| <<<Back 1 day (to 2014/03/18) | 2014/03/19 |
Robin_Watts | wow. I step away for 3 hours and you get it all working :) | 00:02.30 |
| just imagine how productive you'd be if I stayed away for longer. | 00:03.01 |
Jogux | not hugely different given I'm about to collapse into my keyboard :-) | 00:03.24 |
Jogux | notes that the ATS stuff seems to be devoid of a mac VM (perhaps not surprisingly). | 00:05.08 |
| I guess we need a real Mac | 00:05.27 |
Robin_Watts | Jogux: One of the cluster nodes is a MacPro. | 00:07.04 |
Jogux | is that in marcos's garage too? :-) | 00:07.22 |
Robin_Watts | No, that's in henrys house :( | 00:07.31 |
| which is in denver. | 00:07.42 |
Jogux | ah :-) | 00:11.22 |
| I think that's enough for me. night all. | 00:14.12 |
Robin_Watts | night. | 00:15.15 |
ray_laptop | Jogux: did you cc tech on the API docs that you sent to henrys ? Generally we do that to make sure everyone has access to info. | 00:30.41 |
Robin_Watts | ray_laptop: Jogux may not be able to post to tech. | 00:37.38 |
ray_laptop | mvrhel_laptop: in gximono.c the compiler notes that 'xrun' may be uninitialized, and it looks to me like it will be for the landscape case. Can you double check me on that ? | 01:35.26 |
| mvrhel_laptop: I'm going to try and come up with a test case. | 01:35.47 |
| (unless you have one laying around) | 01:36.25 |
| bbiaw... | 01:36.44 |
mvrhel_laptop | ray_laptop: it does appear that way. gxht_thresh_planes is using xrun and it is not set in the landscape case | 03:21.27 |
ray_laptop | mvrhel: (for the logs) Thanks for cross checking me. I've thrown together a test case so I can figure out what it should be set from. | 03:29.52 |
| The 4-bit halftoning is ready for review on my repo (Robin_Watts, mvrhel, or anyone). Gray case is improved by 1,8 x. CMYK by 3.2 x (even without SSE2). | 03:43.39 |
| BTW, the chunky mode CMYK for the image is 2.5x the original planar mode (sort of surprised at that) -- Our tile caching must be pretty effective I guess. That makes the fast planar code come in at 2.1x the chunky mode. | 03:51.58 |
| that's for portrait images. For landscape images, CMYK is 2.8x faster, Gray is 1.8x faster | 03:55.00 |
| makes me want to do the SSE2 next :-) | 03:56.09 |
| but first, I need to look into printing to the Pi over wireless (zero conf discovery mode, first) | 03:58.30 |
| henrys: unless you know of some other high priority for me. | 03:59.06 |
| henrys: I also thought about trying the example printer discovery on Android. The first was brought up at the staff meeting. | 04:00.53 |
| I might be back online later -- in nay case, I'll check the logs when I come back. Bye for now | 04:01.37 |
marcosw | chrisl: ray_laptop's latest commit not only fixed the segfaults but also the problems that mentioned with some files during this morning's meeting (the ones that I thought were caused by michael's commit). | 04:01.45 |
ray_laptop | marcosw: Hurray! Good news to finish the day. I might be back on later, but I'll check the logs when I come back | 04:02.36 |
chrisl | marcosw: thanks, I've pulled that fix and mvrhel's black point fix in. *But* I realised after I'd pushed that I had amended the documentation commit, so you'll need to reset to an earlier commit, and then pull the new commits from the branch. Sorry, I won't do that again...... (p.s. I'll send a mail with this in it, too) | 09:13.14 |
sebras | kens2: did you get inspired by tor5/tor8? :) | 10:45.12 |
kens2 | No just had a network hiccup | 10:45.27 |
Robin_Watts | tor8, paulgardiner_lap: Am I right in thinking that we never compress any streams? | 11:18.46 |
| In particular any appearance streams we generate are uncompressed? | 11:18.57 |
paulgardiner_lap | AFAIK, we dont | 11:19.24 |
tor8 | Robin_Watts: you mean xref objects that we update with new stream data? | 11:19.30 |
| afaik, we don't do anything to compress them. | 11:19.37 |
paulgardiner_lap | Although, I would imagine that is more to do with saving than appearance-stream synthesis | 11:20.06 |
pedro | paulgardiner_lap: just a note on where I'm at just now; spent some time on Friday looking at device build for android on windows and investigating some of the GhostDocs layout bugs; I stopped the android stuff for now as Jogux said you'd been working on it too. | 11:22.48 |
| I have vs2010 and vs2012 builds of dispman_ue2fileviewer running now on windows7, and have been going through layotu issues to identify the bigger issues | 11:23.33 |
paulgardiner_lap | pedro: yes, there are some commit to the main repo now for windows-hosted android builds. I have nother further since commiting them | 11:23.54 |
| pedro: that sounds great | 11:24.20 |
pedro | various missing content problems look related to differences in parsing between 'new' and 'legacy' ooxml DTDs, so that seems like a good one to tackle first | 11:24.24 |
| also the missing header/footer problems look like mis-hanlding of some external refs in the document fragments | 11:24.54 |
| if its ok i'll go for them first then move on to some of the other PPT problems? | 11:25.24 |
paulgardiner_lap | Yes, all sounds good to me. | 11:25.48 |
pedro | cool | 11:25.55 |
paulgardiner_lap | Do you have any notes/commits connected with getting the windows buids working. | 11:26.34 |
| I? | 11:26.48 |
Jogux | paulgardiner_lap Robin_Watts pedro : I'm very tempted to suggest that we rewrite the all the git history to remove the 'Summary:\n' at the start of all the picsel commits, as some of the git tools (eg. the gitk I have) like to display just that when showing a oneliner summary for the commits, which is less than helpful. | 11:27.33 |
pedro | Robin_Watts: I did also try (briefly) to get a vs2005 on Windows7 build, but its full of issues and it seemed more productive to focus on the vs2010/win7 config if that's ok | 11:27.37 |
Robin_Watts | Jogux: That's probably sensible. | 11:28.27 |
| pedro: I can cope with vs2010 I think. | 11:28.46 |
Jogux | seems better to do it now than once development is in full swing | 11:28.48 |
pedro | paulgardiner_laptop - just going to commit the build changes if that's ok - I'll stick build instructions up on the twiki if that's ok? (basically a tgvbuild.py -vs2010 --dispman-ue2fileviewer app/simple' build I'm using) | 11:29.06 |
Robin_Watts | Jogux: Yes. The existing epage -> git conversion took about 15 hours to run. | 11:29.19 |
Jogux | pedro : we don't have a twiki :-) | 11:29.28 |
Robin_Watts | If we have to reconvert I wonder if it would be worth rolling resources into the epage git repo too ? | 11:30.05 |
paulgardiner_lap | Jogux: my git mirroring we used to have at Picsel removed the "Summary" line from each commit | 11:30.22 |
pedro | Jogux: yeah, but will have soon-ish I guess? ;) | 11:30.24 |
Jogux | robin_watts : I would suggest just editting the logs rather than reconverting | 11:30.25 |
Robin_Watts | Jogux: OK. | 11:30.50 |
Jogux | but... we could do your thing too. resources is a pain the way it is at the moment. ish. | 11:31.04 |
| I made the build checkout resources automatically from your PicselCVS when needed, which makes it less of a pain. | 11:31.26 |
Robin_Watts | If we roll resources in, we are still left with a build system that won't actually build though. | 11:31.32 |
Jogux | (it only checks out the files it wants) | 11:31.35 |
