app? | 08:34.30 |
tor7 | integrated how? the mupdf android app has support for adding annotations, if that's what you mean? | 08:40.36 |
IMAC_ | <tor7> Could you please tell me what kind of annotation support does MuPDF provide? | 10:00.07 |
Robin_Watts | IMAC_: We support adding/editing highlight/underline/strikeout currently. | 10:01.21 |
| paulgardiner is adding support for more types (including digital signatures). | 10:01.46 |
| I think freetext may be coming soonish. | 10:02.03 |
| and we do ink, if memory serves. | 10:02.08 |
| but this is still a work in progress. | 10:02.17 |
| tor7: ping | 10:09.00 |
tor7 | Robin_Watts: pontg. | 10:09.08 |
Robin_Watts | tor7: 2 small patches on robin/master. both untested as yet, but you may have thoughts on them. | 10:09.32 |
paulgardiner | We also do test markup: strikethrough, underline and highlight. | 10:09.33 |
tor7 | Robin_Watts: we should put the android bugfix on a 'release' branch | 10:10.20 |
Robin_Watts | tor7: The curl commit from yesterday uses //. I thought we should avoid that ? | 10:10.27 |
tor7 | and the curl fix, but I want to try building on windows with mingw as well | 10:10.32 |
| and fix the // as well | 10:10.40 |
paulgardiner | IMAC_: freetext is working, but restricted to base 14 fonts | 10:10.42 |
Robin_Watts | tor7: I'm gonna quickly test these things, then I have to look at this P1 gs customer bug. | 10:11.21 |
| Not that I have a 7 button mouse of course... | 10:11.53 |
tor7 | Robin_Watts: *proud owner of a 3 button mouse without a scroll wheel* | 10:12.22 |
| have you looked at his second patch, that does the page flipping on j/k as well? | 10:12.58 |
Robin_Watts | tor7: I have several of those, but they aren't PS/2 | 10:13.06 |
| I haven't. | 10:13.11 |
tor7 | aren't ps/2? | 10:13.24 |
| I have an optical usb 3-button mouse without a scroll wheel. rare beasts. | 10:13.38 |
Robin_Watts | I've got a couple of original Acorn 3 button mice with aftermarket "heavy" mouseballs in somewhere. Cutting edge. | 10:14.36 |
| but they are a strange acorn specific interface. | 10:14.59 |
tor7 | ah, kind of like the old apple mice with the Apple Desktop Bus connector? | 10:16.26 |
Robin_Watts | yeah. | 10:16.52 |
IMAC_ | Robin_Watts: Ok. Can so we can add/edit the PDF right? | 10:18.48 |
Robin_Watts | IMAC_: In the android app, you open a PDF, then click the appropriate menu item to go into the annotation mode you need. | 10:19.30 |
| then (for the text ones) you select some text, and click a button to apply the changes. | 10:19.50 |
IMAC_ | Robin_Watts: Ok. Currently in my MuPDF Reader after I opened the PDF I'm having 5 buttons at the top ie; 1: To highlight the links 2. An arrow 3. Getting the contents list 4. An icon with some arrows pointing downwards 5. Search | 10:22.39 |
| and at the bottom I'm having a Seek bar using which I can scroll to a particular page | 10:23.15 |
Robin_Watts | My right hand icon is something like (v) | 10:24.18 |
IMAC_ | Are these the Annotation modes ur talking about? | 10:24.47 |
Robin_Watts | (v) changes the top bar into a second set of options. | 10:25.12 |
| Including printing, copy to clipboard, and (+) annotations. | 10:25.33 |
| (+) Then takes you to a mode to do highlight, underline, strikeout, ink. | 10:26.03 |
IMAC_ | Ok. I don't seem to be having a "V" icon | 10:28.59 |
| The 5 icons that I have are the above ones | 10:29.21 |
| Can I create a PDF file using MuPDF Reader? | 10:30.05 |
Robin_Watts | IMAC: create files using MuPDF Reader? No. | 10:34.29 |
| reboot | 10:34.34 |
| Where does local.properties keep going? | 11:23.35 |
| oh, ignore me. | 11:25.42 |
| tor7: OK, the mupdf android fix tests out. | 11:41.43 |
| and the linux scroll wheel change doesn't cause any bad effects for me. | 11:42.01 |
| so I'm leaving them to you and going to look at the P1 gs bug. | 11:42.13 |
kens | Robin_Watts : ping | 11:49.37 |
Robin_Watts | Just replying to Phil now. | 11:49.48 |
kens | AH, tehn I won't bother, thanks | 11:50.00 |
tor7 | chrisl: Robin_Watts: have any of you had any luck installing vs2005 on win8? | 12:20.47 |
kens | not tried | 12:20.55 |
| I don't even have 2005 on WIndows 7 | 12:21.06 |
paulgardiner | tor7: I have | 12:21.08 |
tor7 | paulgardiner: any special tricks you had to do? | 12:21.24 |
paulgardiner | You need all the windows 7 fixes, but besides that should be okay | 12:21.29 |
| Ah. Don't install the SQL thingy | 12:21.40 |
tor7 | yeah, I've pared it down to just the C++ compiler and win32 SDK | 12:21.55 |
| got a failure on .NET 2.0 but restarting the install seems to work better | 12:22.07 |
paulgardiner | Sorry. Can't remember exactly what it is. One of the install check boxes needs to be left unchecked | 12:22.10 |