Robin_Watts | cos the build system will try to checkout from cvs, and won't realise we have it in git already. | 11:32.18 |
| So we might as well just add resources to epage as we need them, fixing the build system as we go. | 11:32.34 |
paulgardiner_lap | pedro: just rough notes in an email would be good, pending the reanimating of the wiki. I need to get the windows builds working here too and would be useful to know what you had to install on your machine and any config | 11:33.42 |
Robin_Watts | Jogux: Do you want to go ahead and do that git change? Would make sense to do it sooner rather than later. | 11:34.03 |
tor8 | pedro: Robin_Watts: filtering out the Summary: lines from the epage git is easy to do with 'git filter-branch' | 11:34.47 |
| Jogux: ^ | 11:35.17 |
| Robin_Watts: you got a minute to talk about printf? | 11:35.45 |
Robin_Watts | tor8: sure. | 11:35.52 |
tor8 | Robin_Watts: so, having grepped through the code I see why you want separate %c for raw bytes and utf-8 encoding | 11:36.16 |
Jogux | robin_watts : the build system checks if the files are there before trying to get them from cvs | 11:36.45 |
tor8 | and now I'm wondering if we don't want a "quoted string" formatting character | 11:36.45 |
| that will escape binary as octal and do the usual \n and quote character escapes | 11:37.18 |
Jogux | tor8 : yeah, filter-branch is what I was thinking off | 11:37.24 |
paulgardiner_lap | robin_watts: I saw your note about making points readable by everyone. Should it also be made non-writeable seeing as we have only the one backup and we are intending to make copies of anything we change? | 11:37.33 |
Robin_Watts | Jogux: So maybe it *does* make sense to reconvert with resources attached? | 11:37.51 |
| paulgardiner_lap: possibly. | 11:38.07 |
| tor8: quoting for what system though? | 11:38.17 |
paulgardiner_lap | filter-branch could attach be used to combine resources and epage | 11:38.38 |
Robin_Watts | I'd like to see %c and %s stay exactly as they are. | 11:38.40 |
| i.e. C standard. | 11:38.47 |
tor8 | Robin_Watts: that's my main question :) C and PS/PDF have similar rules (octals, \r\n set) | 11:38.57 |
| but differ in which quote characters need quoting | 11:39.07 |
Jogux | robin_watts : I'm happy either way. adding resources would be another 15% or so on top of the current ~1G to clone the repo | 11:39.37 |
tor8 | we could make sure to quote all possible quote characters: ', ", (, ) or have separate codes for C and PDF quoting | 11:39.37 |
Robin_Watts | We were going to have and %S for utf8 ones, but we could use %u and ? | 11:39.40 |
paulgardiner_lap | Jogux: where did you put the notes on getting the iOS library build working? | 11:39.52 |
Robin_Watts | Jogux: 15% is nothing if it saves us grief along the way. | 11:39.54 |
tor8 | %q and %Q was my first thought, but then maybe %(s and %"s came to mind | 11:40.03 |
Jogux | paulgardiner_lap : I emailed them to you :-) | 11:40.07 |
Robin_Watts | If we use %u and $U then and %S would be available for quoted stuff. | 11:40.18 |
Jogux | and put them in my home directory | 11:40.20 |
| robin_watts : do you want to do that then? | 11:40.29 |
tor8 | %c and %C, then it would make sense with %S for UTF-16 strings to be printed as utf-8 but I'm still not convinced of the need | 11:40.46 |
paulgardiner_lap | Jogux: ah okay ta | 11:40.47 |
Robin_Watts | Jogux: I can't say it's a burning desire :) | 11:41.03 |
tor8 | 100% would be more worrisome, but the repo is already ginormous enough :( | 11:41.11 |
Robin_Watts | let me see if I can find my script from before. | 11:41.13 |
Jogux | robin_watts : :-) | 11:41.22 |
Robin_Watts | This presuably means we'd need to introduce a new top level dir. | 11:42.33 |
Jogux | nods. | 11:42.50 |
Robin_Watts | sot say. so we'd have sot/epage and sot/resources all within the sot repo. | 11:42.53 |
Jogux | yup | 11:43.29 |
Robin_Watts | is that going to give the tools conniptions ? | 11:44.02 |
Jogux | robin_watts : btw, I noticed the conversation didn't track renamed files at all afaics - I don't support there's an option in the tool for that? | 11:44.02 |
Robin_Watts | Jogux: Not that I know of. | 11:44.19 |
| but it might benefit from another set of eyes on it. | 11:44.27 |
Jogux | which tool? | 11:44.33 |
| robin_watts : the use of git is going to make ATS throws fits; anything else is relatively minor by comparison I think. | 11:44.59 |
tor8 | Robin_Watts: submodules? :) | 11:45.23 |
Robin_Watts | cvs2git --blobfile=cvs2svn-tmp/git-blob.dat --dumpfile=cvs2svn-tmp/git-dump.dat --username=cvs2git /home/robin/PicselCVS | 11:45.47 |
tor8 | (I have no idea how ofter the resources change, but it's worth mentioning) | 11:45.48 |
Robin_Watts | tor8: The problem is that we have all the source in a cvs module called epage. | 11:46.07 |
| and we have resources for that in another cvs module called resources that lives at the same level. | 11:46.28 |
Jogux | tor8 : resources rarely changes if ever. | 11:46.33 |
Robin_Watts | so to put them into a single repo we'd need sot/epage and sot/resources. | 11:46.48 |
| It involves extending the directory structure upwards by 1 dir. | 11:47.01 |
tor8 | Robin_Watts: hence my suggestion for a super repo which would use git submodules | 11:49.17 |
| and then you could check out the sot/epage or sot/resources repositories independently as well | 11:49.32 |
| dunno if that's useful | 11:49.37 |
Jogux | has no obections to submodules. | 11:49.45 |
tor8 | and it makes it easy to add more modules to the sot/ super-repo as we want | 11:50.03 |
Jogux | but also no option on whether they help or hinder | 11:50.04 |
| s/option/opinion/ | 11:50.16 |
Robin_Watts | It's an interesting idea. | 11:50.33 |
Jogux | robin_watts : that doesn't appear to have any options for better handling of renamed files :( | 11:50.44 |
Robin_Watts | I think it would be very hard to generate a super repo that had epage and resources as submodules that tied the dates of the 2 submodules together in the correct way. | 11:51.33 |
Jogux | I'm not necessarily sure we need to. | 11:52.08 |
| (anytime we need to go back in time we'd be using tags I would think) | 11:52.30 |
Robin_Watts | tags don't cross repos | 11:52.40 |
Jogux | actually my statement probably wasn't 100% true anyway. | 11:52.54 |
Robin_Watts | paulgardiner_lap: You had an alternative cvs -> git tool, didn't you? | 11:53.19 |
Jogux | perhaps just go for Robin's way for now - we can always submodule later (there's a git filter-something for doing it) | 11:53.23 |
Robin_Watts | Jogux: what was my way? | 11:53.47 |
paulgardiner_lap | robin_watts yes, a ruby-based thing called from-cvs/to-git | 11:53.52 |
Jogux | robin_watts : a single repository containing epage & resources directories. | 11:54.31 |
paulgardiner_lap | It kept an up-to-date git mirror of the cvs repo | 11:54.37 |
Jogux | pailgardiner : the config for that could well be on the NAS :-) | 11:55.09 |
Robin_Watts | IIRC I did this on peeved, not casper cos it was faster. | 11:55.19 |
tor8 | Jogux: git subtree | 11:55.23 |
Robin_Watts | Jogux: oooh. | 11:55.40 |
Jogux | tor8 : that sounds right, yep, thanks | 11:55.44 |
paulgardiner_lap | Jogux: I also have the config here | 11:55.48 |
tor8 | but yes, if you want to keep them in sync with submodules that needs more scripting voodoo | 11:55.52 |
Jogux | ah, even handier | 11:55.54 |
paulgardiner_lap | Jogux: it used to run on snake at picsel's offices | 11:56.10 |
| I think git filter-branch could probably link resources into epage as a submodule with the versions linked correctly | 11:57.06 |
| Jogux: just read your previous comment. Yes, robin_watts suggestion first then submodule later if we wish | 11:58.12 |
tor8 | if they both have the same tags, you could iterate over the tags with a script in the super-repo, check out the submodules to the tag and make commits and tags in the super-repo from that | 11:58.47 |
| shouldn't be more than an afternoon's work (famous last words) | 11:58.58 |
paulgardiner_lap | Does the resource repo have the same branch structure as the epage repo? | 12:00.04 |
Robin_Watts | paulgardiner_lap: resource and epage are the same CVS repo. | 12:00.23 |
Jogux | paulgardiner_lap : yes. | 12:00.29 |
Robin_Watts | Different modules within that repo. | 12:00.30 |
Jogux | or should. people were meant to tag / branch both if they tags/branchedd one. | 12:00.48 |
Robin_Watts | IIRC, paulgardiner_lap's solution was MUCH faster than my conversion tool. | 12:00.55 |
paulgardiner_lap | Ah, I forget how cvs works | 12:00.58 |
Jogux | Paul's one possibly didn't do branches, or am I misremembering? | 12:01.22 |
Robin_Watts | http://cvs2svn.tigris.org/cvs2git.html <- That's the docs for mine. | 12:01.53 |
paulgardiner_lap | Yeah it did branches. | 12:01.59 |
Jogux | Ihave no idea if Paul's one was any better at renamed files. | 12:03.21 |
paulgardiner_lap | has no idea either | 12:03.42 |
| Hmmm, I think the android lib we just released was built with the master branch of resources and the smart-products branch of epage. Oops! I'm definitely in favour of combining them. | 12:06.42 |
kens2 | heads for lunch | 12:10.19 |
paulgardiner_lap | Too late now, but I'd have thought it was better for resources not to be under version control at all. Make it something you only ever add to. If you want to change someting, add a new version with a new name and change the point of use within the source | 12:12.46 |
| That way every version of the source can use the latest version of resources | 12:13.16 |
Robin_Watts | paulgardiner_lap: No. You want to be able to rewind to a particular place in history and reproduce what you shipped. | 12:15.34 |
| hence resources and code should be in sync. | 12:15.52 |
paulgardiner_lap | But you still can | 12:15.57 |
| The add-only resources will always give the correct thing | 12:16.15 |
| It's always a superset of past versions | 12:16.39 |
Robin_Watts | Oh, I see what you mean. Bloat city arizona. | 12:16.49 |
| revision control is better, I feel. | 12:17.08 |
| If you want to do a build with the latest resources, just update the resources on the appropriate branch and do a commit. | 12:17.29 |
paulgardiner_lap | Resources will bloat a repo anyway because they are often binary | 12:17.34 |
| robin_watts still seems simpler to me "just always use the latest version of resources" | 12:18.12 |
tor8 | paulgardiner_lap: but your approach will bloat both the repo and the workspace | 12:18.14 |
paulgardiner_lap | tor8: true, but so will a git repo | 12:18.44 |
| Actually I doubt it. I bet most changes to the resources are additions in any case | 12:19.16 |
tor8 | a git repo will only bloat the .git, you don't have to pay twice for the working directory as well | 12:19.32 |
paulgardiner_lap | But with resources not under version control you pay twice even less | 12:20.09 |
| It's a thought anyway... maybe not a great one. :-) | 12:20.55 |
Robin_Watts | A git repo only ever stores one version of a blob. | 12:22.01 |
| so if the same font is used in lots of different places git only ever stores it once. not sure how we can do better than that with what paulgardiner_lap is suggesting. | 12:22.35 |
tor8 | paulgardiner_lap: Jogux: git doesn't track file renames at all in the repository structure; that's all done in the clients if at all | 12:23.33 |
| so any file rename information from cvs would be lost no matter what you do, and at the mercy of the file rename heuristics in the git client tools | 12:24.03 |
Crocell | hi | 12:24.34 |
ghostbot | niihau | 12:24.34 |
tor8 | Robin_Watts: mind taking a look at tor/master, there's a slightly revised printf (that does %c and %C separately) and a draft of %q | 12:25.27 |
Robin_Watts | looking now. | 12:25.41 |
paulgardiner_lap | robin_watts I was assuming resources didn't contain repeated files, at least not many... but in any case we couldn't do what I suggest now even it it were a good idea | 12:25.57 |
Robin_Watts | tor8: The printf one looks good. Looking at next one now. | 12:27.31 |
Jogux | tor8 : ah. I suspect the problem I saw was actually because I'd missed off --follow to git log. d'oh. | 12:27.35 |
| so. yes. ignore all my whittering about renames. oops. | 12:27.50 |
Robin_Watts | tor8: Looks good to me. | 12:29.34 |
| While I am doing this... is there anything other than resources we should include? | 12:30.07 |
paulgardiner_lap | Here's the alternative conversion tool, btw: http://mercurial.selenic.com/wiki/fromcvs | 12:31.44 |
Jogux | robin_watts : epage-docs is the only thing that might be relevant. not sure it's worth it. | 12:32.28 |
tor8 | Robin_Watts: thanks. running a cluster test now with your content cleaning | 12:34.07 |
Robin_Watts | Jogux: Not customer-aliens? | 12:36.28 |
| us2sdk? | 12:37.41 |
| ue2sdk? | 12:37.43 |
| thirdparty? | 12:37.52 |
Jogux | afaicr there's nothing useful in any of those. | 12:38.53 |
| unless we happen to still have any of the customers relevant to customer-aliens. | 12:39.14 |
paulgardiner_lap | Ah, hadn't thought before, combining epage and resources is going to interleave the history | 12:39.58 |
Jogux | ue2sdk is just the c# editor thing. never finished. not sure about thirdparty but I suspect unlikely to be relevant. | 12:40.22 |
Robin_Watts | paulgardiner_lap: That's the whole point. | 12:41.24 |
| ok, conversion running now. If this is anything like before, I'll be restarting it a few times and it'll take a few days to actually get us something that works. | 12:42.26 |
| Then we'll need to do the filtering of the names. | 12:42.39 |
| Then we'll need to pull in the changes from the existing repo. | 12:42.52 |
Jogux | presumes the changes can just be cherry-picked across fairly quickly. | 12:43.36 |
Robin_Watts | well, with some editing because there will now be an 'sot' in the path or something. | 12:44.27 |
Jogux | robin_watts : oh yes, of course. | 12:45.31 |
Crocell | i'm trying to use ghostscript to generate 128px (largest dimension) png thumbnails of the first pages of pdf files of various aspect ratios. | 12:46.36 |
| trouble is: while i do get a square 128x128 thumbnail with -g128x128 -dPDFFitPage=true, the aspect ratio is lost, i.e. the background is white or transparent on the _whole_ square so that the original aspect ratio of the document is no longer visible. | 12:46.46 |
| is there an easy way to | 12:46.55 |
| * either have thumbnails with 128px largest dimension and the smaller dimension being smaller so as to keep the aspect ratio of the original document | 12:47.02 |
| * or have square 128x128 thumbnails, but with the padding being transparent or black so that the original aspect ratio of the document remains visible? | 12:47.08 |