| tor7: I've installed it several times and always had to restart I believe | 12:22.40 |
tor7 | and eek, win8 is awkward to use. | 12:22.43 |
kens | Yes, I think you'll have gathered that form my comments :-) | 12:22.58 |
tor7 | not to mention how ugly the old win7 icons look on the "flat" metro style window title bars | 12:23.14 |
kens | Well, IMO the new Windows 8 flat menu bars are dreadful | 12:23.35 |
paulgardiner | I'm slightly embarrassed tp say that I've sort of gotten used to Windows 8, possibly because I stick most of what I use on the taskbar | 12:23.59 |
| Ah right. I never use the flat menus | 12:24.12 |
tor7 | kens: if they'd redone all the desktop icons in the new flat style it wouldn't be so terrible. but the way they've gone and hidden every single thing you want to do behind five steps of the tile start menu thing | 12:24.19 |
| urgh. | 12:24.22 |
kens | TO be honest, I just dislike the entire look and feel | 12:24.37 |
tor7 | but yeah, it's dreadful. win7 is going to stay here a long time... | 12:24.57 |
paulgardiner | Oh yeah. I certainly dislik the entire look and feel, but I was expecting that. :-) | 12:25.01 |
tor7 | kens: there is one look and feel that's worse: iOS 7. | 12:25.15 |
kens | Every timeI start it up I'm reminded anew how much I hate WIndows 8 | 12:25.25 |
| Fortunately I don't have to deal with iOS 7 :-) | 12:25.45 |
tor7 | I'm very happy with my new linux machine setup (no desktop environment anywhere in sight) | 12:26.42 |
Robin_Watts | tor7: Eh? No desktop? | 12:48.29 |
tor7 | no gnome, no kde, just a .xsession | 12:48.55 |
| tried them all. hated them all. doing what I did in the '90s ... window manager and some xterms | 12:49.44 |
chrisl | tor7: what window manager are you using? | 12:52.10 |
tor7 | Robin_Watts: mupdf.com/download/ has a 1.3 source tarball. could you compile that on vs2005 for me? | 12:52.18 |
| chrisl: evilwm | 12:52.27 |
Robin_Watts | tor7: Ah, right. | 12:52.39 |
tor7 | I don't like title bars | 12:52.41 |
| Robin_Watts: it's going to take a while to patch my vs2005 back up | 12:52.54 |
chrisl | tor7: that's a bit too minimalist for me - I like *small* title bars..... | 12:53.16 |
Robin_Watts | tor7: Sure. Give me a while to at least have a crack at this interpolation thing. | 12:53.18 |
tor7 | Robin_Watts: no rush. in fact, just prod me tomorrow if I haven't got the win32 binary uploaded by then | 12:53.46 |
| chrisl: ctl+alt+i to show the title bar :) | 12:54.06 |
| chrisl: but yes, it's definitely not newbie friendly in its minimalism | 12:54.28 |
| chrisl: then I have pidgin, sylpheed, firefox and xterm running. gmrun to start apps and xbindkeys for some extra key bindings. | 12:55.21 |
chrisl | tor7: I might give it a go next time I install. I keep hitting head-banging annoyances on the desktop "usual suspects" | 12:55.46 |
tor7 | chrisl: well, any "old" window manager will usually do | 12:56.10 |
| icewm looks like win95 with a taskbar, I used that one a lot back in the days. | 12:56.51 |
chrisl | tor7: I certainly don't want a Windows clone..... | 12:57.18 |
| tor7: I might go back to WindowMaker - it defaults to rather huge, clunky decorations, but (at least used to be) configurable to a less screen area hungry setup | 12:59.45 |
tor7 | chrisl: icewm's redeeming feature is its rather nice taskbar, but I'm quite happy without one (hence evilwm and an xclock to keep track of time) | 13:00.25 |
| I never got the hang of WindowMaker | 13:00.35 |
chrisl | It does seem that too many window managers have disappeared down the "eye candy over function" route :-( | 13:02.29 |
tor7 | chrisl: what's the other useful one, openbox, fluxbox, blackbox? somethingbox anyway. that one didn't go overboard with eye candy over function and is still relatively conventional. | 13:04.11 |
chrisl | They are all variations on the same theme or forks or the same thing, aren't they. Might be worth a look | 13:05.26 |
fdncred | any idea why my mupdf git pull won't pull the thirdpary submodules? | 14:09.35 |
Robin_Watts | fdncred: What error are you seeing? | 14:11.05 |
| have you done "git submobule update --init" at the top level? | 14:11.19 |
| have you done "git submodule update --init" at the top level? | 14:11.29 |
fdncred | fatal: Needed a single revision | 14:12.28 |
Robin_Watts | For what command ? | 14:13.07 |
fdncred | $ git submodule update --init Submodule 'thirdparty/curl' () registered for path 'thirdparty/curl' Submodule 'thirdparty/freetype' () registered for path 'thirdparty/freetype' Submodule 'thirdparty/jbig2dec' () registered for path 'thirdparty/jbig2dec' Submodule 'thirdparty/jpeg' () registered for path 'thirdparty/jpeg' Submodule 'thirdparty/openjpeg' () registered for path 'thirdparty/openjpeg' Submodule 'thirdparty/zlib' () regi | 14:13.55 |
| sorry | 14:14.10 |
| they all say registered for the --init submodule | 14:14.20 |
| then there's a unable to find current revision for curl | 14:14.35 |
| i'm sure the problem is i'm a git-idiot (gitiot?) | 14:15.09 |
Robin_Watts | fdncred: Try deleting the contents of thirdparty and retrying ? | 14:15.47 |
fdncred | fatal: unable to connect to git.ghostscript.com: git.ghostscript.com[0: 184.73.189.105]: errno=No error Unable to fetch in submodule path 'thirdparty/curl' | 14:16.29 |
| that was from the command you suggested | 14:16.50 |
Robin_Watts | fdncred: That would seem to imply that you haven't got a thirdparty/curl dir? | 14:17.10 |
fdncred | i deleted everything in thirdparty so there was nothign. now there is only a curl dir that is empty except for .git | 14:17.54 |
Robin_Watts | retry the deletion. confirm that it's really gone. | 14:18.24 |
| then try git submodule update --init again. | 14:18.40 |
fdncred | same outcome - everything says registered for path thirdparty/blah | 14:19.35 |
| then it pukes | 14:19.40 |
| with unable to connect to git.ghostscript.com | 14:19.50 |
| so for some reason i can't connect - is there a way to make that http versus git (assuming it's using git protocol) | 14:20.44 |
Robin_Watts | git log -1 | 14:20.47 |
| What commit are you on? | 14:20.57 |
fdncred | $ git log -1 commit 00fbe2446c7ca2c291b785da99f6e0b293574001 Author: Paul Gardiner <paulg.artifex@glidos.net> Date: Tue Aug 13 13:44:40 2013 +0100 | 14:21.19 |
| if i use TortoiseGit it says "git.exe submodule update --init" | 14:22.12 |
tor7 | fdncred: two ways, you can "cd thirdparty/curl" and do "git fetch http://..." from there | 14:22.14 |
| or edit the .gitmodules and rerun "git submodule init" | 14:22.33 |
fdncred | and everything is registered but then "fatal: Need a single revision" | 14:22.44 |
Robin_Watts | fdncred: Are you sure your .gitmodules file is intact ? | 14:23.03 |
fdncred | i'm looking at .gitmodules now | 14:23.20 |
| has 18 lines | 14:23.27 |
tor7 | fdncred: iirc that error can happen when the project references a submodule commit that doesn't exist in the submodule repo | 14:23.36 |
fdncred | i think the problem is the urls start with git and not http | 14:23.42 |
tor7 | usually because the submodule hasn't or can't be updated | 14:23.47 |
| fdncred: you sitting behind a firewall that blocks the git:// port? | 14:24.14 |
fdncred | i think git:// may be blocked - it's my guess because it looks like i just can't get through | 14:24.34 |
tor7 | fdncred: in the "url = git://..." lines, change git to http | 14:24.36 |
fdncred | doing that now | 14:24.45 |
tor7 | if you "cd thirdparty/curl" and do "git fetch http://git.ghostscript.com/thirdparty/curl.git" does that work? | 14:24.59 |
fdncred | my .gitmodules looks like this now | 14:25.45 |
| [submodule "thirdparty/jbig2dec"] path = thirdparty/jbig2dec url = http://git.ghostscript.com/jbig2dec.git | 14:25.45 |
| except those are on two lines | 14:25.54 |
| 3 lines | 14:26.07 |
| and i do a "$ git submodule update --init" i still get unable to connect to git.ghostscript.com | 14:26.56 |
| trying to cd to thirdparty/curl and try that git fetch | 14:27.21 |
Robin_Watts | fdncred: If you change the URL in the .gitmodules you'll need to wipe thirdparty and retry, I think. | 14:27.44 |
fdncred | ok, i wiped my thirdparty dir again | 14:28.31 |
| cd and did the git fetch like tor7 said | 14:28.45 |
| $ git fetch http://git.ghostscript.com/thirdparty/curl.git fatal: http://git.ghostscript.com/thirdparty/curl.git/info/refs not found: did you run git update-server-info on the server? | 14:29.00 |
| was i supposed to do a submodule init first? | 14:29.15 |
tor7 | fdncred: try the git fetch again, I think we forgot to set up a git hook when creating the curl repository | 14:30.32 |
fdncred | back - friggin connection dropped | 14:35.31 |
| i just did a git submodule init and it seems to have worked | 14:36.11 |
| now i need to update them | 14:36.21 |
Robin_Watts | fdncred: Thanks. I think it was a problem at our end. | 14:36.27 |
fdncred | ok | 14:36.46 |
| i just did a git submodule foreach git pull | 14:37.38 |
| nothing happened | 14:37.42 |
| how do i update my submodules? | 14:37.50 |
tor7 | git submodule update | 14:37.56 |
| that will do a "git fetch" if the parent project uses a submodule at a commit that it doesn't currently have | 14:38.24 |
| "git submodule foreach git remote update" is what you want rather than pull | 14:39.06 |
| since a submodule will be using detached HEAD pulling doesn't make sense | 14:39.26 |
fdncred | dang it | 14:40.42 |