| ? | 12:47.14 |
Robin_Watts | Crocell: Don't use gs. | 12:47.30 |
| Use mupdf, specifically mudraw. | 12:47.38 |
| mudraw -o out%d.png -w128 -h128 in.pdf | 12:48.30 |
Crocell | ok thx. but i had hoped gs would do it too (as i don't just have to do it on _my_ machine but also via calling program on other machines where i know ghostscipt is installed as it is used for ps->pdf printout conversion but don't believe mudraw is installed and i don't know if i can get it installed everywhere) | 12:50.25 |
Robin_Watts | Well, it's harder with gs. You probably need to get gs to tell you the aspect ratios, then select w/h in a second run yourself. | 12:51.31 |
chrisl | Crocell: If you're forcing the page size to 128x128 pixels, surely you're losing the aspect ratio anyway? | 12:51.51 |
Crocell | Robin_Watts: ok thx. i'm cool with that. i just don't know how to get the aspect ratio information via gs | 12:52.14 |
Robin_Watts | chrisl: No, I think he wants a thumbnail that fits within 128x128 but maintains the aspect ratio. | 12:52.27 |
| it's a common request. | 12:52.36 |
Crocell | chrisl: yeah i don't mind as long as the padding is black or transparent so that the _visible_ aspect ratio remains that of the document | 12:52.50 |
| chrisl: even if the image file has a fixed square aspect ratio | 12:53.17 |
| Robin_Watts: yup exactly | 12:53.42 |
chrisl | It should be possible, but maybe the stupidities of the PDF interpreter get in the way :-( | 12:56.48 |
Robin_Watts | tor8: PDF clean commit is ready to go I think: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=33a77bc8c2ca1815ff902acfc5be3c208740e59b | 12:57.13 |
Robin_Watts | runs. | 12:57.46 |
| Crocell: Mupdf is still a better idea. | 12:58.24 |
| It'll do antialiasing better than gs too. | 12:58.37 |
chrisl | Crocell: maybe "-sDEVICE=png16m -dDEVICEHEIGHT=128 -dDEVICEWIDTH=128 -dFIXEDMEDIA -dPDFFitPage=true" ? Or that might do the same things..... | 12:59.45 |
kens | THat will produce a fixed 128x128 output file | 13:00.00 |
| You need to get the page size, figure out the scaled down size and set the page size to that, then do -dPDFFitPage (I think) | 13:00.26 |
| It would be fairly easy to do thi in the PDF interpreter I think | 13:01.12 |
chrisl | kens: with the commandline above I get a 128x128 pixel image, but the PDF contents are drawn with their original aspect ratio | 13:02.33 |
kens | I htought the desire was for the page size to be a maximum of 128 X or Y but the other dimension to be scaled appropriately | 13:03.26 |
chrisl | kens: above Crocell said "even if the image file has a fixed square aspect ratio" | 13:03.59 |
kens | OK fair enough, I was at lunch | 13:04.09 |
chrisl | For thumbnails, I'd have thought a consistent, tilable image size would be preferable, anyway | 13:05.26 |
Crocell | ahm i'm not sure to understand that above stuff. with that command line i get a square file where the padding is white (and thus the aspect ratio of the original document is not visible inside the square) | 13:11.56 |
| as for solutions that rely on previous knowledge of the pdf page or box size: would be nice if getting that information was possible with gs | 13:13.41 |
kens | It is possible | 13:13.50 |
chrisl | It might depend on the content of the file: for the example I tried here, the pixels outside the original aspect ratio were transparent with pngalpha | 13:13.53 |
kens | ghyostpdl/gs/toolbin/pdf_info.ps will retrieve a lot of information about a PDF file | 13:14.44 |
| That includes Media sizes | 13:14.52 |
Crocell | chrisl: yeah but with pngalpha, the background pixels _inside_ the original aspect ratio are transparent too on some documents | 13:15.20 |
| chrisl: thus also losing the aspect ratio there | 13:15.39 |
kens | Start by running a PostScript file which paints the contents black.... | 13:15.49 |
| To get the size of a given page :(filename) | 13:17.08 |
| Ooops | 13:17.12 |
| (filename) (r) file runpdfbegin pagenumber pdfgetpage /MediaBox get == | 13:18.07 |
chrisl | kens: I'm not sure painting the backdrop works, I suspect the PDF interpreter foxes that one somehow :-( | 13:18.30 |
kens | chrisl it might, but its worth a try. | 13:19.00 |
chrisl | Doesn't *seem* to work for me :-( | 13:19.49 |
kens | OK guess not then | 13:19.58 |
chrisl | PDF interpreter's setpagedevice probably buggers that | 13:20.41 |
kens | Ah yes, that's true, that's an implicit erasepage | 13:20.59 |
| Best bet then is probably to get the media size, figure out the desired dimensions and set the page size to that | 13:23.08 |
Crocell | so pardon my n00bish ignorance: how do i tell gs from command line call to give me that media size? | 13:23.52 |
chrisl | Crocell: if you find the PS utility pdf_info.ps and look at the comments at the beginning of the file | 13:25.15 |
kens | Crocell : that line of text above, feed ti to GS inside "" preceded by -f and followed by -c. Or put it in a file and send the file to GS | 13:25.41 |
| Obviously replacing 'filename' and 'pagenumber' with the relevant data | 13:26.08 |
henrys_ | chrisl: not sure if you saw my question to you buried in the email to Peter at URW. | 13:27.31 |
chrisl | henrys_: Yeh, I'm about to look at getting the images now | 13:28.10 |
tor8 | Robin_Watts: I'd rather add another flag for content stream sanitising than tack on another 'g' | 13:28.39 |
| -s clean content streams | 13:28.58 |
Crocell | ok thx (installed ghostpdl now, trying) | 13:32.36 |
kens | Umm you mean Ghostscript ? | 13:32.57 |
Crocell | well i did have ghostscript installed, but pdf_info.ps was not in there | 13:33.18 |
kens | It should be, and it won't be in 'GhostPDL' because that's not a real thing, its a collective term for the family of products... | 13:33.44 |
| Where did you get GS from ? | 13:33.53 |
chrisl | kens: most distributions don't include toolbin for some reason | 13:34.45 |
kens | Well tht would do it | 13:34.52 |
Crocell | well, on windows i have it from here: http://sourceforge.net/projects/ghostscript/files/GPL%20Ghostscript/9.07/ | 13:37.10 |
kens | Ah, chrisl that looks like we don't distribute toolbin in our installer ;-) | 13:38.13 |
chrisl | No, I don't think we do. TBH, I think PS utilities like that should be in lib.... | 13:38.37 |
kens | is inclined to agree , the 'bin' part of the name is misleading | 13:38.59 |
chrisl | I'll do that for the next release. | 13:40.33 |
marcosw | chrisl: I did the git reset ... ; git pull ... that you mentioned in your email but I do not see a ghostpdl-9.12rc2 tag. Any ideas? | 13:46.00 |
chrisl | marcosw: I haven't done an RC2 - I wanted to ask if you were planning to restart the testing? | 13:46.31 |
marcosw | chrisl: yes, I am. | 13:46.45 |
chrisl | marcosw: okay, give me 20 minutes or so, and I'll do the new tag | 13:47.10 |
| marcosw: it should be there now (remember the "-t" to fetch new tags) | 13:49.14 |
marcosw | btw, I do not see the git pull fail you mention in your email: http://pastebin.com/wtr4ZwyC | 13:49.20 |
chrisl | That's odd... oh well.... | 13:50.11 |
| marcosw: Oh, no fetch will be okay, but when you try to rebase into your working copy, you'd get the error | 13:50.50 |
| marcosw, henrys_: do you think I should do a full RC2 archive, too? I'm leaning towards "not" given how niche Ray's changes were.... | 13:52.09 |