| still puking | 14:40.46 |
| $ git submodule foreach git remote update Entering 'thirdparty/curl' Fetching origin fatal: unable to connect to git.ghostscript.com: git.ghostscript.com[0: 184.73.189.105]: errno=No error error: Could not fetch origin Stopping at 'thirdparty/curl'; script returned non-zero status. | 14:40.59 |
tor7 | fdncred: do a "cd thirdparty/curl" and "git remote -v" and paste the output here | 14:41.45 |
| I have a suspicion that the remotes haven't been updated | 14:42.05 |
fdncred | $ git remote -v origin http://git.ghostscript.com/thirdparty/curl.git (fetch) origin http://git.ghostscript.com/thirdparty/curl.git (push) | 14:42.32 |
tor7 | rigel ~/src/mupdf/thirdparty/curl $ git fetch http://git.ghostscript.com/thirdparty/curl.git | 14:43.03 |
| From http://git.ghostscript.com/thirdparty/curl | 14:43.03 |
| * branch HEAD -> FETCH_HEAD | 14:43.03 |
| worked for me... odd that it doesn't work for you | 14:43.10 |
fdncred | $ git fetch http://git.ghostscript.com/thirdparty/curl.git From http://git.ghostscript.com/thirdparty/curl * branch HEAD -> FETCH_HEAD | 14:44.17 |
| but there are no files in the curl dir other than .git | 14:44.33 |
henrys | kens:did you forward this abdo guy to scott and miles. | 14:51.39 |
| ? | 14:51.55 |
kens | henrsy no not yet | 14:54.55 |
| henrys | 14:54.59 |
henrys | I just did it. | 14:56.50 |
kens | Thanks henry | 14:57.05 |
kens | is wrestling with EndPage | 14:57.34 |
chrisl | kens: We have a bug with EndPage? | 14:58.39 |
kens | No, its more complex than that | 14:58.48 |
chrisl | :-) | 14:59.04 |
kens | pdfwrite of course | 14:59.19 |
chrisl | Not related to henrys blank pages insertion thing, is it? | 14:59.49 |
kens | Not exactly, someone who doesn't want a page | 15:00.04 |
chrisl | Oh, by returning false, or something? | 15:01.20 |
kens | well the problem is (I think) that caption.ps has a bogus EndPage, but proving it is 'interesting' | 15:02.17 |
| I think its not caring for device deactivation | 15:02.49 |
fdncred | well i have curl now but none of the other thirdpary submodules... grrrrr | 15:02.57 |
henrys | chrisl: let me know if you want me to look at the PCL performance problem. I did profiling and it looks as though we are spending a lot of time purging cache entries. Particularly that linear probe for the character. | 15:06.53 |
chrisl | henrys: yes, I know why that's happening, and I can fix it fairly easily. *But* it doesn't really improve things much :-( | 15:13.05 |
| henrys: if you could point me to the mail subject line, so I can get the full test file, that would be good, though | 15:20.34 |
henrys | Fw: performance issue for GhostPCL 9.05 vs 8.70 | 15:23.18 |
| chrisl:you can just run the simplified file repeatedly... | 15:23.59 |
chrisl | henrys: so, the full file is nothing more than the same thing repeated? | 15:24.48 |
henrys | from what I looked at, it is 400 pages I didn't examine all the pages. | 15:25.42 |
chrisl | Ugh, download expired..... | 15:26.44 |
| henrys: could you upload the full test to casper? I'd want to test with it before getting back to the customer, anyway | 15:29.29 |
henrys | will do | 15:30.05 |
chrisl | Thanks | 15:30.49 |
henrys | zip in my home dir now, chrisl | 15:32.06 |
| chrisl: do you know about that "clusterdatestamp" thing with rays commit. It looks harmless enough but it screws up the patch. | 15:35.04 |
| or Robin_Watts | 15:37.10 |
Robin_Watts | henrys: I suspect that ray Ctrl-C'd out of a "git cluster" at some point. | 15:37.10 |
| Those files get created at the start of a git cluster, and should be removed at the end. | 15:38.04 |
| I'll do a commit that removes them. | 15:38.19 |
chrisl | henrys: I've got that test file, now, thanks | 15:39.21 |
Robin_Watts | ray_laptop: Do you understand the non-dropout downscaler? | 16:28.33 |
| I thought for a moment I understood it, but 1) if I understand it, it's broken, and 2) I have a tiny change that almost fixes it, and it really shouldn't. | 16:30.30 |
ray_laptop | Robin_Watts: sorry. wasn't paying attention | 16:44.54 |
| Robin_Watts: I wrote it quite a while ago | 16:45.10 |
Robin_Watts | ray_laptop: Ah! | 16:45.22 |
| then you're exactly the person I should be speaking to :) | 16:45.30 |
ray_laptop | Robin_Watts: Is that where the image squashing is coming from ? | 16:45.44 |
Robin_Watts | It is. | 16:45.50 |
ray_laptop | or the person you should turn the bug over to. | 16:46.01 |
Robin_Watts | sidscale.c idownscale_x | 16:46.04 |
| AIUI, this routine copies from src into tmp. | 16:47.10 |
ray_laptop | OK. I'm looking at it. Should I fire up a debugger with cust 531's file ? | 16:47.16 |
Robin_Watts | Let's speak conceptually first. | 16:47.29 |
| If it's the first time we're copying for a given line, we copy src into tmp. | 16:47.47 |
ray_laptop | Robin_Watts: OK | 16:48.00 |
Robin_Watts | actually, that's not quite right. Let me restate that. | 16:48.26 |
ray_laptop | oops. someone at the door. | 16:48.37 |
Robin_Watts | We run through the source pixels. | 16:48.46 |
| We have to "copy" 1 or more source pixels into tmp, thus downscaling. | 16:49.26 |
ray_laptop | ok, back | 16:50.19 |
Robin_Watts | Each "group" of such pixels is given by the pixels that lie between the boundaries of the bresenham. | 16:50.24 |
ray_laptop | OK so far | 16:50.37 |
Robin_Watts | i.e. in this case we're downscaling by roughly a factor of 5, so we "copy" pixels 0 to 4 into the first tmp pixel, then pixels 5 to 9 etc. | 16:51.02 |
ray_laptop | Robin_Watts: right, where 'copy' involves keeping the darker color | 16:51.42 |
Robin_Watts | Right. | 16:51.48 |
| There is some stuff in there with "polarity_additive" to say whether it's darker or lighter. | 16:52.08 |
| but that seems wrong to me. | 16:52.16 |
ray_laptop | Robin_Watts: this case (monochrome) s/b additive. Thus a darker color is closer to 0 | 16:53.24 |
Robin_Watts | The logic in the loop is: For each pixel in the row { do the first pixel of each group; do the rest of the pixels in the group; } | 16:53.51 |
| The polarity code is in the first pixel case, but is missing from the 'rest of the pixels in the group' case as far as I can see. | 16:54.20 |
ray_laptop | Robin_Watts: you are right. the 'while (i < endx)' doesn't check polarity. It works for additive | 16:55.15 |
Robin_Watts | ok, so that's a bug? | 16:55.58 |
ray_laptop | Robin_Watts: I think so. We'd have to test with subtractive polarity. | 16:56.55 |
Robin_Watts | So I now think I broadly understand the code. | 16:57.16 |
ray_laptop | OK | 16:57.20 |
Robin_Watts | But I did a quick test to check whether it was the output of this function that was going wrong, and I cannot explain the results I got. | 16:58.02 |
ray_laptop | I'm curious if fixing that will show any cluster diffs | 16:58.20 |
Robin_Watts | Can you find the test file, and do a quick test for me please? | 16:58.29 |
ray_laptop | the customer's file ? | 16:58.40 |
Robin_Watts | For sanity, the following command is much faster than theirs: | 16:58.41 |
| gs/debugbin/gswin32c.exe -o out.tif -r72 -sDEVICE=tiffpack -dDOINTERPOLATE squish.pdf | 16:58.54 |
| (i.e. we don't need 300dpi to see the issue) | 16:59.14 |
| If you run that on their test file, you should see the squishing effect they are talking about. | 16:59.34 |
| Note that the image in the file is in landscape mode, so the 'scanlines' we are working on here are actually "scancolumns" in the output bitmap. | 17:00.11 |
ray_laptop | is 'squish.pdf' the file from the bug ? | 17:00.21 |
Robin_Watts | ray_laptop: It is. | 17:00.27 |
ray_laptop | OK. I have the file. renaming it to match your command line... | 17:00.52 |
| rebuilding... | 17:03.42 |
Robin_Watts | while we rebuild, an idle observation. | 17:04.18 |
| if ((polarity_additive && (*pp < *tp)) || | 17:04.28 |
| (!polarity_additive && (*pp > *tp)) ) | 17:04.30 |
| *tp = *pp; | 17:04.32 |
| could be equivalently written: if (polarity_additive ^ (*pp >= *tp) *tp = *pp; | 17:05.08 |
ray_laptop | OK. I ran it and see the squished output. | 17:05.21 |
Robin_Watts | OK, now here's the thing I really don't understand. | 17:05.37 |
| Change: if (firstline) *tp = *pp; | 17:05.50 |
| to be if (firstline) *tp = 128; | 17:06.05 |
| then rebuild and rerun. | 17:06.17 |
| I did that because I wanted to see if the problem was that this function was not setting enough of the buffer, or whether it was that this was setting all the buffer, but only part of it was actually being plotted out. | 17:07.28 |
| What effect do you see? | 17:08.07 |
ray_laptop | yuck. It's mostly gray but bits of the drawing show through as white or black | 17:09.24 |
| and the image is not squished looking | 17:09.45 |
Robin_Watts | Right. Why does that stop the squishing ? | 17:10.29 |
ray_laptop | uhh... | 17:10.59 |
Robin_Watts | The color change I expected :) | 17:11.48 |
ray_laptop | OK. Why don't I take a look with the debugger ? | 17:16.14 |
Robin_Watts | ray_laptop: OK. I'll keep fiddling here. | 17:16.56 |
| Memento gives it a clean bill of health. | 17:17.14 |
| I wonder if it's not this code at all. | 17:23.40 |
ray_laptop | Robin_Watts: I don't think it is. I look at the 'tmp' buffer and all 2592 bytes are ff | 17:31.42 |
| I'm going up a level... | 17:32.03 |
Robin_Watts | ray_laptop: Yeah, I think it may be something later on that's coping badly with being given w=1 h=many. | 17:32.32 |
ray_laptop | I'll keep looking... | 17:33.21 |