marcosw | you mean with git pull --rebase? that worked as well. | 13:52.11 |
chrisl | marcosw: oh, well, <shrug>.... | 13:52.37 |
henrys_ | chrisl: I don't think it's needed. | 13:53.10 |
marcosw | perhaps because I didn't have any local changes in that repository, so the rebase didn't need to do anthing? | 13:53.48 |
chrisl | marcosw: possibly, I'm not too worried - I just didn't want you getting stuck if the error did happen. | 13:54.49 |
| henrys_: Good, thanks | 13:55.11 |
paulgardiner_lap | henrys_: was reading back through the logs regarding the printer module earlier. One of tkamppeter's comments seemed to suggest that Airprint does most of what a cups port would do, or did I misunderstand? | 14:05.18 |
chrisl | paulgardiner_lap: I wouldn't be surprised if AirPrint has a lot of CUPS stuff in it..... | 14:11.15 |
Crocell | ahm another n00bish question: when i try these ghostscript commands to get pdf information, a window pops up where some text is displayed - that window looks like that which i get on errors, but unlike the error cases it disappears again right away. | 14:11.39 |
| how do i a) get the window to not disappear and | 14:11.52 |
paulgardiner_lap | chrisl: oh yes, probably true | 14:12.01 |
chrisl | Crocell: are you using gswin32.exe or gswin32c.exe? | 14:12.13 |
Crocell | b) get it to not pop up but instead get its text redirected to a file | 14:12.21 |
| gswin32 | 14:12.23 |
kens | Crocell : use gswin32c for the console | 14:12.26 |
Crocell | oh is that the difference ok thx | 14:12.32 |
kens | Or don't set -dBATCH | 14:12.52 |
chrisl | And then use the normal pipe to write to a file | 14:13.06 |
Crocell | ok thx | 14:13.19 |
henrys_ | paulgardiner_lap: yes but I'm concerned about legacy printers | 14:15.28 |
| paulgardiner_lap: I also know 0 about windows mobile printing. | 14:16.24 |
| paulgardiner_lap: speaking of which should we be working toward something for the windows store. Picsel said windows mobile is coming soon ;-) | 14:17.03 |
paulgardiner_lap | Oh my god! :-) | 14:17.36 |
henrys_ | coffees | 14:18.28 |
| chrisl: everything I've read about cups and AirPrint talks about a linux server in the mix. | 14:29.22 |
chrisl | henrys_: it's hard to know, but given that CUPS is core to OS X printing, I would not be surprised if they butchered parts of it to make AirPrint | 14:30.39 |
Jogux | airprint is IPP, bonjour and PDF I believe. | 14:31.33 |
ray_laptop | Crocell: PDFFitPage retains the aspect ratio. Perhaps you need pngalpha ? This has transparent bg | 14:32.32 |
| if PDFFitPage is changing the aspect ratio, it's a bug | 14:33.02 |
kens | ray_laptop : it seems this is not the case | 14:33.43 |
| the problem seems to be transparent background (or not transparent) | 14:33.59 |
tkamppeter | paulgardiner_lap, note that AirPrint is rather limited, as it requires either a printer with explicit support for AirPrint (Apple's intended way how it works) or a Ubuntu or Debian machine which shares a printer (my hack of AirPrint printer emulation). If we would use our stack, we could print on a wide range of network printers and on all printers shared from machines running CUPS (including Mac OS X). | 14:35.42 |
ray_laptop | tkamppeter: and printers shared by Windows machines ? Or are those printers not supported ? | 14:37.45 |
| tkamppeter: and what do you mean by "a wide range of network printers" ? | 14:38.32 |
Robin_Watts | tor8: I started like that, but changed my mind. Will fix. | 14:38.49 |
ray_laptop | kens: which is not the case? that PDFFitPage does not change the aspect ratio ? or that pngalpha has transparent background ? Note that 9.12 will be different to older GS since I fixed a bug on that | 14:40.28 |
kens | ray_laptop : that the aspect ratio is unchanged, it is unchanged | 14:40.52 |
ray_laptop | so it is not the case that we have a bug ? | 14:41.16 |
kens | ray_laptop : depends, this is not a supported use case :-) THerefore its not a bug | 14:41.37 |
| The problem seems to be whether or not the unused area is white, or transparent. | 14:41.58 |
ray_laptop | kens: sorry for being dense. Coffee hasn't hit the brain yet ;-) | 14:41.58 |
| the unused are is supposed to be transparent with pngalpha | 14:42.38 |
kens | Its also possible that this is something which you fixed | 14:42.47 |
ray_laptop | kens: the fix is in 9.12 commit 174cfd14e4d34a37cdcf2e0372bd0b4334310e8b | 14:43.36 |
kens | ray_laptop : yes so chrisl will be using that, and Crocell weon't be | 14:43.53 |
chrisl | kens: no, I specifically used an older release | 14:44.18 |
kens | No idea then | 14:44.27 |
ray_laptop | chrisl: and it worked for you ? | 14:44.47 |
chrisl | ray_laptop: IIRC, the command line Crocell posted gave the wrong aspect ratio to the page markings | 14:45.36 |
norbertj | hello henrys_ | 14:45.38 |
ray_laptop | the older pngalpha would have transparent outside the specified media size (I think) but the entire media rectangle filled would not be transparent. | 14:46.04 |
norbertj | henrys_: I just entered a bug 695100 on rastergfx handling (jpeg color image not correctly rendered). | 14:46.43 |
ray_laptop | the "fixed" pngalpha has transparent everywhere on the page not actually painted by the page Contents | 14:47.01 |
tkamppeter | ray_laptop, for machines shared by Windows CUPS needs parts of Samba in addition, especially the "smb" backend which is a symlink to smbspool (or smbclient?) and smbclient for the discovery. | 14:47.14 |
ray_laptop | chrisl: So, are we good to go on the release ? no more blockers ? | 14:47.34 |
chrisl | ray_laptop: as far as I know, yes | 14:47.45 |
ray_laptop | tkamppeter: OK. Thanks | 14:47.55 |
tkamppeter | ray_laptop, with "a wide range of network printers" I mean IPP printers which understand at least one of these languages: PWG Raster (IPP Everywhere), PDF, PostScript, PCL XL. PCL 5e, or PCL 5c. | 14:48.25 |
ray_laptop | chrisl: Great! So we are waiting for marcos' final release testing ? Or is that done ? | 14:48.29 |
chrisl | ray_laptop: marcosw is restarting the testing with your fixes and mvrhel's fix | 14:49.01 |
ray_laptop | chrisl: OK. Thanks. | 14:49.12 |
Crocell | oops sry was away | 14:56.50 |
paulgardiner_lap | tkamppeter: ah right. Thanks for the explanation. | 14:57.29 |
Crocell | ray_laptop: yes pngalpha renders the padding as transparent. it works for some pdfs (document part with white background, padding with transparent background) but on others there is a transparent background everywhere | 14:58.40 |
henrys_ | hi norbertj, yes I just saw that. | 14:58.53 |
Crocell | i guess some pdfs just don't set a background color, and in that case the aspect ratio is not visible any more | 14:59.13 |
kens | Crocell that's the way its meant to work I believe, only the areas marked are not transparent | 14:59.32 |
henrys_ | norbertj: is that a regression? | 15:00.24 |
norbertj | henrys_: I don't think so, it was already in 9.04 | 15:01.16 |
Crocell | my problem is that i have no way so far (that is: without extracting and re-injecting size information, which i haven't got working yet) to handle the cases where the original pdf sets a background color and those pdfs that don't | 15:02.09 |
kens | You can't get the media size, or are having trouble setting it ? | 15:03.41 |
Crocell | kens: so far i had problems getting the MediaBox | 15:03.59 |
kens | Let me try it, give me a mnute | 15:04.09 |
Crocell | i do get the whole sizes with -dDumpMediaSizes | 15:04.20 |
kens | Right | 15:04.28 |
Crocell | but that other line that was supposed to just print the MediaBox size just gives me | 15:04.51 |
| --nostringval-- | 15:05.02 |
kens | You shold have said ;-) | 15:05.14 |
| Lost the line, have to go to the logs to see it | 15:05.43 |
henrys_ | norbertj: so if I had this entire image that white line would repeat? | 15:05.48 |
norbertj | Yes, on the complete page there are multiple black stripes (i.e. the image is underlayed with a black rectangle). But it's (a piece of) a customer job which we found when printing on a b&w printer. | 15:06.42 |
Crocell | kens: (filename) (r) file runpdfbegin pagenumber pdfgetpage /MediaBox get == | 15:07.33 |
kens | Its OK I read it from our logs | 15:07.44 |
| This: | 15:08.24 |
| %! | 15:08.24 |
| (/temp/source.pdf) (r) file runpdfbegin 1 pdfgetpage /MediaBox get == | 15:08.24 |
henrys_ | norbertj: did you already try pulling that ROP out? | 15:08.30 |
kens | returned hte result [0 0 500 328] for me | 15:08.37 |
| So lower left x,y is 0,0 and upper right x,y is 500,328 | 15:08.55 |
norbertj | henrys_: n, but I don't think its the rop. I added some extra info to 695100 | 15:09.22 |
kens | Crocell you did remember to put a fully qualified path in replacement for (filename) and changed pagenumber to an integer value (1 is good) ? | 15:10.00 |
Crocell | oh ok just tried again, prob was a typo: | 15:10.12 |
| when i added the "quit" without which i got an error, i somehow removed one "=" inadvertently | 15:10.46 |
| ok i get a box now too | 15:10.58 |
kens | RIght, so there's you r media size | 15:11.11 |
Crocell | ok thx | 15:11.17 |
kens | You probably want to ignore the lower left | 15:11.19 |
norbertj | henrys_: if I remove the ROPs from the file I loose the color, but the white stripe is still there (I'll attach this testfile also). | 15:12.21 |
kens | Theoretically you can probably use tht to set the media size, but I'm not sure you can set fixed media frominside PostScript. Maybe Ray knows | 15:12.28 |
norbertj | henrys_: can you reproduce? | 15:14.38 |
Crocell | btw about the direct conversion (without injecting size information): if there was a way to pass a parameter that said "if the pdf document doesn't specify a background color, then use white as a background color", then pngalpha would work as is and show the aspect ratio | 15:16.05 |
kens | But then the people that want' pngalpha to only show the marked area woudl complain. | 15:16.40 |
henrys_ | norbertj: yes I'll look at it. | 15:16.53 |
kens | Documents *never* specify a background colour, that's not the way the imaging model works | 15:17.02 |
| Becausew this is, at heart, a print model, so the page is whatever colour paper you use | 15:17.25 |
ray_laptop | to set FIXEDMEDIA from within PS, use: /FIXEDMEDIA true def | 15:17.29 |
kens | OK so there you go Crocell | 15:17.39 |
Crocell | ? | 15:17.43 |
kens | Take the last two entries from the width array, use those as the argument to /MediaSize in setpagedevice, and set /FIXEDMEDIA true | 15:18.12 |
Crocell | ah ok thx | 15:18.53 |
kens | That will get you a PNG whihc has hte aspect ratio you want. Obviously you'll have to play with the media sizes, since those are in 72nds of an inch | 15:18.55 |
| You'll need to work with the resolution to figure out what size that is, then divide the two values by the scale factor so that the largest is 128. | 15:19.40 |
Crocell | ok thx | 15:20.40 |
| so do i understand correctly that those pdfs that show a white background actually draw a white rectangle as a background? | 15:22.41 |
kens | Yes, that's right | 15:22.49 |
henrys_ | norbertj: strange how destination width is 4761 in the PXL, likely irrelelevant | 15:22.56 |
kens | Sometime people do that | 15:23.00 |
ray_laptop | Crocell: actualluy, I just started gs with -dPDFFitPage, then did: << /PageSize [ 400 400 ] >> setpagedevice /FIXEDMEDIA true def (examples/annots.pdf) run | 15:26.18 |
| Crocell: and the pages were scaled to the page size I set | 15:26.37 |
Crocell | ahm i don't get that with the << and >> | 15:27.59 |
| i mean was that one same command line? | 15:28.36 |
kens | << and >> are dictionary delimiters | 15:28.51 |
| setpagedevice takes a dictionary as argument (this is PostScript) | 15:29.21 |
ray_laptop | from what kens said, I thought you needed PS. | 15:30.12 |
kens | ray_laptop : the input is PDF | 15:30.29 |
Crocell | thumbnails: input: pdf, output: png | 15:30.49 |
ray_laptop | kens: right, that's why I used annots.pdf as the input for my test | 15:30.59 |
kens | Seems reasonable | 15:31.09 |
ray_laptop | Crocell: so, it sounds like you want thumbnails of a maximum size, (say 128 x 128) but you want those thnumbnails to be of varying dimensions according to the MediaSize (or CropBox) in the PDF | 15:32.49 |
| gs can do that :-) | 15:33.03 |
Robin_Watts | tor8: You didn't test printf under vs did you? | 15:33.25 |
Crocell | ray_laptop: yeah either that (i.e. the smaller dimension being <=128 following the aspect ratio) | 15:33.35 |
ray_laptop | the "magic" of InputAttibutes with page sizes that are ranges | 15:33.46 |
| Crocell: give me a minute or two... | 15:33.59 |
Crocell | ray_laptop: or square icons are fine too, if the padding is transparent and the background color is white | 15:34.11 |
kens | Hmm, never thought of that, I wonder if it will work | 15:34.17 |
ray_laptop | begins tracing pentagram on floor in chicken blood while mumbling incantations :-) | 15:34.50 |
Crocell | lol | 15:34.56 |
Robin_Watts | or... you could call mupdf and it would do exactly what you want... | 15:35.12 |
henrys_ | norbertj: there is an underflow in the image so the problem is further upstream, the jpeg decoder or more likely the pxl interface to the decoder, running -ZB "premature end_image" | 15:35.41 |
Jogux | Robin_Watts is intend on driving the chicken blood suppliers out of business | 15:35.44 |
ray_laptop | robin_watts: Pleasse go ahead and tell me how | 15:35.48 |
Robin_Watts | mudraw -o out%d.png -w128 -h128 in.pdf | 15:35.56 |
| as I said about 3 hours ago :) | 15:36.20 |
ray_laptop | I missed it | 15:36.27 |
Robin_Watts | :) | 15:36.31 |
kens | THe only problem there was that MuPDF miught not be available, and GS is (so we are told) | 15:36.34 |
Crocell | yeah that for one | 15:36.51 |
Robin_Watts | at some point it becomes easier to say "you need mupdf installed". It's not like it's a hard build. | 15:37.02 |
kens | depends if you're on WINdows and have a compiler ;-) | 15:37.46 |
henrys_ | norbertj: do you want me to look at it now, or did you want to study it more? We can give a bounty to your favorite charity if you want to look at it. | 15:38.08 |
Robin_Watts | for windows it's a simple binary. | 15:38.20 |
kens | was joking | 15:38.28 |
henrys_ | norbertj: I don't think it is going to be that painful a problem but I could be wrong. | 15:38.39 |
Robin_Watts | curses MSVC and it's non standard C runtime. | 15:38.54 |
norbertj | henrys_: I think I have some time this week, so I could give it a try. ;) | 15:39.02 |
henrys_ | norbertj: okay and if you need anything let me know. | 15:39.32 |
norbertj | henrys_: will do, bye | 15:39.48 |