Robin_Watts | yeah, if I set the output buffer to be full of ff's I see the output squished. | 17:35.25 |
| There is a special case wrong somewhere I bet. | 17:35.33 |
| ray_laptop: OK, so this looks like it's not "your fault" any more. Sorry to have bothered you. Feel free to leave it to me to look at again. | 17:36.52 |
henrys | I'm probably going to complain and whine constantly until marcosw returns. | 17:53.51 |
ray_laptop | Robin_Watts: OK. | 17:54.31 |
| All: Michael noticed that J12 wasn't in the JEITA files in ~ray/public. I just updated it to include J12_acrobat.pdf | 17:55.59 |
| mvrhel: (for the logs) OK. Updated tarball is there now with J12 | 17:56.35 |
Robin_Watts | ray_laptop: Why not in tests_private/pdf/jeita? | 17:57.15 |
| oh, ass. I see the problem. | 18:00.22 |
ray_laptop | mvrhel: debugbin/gswin32c -dNOPAUSE -dBATCH examples/tiger.eps -c "(a) true .setdebug 65535 string (a) false .setdebug" | 18:28.59 |
| Robin_Watts: I'll leave adding them to the regression up to Marcos. I don't think they really add any value | 18:29.46 |
Robin_Watts | ray_laptop: Adding them to the tests_private does not add them to the regression tests. | 18:30.05 |
ray_laptop | Robin_Watts: not in the clist code I hope | 18:30.22 |
Robin_Watts | It just means they are in the same consistent place as all the rest of our test files and hence easy to find. | 18:30.26 |
| ray_laptop: No, it's to do with the horrible macroised code. | 18:30.47 |
| The image code accumulates scanlines and then sends them at once. | 18:31.04 |
| the landscape image code does the same thing, but then sends those scanlines as scancolumns. | 18:31.42 |
| And that breaks down horribly of course when working with bpp < 8. | 18:31.58 |
ray_laptop | have to pick up my son from tennis. bbiaw | 18:34.47 |
mvrhel | hi ray_laptop: found my problem. I was running this in a shell script and the quotes were causing me issues. adding eval fixed my issue | 19:56.56 |
| so the max = value is the maxed used | 19:57.27 |
| and will that be across all the pages in a multi-page document? | 19:57.41 |
ray_laptop | mvrhel: yes, all pages. | 19:58.47 |
mvrhel_laptop | ok thanks | 19:59.19 |
ray_laptop | mvrhel: If you want something for individual pages, then you'd have to run it a page at a time with -dFirstPage=# -dLastPage=# | 19:59.35 |
mvrhel_laptop | right | 19:59.41 |
ray_laptop | c:/Artifex/bugs/kyocera/PDF_Performance/J11_acrobat.pdf has 12 pages | 20:03.13 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J12_acrobat.pdf has 20 pages | 20:03.14 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J1_acrobat.pdf has 1 page. | 20:03.16 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J2_acrobat.pdf has 2 pages | 20:03.17 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J3_acrobat.pdf has 2 pages | 20:03.19 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J4_acrobat.pdf has 1 page. | 20:03.20 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J5_acrobat.pdf has 1 page. | 20:03.22 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J6_acrobat.pdf has 1 page. | 20:03.23 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J7_acrobat.pdf has 1 page. | 20:03.25 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J8_ghostscript.pdf has 1 page. | 20:03.26 |
| c:/Artifex/bugs/kyocera/PDF_Performance/J9_acrobat.pdf has 5 pages | 20:03.28 |
mvrhel | running J12 now on the raspberry pi | 20:04.42 |
ray_laptop | mvrhel: to use bash with a numeric loop, use: x=0 ; while [ $x -le 5 ] ; do echo $x ; x=$(( $x + 1 )) ; done | 20:05.21 |
| where '5' is whatever limit you have, and the "echo $x" can be running gs | 20:06.08 |
mvrhel | right. I have a loop now that goes through the files. I am just going to do the max on the pages for memory though. I am feeling like I am running of time right now | 20:06.14 |
ray_laptop | mvrhel: just wanted to mention it in case it was useful | 20:06.42 |
mvrhel | ok thanks | 20:06.47 |
ray_laptop | mvrhel: was the bitrgb variation with -dBufferSpace with all files, or just some ? | 20:08.41 |
mvrhel_laptop | let me look hold on | 20:10.26 |
| some of the other docs follow the same patter | 20:11.50 |
| ray_laptop: that is, 16 case is quite faster for bitrgb than the 8 or 32 cases | 20:12.16 |
| like I said, I don't see this on my laptop though | 20:12.35 |
ray_laptop | I don't recall which doc was the one you mentioned | 20:12.36 |
mvrhel_laptop | well J9 was the one I mentioned ealier | 20:12.49 |
| but it is true with the following | 20:12.57 |
ray_laptop | OK thanks. | 20:12.58 |
mvrhel_laptop | J10 | 20:13.16 |
| J11 | 20:13.31 |
ray_laptop | I am going to buy a Raspberry and have a look. Probably no results until you are done with your trip, but ... | 20:13.48 |