Robin_Watts | FWIW: I have no mobile phone for the next 3 days while I wait for a new SIM. | 15:42.11 |
kens | changed network ? | 15:42.27 |
henrys_ | Jogux, pedro any preliminary thoughts on how difficult these bugs will be? | 15:42.33 |
Jogux | henrys_ : did you see what pedro said earlier? | 15:43.58 |
Robin_Watts | kens: changed handset. | 15:44.09 |
Jogux | robin_watts : boggle | 15:44.26 |
henrys_ | Jogux: you were akin about a mac earlier marcosw does have an iMac on his network at home. Is that enough horsepower? | 15:44.33 |
kens | Oh thought you could just use same SIM, or is it a micro or something ? | 15:44.34 |
henrys_ | Jogux: I have an 8 core mac pro | 15:44.47 |
Robin_Watts | Jogux: S3 takes a micro sim, I had a regular. | 15:44.47 |
henrys_ | Jogux: I'll check the logs | 15:44.58 |
Jogux | henrys_ : imac should work fine. just needs to do occasional iOS builds really. | 15:44.59 |
| robin_watts - yeah... but usually they send out the new sim, and switch over after it's arrived... | 15:45.37 |
Robin_Watts | Jogux: Nope, they cancel this sim as soon as they send the new one. Which is fine, cos I tried cutting the sim down to size last night... and now I don't have a working sim anyway :) | 15:46.19 |
Jogux | henrys_ : it's best (at least for the moment) if the mac is close to the rest of the ATS system, network wise - ie. it's on marcos's lan. | 15:46.25 |
| Robin_Waitts : lol | 15:46.29 |
kens | ROFL | 15:46.36 |
Robin_Watts | The Internets said it would work! | 15:46.49 |
kens | If you were Americna you would now need to sue someone :-) | 15:47.14 |
Jogux | has utterly, utterly failed to get the ATS windows build VM to run in parellels or vmware fusion. And it sounded like such a nice way to get a pre-setup working windows build environment :-( | 15:50.00 |
Robin_Watts | Jogux: How come? | 15:50.43 |
| I mean, any idea why? | 15:50.51 |
Jogux | robin_watts : parallels insists the vm needs to be upgraded, does so, shutdowns down, then starts up again, says the vm needs to be upgraded, .... | 15:51.12 |
| robin_watts : vmware blue screens almost on boot saying the disc is corrupt | 15:51.23 |
kens | O.O | 15:51.32 |
Robin_Watts | Is it possible that the disc is corrupt? | 15:51.40 |
Jogux | I possibly used too old a version of qemu-img to do the conversion to vmware. | 15:51.53 |
Robin_Watts | i.e. might it have been broken during transfer? | 15:51.54 |
Jogux | Robin_Watts : I really hope not. | 15:52.00 |
Robin_Watts | md5 it? | 15:52.06 |
Jogux | I'll try booting it up in kvm tonight | 15:52.07 |
| Robin_Watts : ah, the md5 sum I have matches what's on the NAS, yes | 15:52.19 |
| Robin_Waitts : it's the "working vm" -> NAS bit I'm worrying about. | 15:52.31 |
| I'm 90% certain it'll be fine. | 15:52.42 |
Robin_Watts | yes, me too :( | 15:52.43 |
pedro | henrys_: there are a few bugs (aka unimplemented features) in the ooxml handling for linked images and some broken external reference handling - they're basically a case of identifyin the missing paths in the parsing code. Not 'hard' but fiddly, time consuming changes. The other layout probelms (lineheight etc) I haven't looked at yet. We'll likely have some issues with text flowing around... | 15:53.04 |
| ...objects and column handling as some of it isn't currently supported | 15:53.05 |
Robin_Watts | pedro: Lineheight is probably the most crucial issue, I think. | 15:54.51 |
pedro | moreso than missing half the contwent? :) | 15:55.08 |
| content | 15:55.18 |
kens | likes the sound of contwent :-) | 15:55.48 |
| But ownders where it went | 15:55.55 |
pedro | :) | 15:56.00 |
chrisl | Is that what you get when Word crashes? | 15:56.19 |
Robin_Watts | pedro: Oh, I hadn't noticed the "missing half the content" stuff so much. but that does sound worse. | 15:56.53 |
Jogux | Robin_Watts : yeah, one of the top 50 docs is missing 3 pages of tables and 2 pages of graphs just for a start :-) | 15:57.20 |
Robin_Watts | But I'd count lineheight problems as being worse than the occasional missing image. | 15:57.23 |
| but maybe the missing images are easier to solve, where the lineheight stuff is an infinitely deep rabbit hole. | 15:57.48 |
| damn. I thought I had this cleaning stuff parked, but it's fighting back. | 15:58.29 |
pedro | yup - I'm just concerned by the type of failure (works ok if you re-sav the ooxml docs *not* in compatibility mode - that's potentially likely to show a cause a lot of problems | 15:58.29 |
| think my wireless keyboard is losing way too many keystrokes :( | 15:59.19 |
kens | I have that problem too (or at least I blame the keyboard) | 16:00.56 |
pedro | ;) | 16:01.05 |
Jogux | my wired keyboard just keeps on producing wrong letters. and words. | 16:03.20 |
Robin_Watts | sad f dsfsd f assad df sdf | 16:05.05 |
Jogux | conjectures that Robin_Watts now has a cat | 16:05.28 |
Robin_Watts | half life of cats in this house would be measured in minutes. | 16:05.48 |
kens | robin using MSVC again :-) | 16:06.13 |
paulgardiner_lap | pedro: I want to build SO for Windows. Any source changes I need and/or host installs and configurations? | 16:19.38 |
pedro | paulgardiner: yep - it'll be on casper in 2 minutes | 16:31.29 |
paulgardiner_lap | pedro: magic thanks | 16:31.56 |
pedro | do you havewin7/vs2010? | 16:32.25 |
paulgardiner_lap | Yes. I might see if I can tease vs2005 on W8 into life too, but w7/vs2010 is fine | 16:33.16 |
| Well w7 will be fine when evetually my shoulder recovers and I'm able to sit at my desk for longer than 5 mins without pain | 16:34.16 |
pedro | ah :( | 16:34.34 |
Jogux | introduces Paul to vnc ;-) | 16:35.18 |
chrisl | paulgardiner_lap: what you done to your shoulder? | 16:36.15 |
paulgardiner_lap | chrisl: Don't know. Just slowly became painful, like sciatica but upper body. Had Physio yesterday, so hoping it'll improve soon. It's fine in reclining chair though | 16:42.37 |
kens | Sounds like a frozen shoulder | 16:43.38 |
paulgardiner_lap | Jogux: that's a point. W7 to W8 I should be able to just use remote RDP or whatever its called | 16:43.53 |
chrisl | paulgardiner_lap: nasty, I hope it improves soon..... | 16:43.59 |
Jogux | paulgardiner_lap : er, yeah, that's even better :-) | 16:44.31 |
Robin_Watts | paulgardiner_lap: Remote desktop. | 16:44.36 |
| Windows versions < 8, you couldn't remote desktop into by default unless you had Pro, rather than Home. | 16:45.32 |
| You want to remote desktop from 8 to 7, right? | 16:46.01 |
| There is a hack you can do to enable remote desktop on Windows & Home. | 16:46.20 |
| s/&/7/ | 16:46.28 |
paulgardiner_lap | Let me take a look. I think I have Premium | 16:48.13 |
| Okay providing enabling it takes 5 mins or less :-) | 16:48.48 |
Robin_Watts | http://www.raymond.cc/blog/enable-remote-desktop-connection-in-windows-7-home-premium/ | 16:50.57 |
paulgardiner_lap | thanks but turns out I have professional | 16:52.39 |
pedro | paulgardiner_lap: ~pedro/dev/epage - the toolchain_vs2010 branch has the changes for vs2010/win7 | 16:54.24 |
| paulgardiner_lap: tgvbuild.py -debug -vs2010 --dispman-ue2fileviewer app/simple | 16:55.16 |
| paulgardiner: tgvbuild.py -debug -vs2010 --dispman-ue2fileviewer app/simple | 16:56.19 |
Robin_Watts | tor8: ping. | 16:56.20 |
| tor8: New version of the clean commit on robin/master | 16:56.34 |
| Testing it shows up something nasty though. | 16:56.42 |