mvrhel_laptop | ok good. | 20:13.58 |
ray_laptop | mvrhel: was there a particular configuration Pi you got (other than just a generic B) ? | 20:14.19 |
| mvrhel: was there a particular configuration Pi you got (other than just a generic B ) ? | 20:14.25 |
mvrhel_laptop | that was the one | 20:14.29 |
ray_laptop | it's sort of an unexpected clist banding issue that I want to understand -- besides it gives me an excuse to buy a Pi and play with it | 20:15.53 |
| I _should_ be able to do a profile build and compare | 20:16.23 |
mvrhel_laptop | I will be very interested to see what you find | 20:18.23 |
| Getting this spread sheet together for you to look over | 20:18.44 |
| J12 takes awhile to run... | 20:19.29 |
ray_laptop | mvrhel: 20 pages. I'm not surprised. BTW, I saw a note about overclocking the Pi to 1.2 GHz (when looking at additional heatsinks) | 20:23.11 |
mvrhel_laptop | whoa | 20:23.21 |
ray_laptop | mvrhel: what clock rate are you running ? | 20:23.38 |
mvrhel_laptop | the std. 700 | 20:23.47 |
| or is it 750 | 20:23.53 |
| hold on | 20:23.55 |
| 700 | 20:24.03 |
ray_laptop | mvrhel: I'll play with overclocking on the one I get. We don't want you to burn yours up before the trip :-) | 20:24.31 |
mvrhel_laptop | right | 20:24.38 |
ray_laptop | OK. I ordered one, with the p/s and the heatsinks (for overclocking) | 20:31.13 |
| I should have done so previously | 20:31.40 |
| oh, oh. The some of the cust 532 JEITA PS files that used to run with 9.06 fail with Error: /invalidfont in --.FAPIBuildChar-- | 20:44.56 |
| bug for chrisl I guess | 20:45.08 |
mvrhel_laptop | ray_laptop: so the bitcmyk device is also a bit faster with the 16M buffer space | 20:51.36 |
| anyway, figuring out memory use for J9, J11 and J12 and then I will get this to you | 20:53.32 |
Robin_Watts | mvrhel_laptop, ray_laptop: Do you have oprofile on your rpi image? | 21:31.59 |
| (which oprofile or which ocontrol) | 21:32.10 |
mvrhel_laptop | no | 21:32.37 |
| new high speed cable modem just arrived. | 21:35.29 |
| have to hook it up tonight | 21:35.34 |
Robin_Watts | I can do profiles on the beagleboard, which should be similar, I'd hope. | 21:36.06 |
| oh, opcontrol, sorry. | 21:36.37 |
| looks like oprofile isn't working on the pi yet. | 21:38.01 |
| but supposedly https://code.google.com/p/gperftools/ might work. | 21:38.17 |
ray_laptop | holy cow. Using git bisect I am getting a LOT of segfaults as it goes through :-( (running J3.ps) | 22:05.59 |
| trouble is I don't know which way to go (good / bad) and skip doesn't really help | 22:06.42 |
| I guess Marcos is luckier with the fixes he's bisecting (or I'm doing something wrong) | 22:09.27 |
Robin_Watts | ray_laptop: skip is the right thing to use. | 22:11.03 |
| Marcos has cached binaries so he doesn't need to rebuild, hence it is a much faster process. | 22:11.19 |
ray_laptop | that's a LOT of binaries! | 22:14.51 |
Robin_Watts | ray_laptop: Indeed. | 22:15.00 |
| I have a fix for the squished image. | 22:15.32 |
| It's not nice, but it works. | 22:15.41 |
| Will tidy it up tomorrow. | 22:15.49 |
| I strongly suspect that the correct thing to do is just to recode the routine to avoid the macro porn. | 22:16.10 |
mvrhel_laptop | so with the file based clist, times were a little bit slower | 22:17.19 |
| trying immediate mode with hug MaxBitmap | 22:18.29 |
| s/hug/huge/ | 22:18.35 |
ray_laptop | Robin_Watts: git bisect stopped working for me. I restarted it after a git bisect reset, but now it's "stuck". After git bisect bad it still says the same thing: This means the bug has been fixed between bbfbdf7574e6ac46b65b76b06dea6a4e94d35f31 and [76b4193e5d8cb280f4795b0e959657ad695ad54d]. | 22:23.36 |
mvrhel_laptop | ray_laptop: so it is still using the clist | 22:24.17 |
ray_laptop | Robin_Watts: | 22:24.55 |
| git bisect start | 22:24.57 |
| # bad: [c07d6c816052686d29efdee5e94567066baaf75a] Remove stray cluster products. | 22:24.58 |
| git bisect bad c07d6c816052686d29efdee5e94567066baaf75a | 22:25.00 |
| # good: [76b4193e5d8cb280f4795b0e959657ad695ad54d] Revise release date and product string for 9.07 release. | 22:25.01 |
| git bisect good 76b4193e5d8cb280f4795b0e959657ad695ad54d | 22:25.03 |
| # bad: [bbfbdf7574e6ac46b65b76b06dea6a4e94d35f31] Bug 693284: Prevent read access violation. Check whether there's enough data. | 22:25.04 |
| git bisect bad bbfbdf7574e6ac46b65b76b06dea6a4e94d35f31 | 22:25.06 |
| # bad: [bbfbdf7574e6ac46b65b76b06dea6a4e94d35f31] Bug 693284: Prevent read access violation. Check whether there's enough data. | 22:25.07 |
| git bisect bad bbfbdf7574e6ac46b65b76b06dea6a4e94d35f31 | 22:25.09 |
| when it was working it was giving me: | 22:26.32 |
| Bisecting: 9 revisions left to test after this (roughly 3 steps) | 22:26.33 |
| [a3d00daf5f9abb1209cb750a95e23bc6951c1c63] pdfwrite - convert non-UTF-16BE doc info to UTF-8 assuming PDFDocEncoding | 22:26.35 |
mvrhel_laptop | I wonder if it is slow reading the files off the sd card | 22:34.04 |
| and that is killing us with the parsing time ray_laptop | 22:34.22 |
ray_laptop | hmm... I just realized that I was telling it that the 'base' value I said was good (9.07) was bad. Guess it is confused | 22:34.29 |
mvrhel_laptop | or does that not count | 22:34.33 |
ray_laptop | mvrhel: it counts, but most SD cards are ~20 Mb/sec reading | 22:35.07 |
mvrhel_laptop | ray_laptop: so I went ahead and did J12 at 72dpi | 22:35.18 |
| and all the time is in parsing | 22:35.24 |
ray_laptop | mvrhel: you can see how fast the SD card read is using 'time cp /dev/sd0/... /dev/null' | 22:35.58 |
| mvrhel: parsing includes a LOT more than reading the file. But PDF is random access, so maybe that is a problem (but SD cards don't do real 'seeking') | 22:37.08 |
mvrhel_laptop | no sd0 in /dev | 22:37.53 |
ray_laptop | mvrhel: well, wherever the J file is on your file system | 22:38.15 |
| it's mounted somewhere | 22:38.27 |
mvrhel_laptop | yes | 22:38.31 |
ray_laptop | just copy it to /dev/null and time it | 22:38.44 |
mvrhel_laptop | yes. ok that is a tiny tiny time | 22:39.31 |
| like 0.034 secs | 22:39.42 |
Robin_Watts | mvrhel_laptop: Time copying it to another file. | 22:40.05 |
mvrhel_laptop | ok | 22:40.10 |
| ok 0.075 | 22:40.29 |
Robin_Watts | ok, so genuinely read speed of the card is not going to be the issue :) | 22:40.52 |
ray_laptop | you can get rid of seeking by forcing gs to copy it to a tempfile (/tmp is probably in ram) rather than specifying the input file, use: - < inputfile.pdf | 22:40.56 |
mvrhel_laptop | ray_laptop: as we noted, it scaled up in the case of me running on my laptop: that is the render time was comparable to the parse time with the clist | 22:41.15 |
ray_laptop | that reads sequentially from the inputfile, writes it to a tempfile then runs it | 22:41.25 |
| mvrhel: true. That points to just a dog slow CPU | 22:41.49 |
mvrhel_laptop | right | 22:41.54 |
ray_laptop | one of the outfits you are going to see uses x86 (AMD) chips. | 22:42.48 |
mvrhel_laptop | oh that is nice | 22:42.58 |
ray_laptop | at least, I think they were on your agenda | 22:43.26 |
mvrhel_laptop | ok. so what shall I do with this data. I am torn on this, since I am not sure if this is really comparable to what they had | 22:43.48 |
| and I don't know what they are going to be comparing against | 22:44.04 |
ray_laptop | Well, if they had PDF on and ARM it probably wasn't a lot faster | 22:44.21 |
mvrhel_laptop | right | 22:44.31 |
| ok. so perhaps I will just share the 16MB data since I don't know why it is the sweet spot at this time | 22:45.15 |
ray_laptop | I don't know which company you are 'targeting' -- you mentioned laser 1200 dpi. Color ? | 22:45.20 |
mvrhel_laptop | right | 22:45.33 |
| ray_laptop: see private message | 22:45.51 |
ray_laptop | mvrhel: have you looked at the 'time' results -- see what real, user and sys are ? | 22:54.26 |
| maybe there is something else going on. | 22:55.04 |
mvrhel_laptop | ray_laptop: no, but let me try that | 22:55.04 |
| right. processor is busy with other stuff | 22:55.17 |
ray_laptop | mvrhel: I'll see if my old slow windows box still runs | 22:55.20 |
mvrhel_laptop | ray_laptop: so at 72 dpi with J12 real time was 1m17 user time was 1m16 | 22:59.35 |
ray_laptop | OK. so that points to all CPU | 22:59.57 |
mvrhel_laptop | and final time was 77 in the output final time | 23:00.22 |
| me piping out the stderr and stdout to a file should not matter should it? | 23:00.57 |
| I mean the amount of data is tiny | 23:01.15 |
| with just the page times for stderr and the command init in stdout | 23:01.37 |
| anyway, I think I have exhausted all of my understandings on this | 23:02.05 |
| ray_laptop: I would like to see how your old windows box does | 23:02.32 |
| Robin_Watts: don't you have a raspberry pi with gs? | 23:02.41 |
ray_laptop | mvrhel: my oldest system (500MHz AMD) is missing something. I have a 1.6GHz celeron laptop I will try | 23:21.01 |
| I also will try another slow system (also windows) | 23:21.34 |
| have to take my youngest to his piano. bbiaw | 23:25.07 |
mvrhel_laptop | ray_laptop ok | 23:26.47 |
ray_laptop | mvrhel: Well, I guess the 700MHz ARM isn't so bad after all. the 1.6GHz celeron took 69 seconds to do J9 at 1200 dpi. Parse time was 1.75, 1.9, 1.5, 2.0 and 1.6. Page 4 took 50 seconds to render | 23:36.26 |
| Forward 1 day (to 2013/08/15)>>> | |