| When we mutool clean and subset the pages, we correctly remove the named destinations that point outside of our new ranges. | 16:58.09 |
paulgardiner | pedro: thanks | 16:58.23 |
Robin_Watts | We do not however remove the links that point at these names. | 16:58.45 |
paulgardiner | likes his new W7 laptop although the graphics seem a bit slow | 16:59.06 |
Jogux | paulgardiner : :-) | 16:59.17 |
Robin_Watts | So when we next access those pages, we get errors through when we load the annots. | 16:59.48 |
paulgardiner | pedro: might the push have failed? I just pulled from your repo and have pete/w7toolchain at the same point as robin/master | 17:11.51 |
pedro | paulgardiner: seems to be in the log on casper | 17:12.29 |
| (git log I mean) | 17:12.35 |
| there's nothing in the w7toolchain branch (yet) btw | 17:14.45 |
Jogux | thinks Paul wants pete/toolchain_vs2010 | 17:15.25 |
| which is definitely there on casper for me | 17:15.31 |
pedro | nods | 17:15.36 |
Robin_Watts | So, you're doing different bits of development on lots of different branches. | 17:23.44 |
paulgardiner | Something odd going on here. I have only branches pete/master and pete/w7toolchain and no commits on either | 17:24.25 |
Robin_Watts | That's fine, but generally the way we work with gs and mupdf is that we never merge those down to the trunk. | 17:24.43 |
kens | maybe that -t thing marcosw had ot do yesterday paulgardiner | 17:24.48 |
paulgardiner | kens: oh could be | 17:25.09 |
kens | I can't remember what it was exactly | 17:25.21 |
Robin_Watts | We rebase them on master so that they are at the 'tip' of master, then commit the whole branch at once. | 17:25.25 |
pedro | that's for updating tags | 17:25.29 |
Robin_Watts | That way we get a nice linear history on master with no confusing merges etc. | 17:25.42 |
chrisl | If you've pulled branches from the remote, that does not automatically create "local" branches for you | 17:26.05 |
paulgardiner | chrisl: I'm viewing with gitk and looking at the remote branches | 17:26.48 |
chrisl | So you can see the branch <pete's repo>/<branch name> ? | 17:27.51 |
Robin_Watts | Where is petes epage repo on casper? | 17:28.38 |
pedro | ~pete/dev/epage | 17:28.57 |
Robin_Watts | ok, ta. | 17:29.06 |
kens | I've had enough of counting bits today, goodnight folks | 17:29.17 |
Robin_Watts | I see w7toolchain and master as being the same (4fe3e7c) | 17:29.45 |
| With toolchain_vs2010 being 1 commit on from that as 277d555 | 17:30.02 |
pedro | yup, ignore w7toolchain - there's nothing on it yet | 17:30.12 |
| Robin_Watts: yup | 17:30.28 |
chrisl | "Add support for vs2010 and vs2012 builds of dispman-ue2fileviewer......." | 17:30.59 |
paulgardiner | It may just be gitk playing up, It just gave an error about tag secsel. I rember that being a problem when I set up the mirror. There were several tags I had systematically rename | 17:31.40 |
chrisl | So, yeh, definitely stuff on that branch | 17:31.41 |
paulgardiner | Also my gitk is fairly old | 17:32.03 |
sporadicsoftwre | Hello. Is there anybody here that could assist me with a mupdf bug/issue? | 17:32.19 |
Robin_Watts | sporadicsoftwre: sure. | 17:32.31 |
| sporadicsoftwre: Interesting. You make a pdf reader based on MuPDF for blackberry ? | 17:33.46 |
sporadicsoftwre | I have code that allows the user to add annotations to a PDF. This works well. A user has recently complained that they cannot save annotations to a PDF with an owner password. I tried authenticating with the owner password but mupdf did not accept the correct password. | 17:34.54 |
marcosw | Jogux: I've installed kvm on one of the linux boxes in my garage and am trying to run kvm with the debian64ats01.img but the boot gets stuck: | 17:35.08 |
| [ ok ] Starting deferred execution scheduler: atd. | 17:35.10 |
| [ ok ] Starting system message bus: dbus. | 17:35.10 |
| [info] Setting NIS domainname to: picsel.com. | 17:35.10 |
| [....] Starting NIS services: ypbind[....] Starting MTA: | 17:35.10 |
| | 17:35.31 |
Robin_Watts | sporadicsoftwre: Creating annotations in an encrypted PDF... not sure we've tested that. | 17:35.54 |
marcosw | wait, it timed out and got past that with a failed message | 17:35.55 |
Robin_Watts | sporadicsoftwre: Where can I get the source code for your PdfReader please? | 17:36.07 |
marcosw | it's now Starting automount | 17:36.08 |
sporadicsoftwre | I looked at mupdf on android and it was able to save the file. mupdf on android is only at version 1.2 so I noticed that the difference is 1.2 does not use opts.do_incremental. If I remove that then I can save the PDF without using the owner password | 17:36.44 |
| I can send you the link for the source if you like | 17:37.05 |
Robin_Watts | That's because the password is removed on saving I think. | 17:37.08 |
| sporadicsoftwre: Looking on appworld.blackberry.com I can see no reference to the fact that it uses MuPDF. Nor can I see a statement of the fact it's licensed under the GNU GPL. | 17:38.04 |
sporadicsoftwre | It is present in the app itself. Do I need to update the App World page as well? | 17:38.42 |
Robin_Watts | It would be nice to :) | 17:39.24 |
marcosw | Jogux: the debia64-ats01 system booted and I'm now presented with a login: screen. What next? | 17:39.34 |
sporadicsoftwre | I will submit the request to BB, they will have to approve the change so it won't be immediate | 17:39.54 |
Robin_Watts | I don't have a blackberry myself, so I have no way of knowing that it's GNU GPL licensed without buying one, downloading it and running it :) | 17:39.57 |
| paulgardiner: Annotation question here... | 17:40.26 |
| (see the logs) | 17:40.31 |
pedro | marcosw> Jogux is cycling home - probably online in around 30 minutes or so. Guess its the root password you need next? | 17:41.05 |
sporadicsoftwre | True but my understand (and I am very likely wrong) is that the source only has to be given to users of the app upon request so i though the license could be in the app itself | 17:41.11 |
Robin_Watts | You may be right. | 17:41.55 |
| We are a bit sensitive to this as lots of people use MuPDF and never acknowledge it anywhere. | 17:42.33 |
sporadicsoftwre | Understandable. I will update App World regardless | 17:43.06 |
marcosw | pedro: that would helpful, though even with that I'm not sure what's next. Getting kvm running one of the ats images was what I was trying to do and I've done that :-) | 17:43.14 |
Robin_Watts | sporadicsoftwre: Thanks. | 17:43.28 |
| sporadicsoftwre: paulgardiner is the annotation expert. | 17:43.59 |
| but I think that if you don't set opts.do_incremental, we completely rewrite the file upon saving, and passwords etc are all removed. | 17:44.23 |
paulgardiner | I'm just taking a look | 17:44.24 |
Robin_Watts | Hence if you don't set opts.do_incremental, you'll never get prompted for a password, and the problem won't occur. | 17:45.02 |
sporadicsoftwre | And mupdf doesn't "care" about the owner password to begin with? | 17:45.05 |
Robin_Watts | AIUI, you're opening a file, giving it the password, and then editing it. | 17:45.42 |
| Is that right? | 17:45.49 |
pedro | marcosw> cool :) Guess we can manually check the builds on each VM and build there, but it'll really need the ATS server set up to get the real value | 17:45.57 |
Robin_Watts | So MuPDF is 'bad' in that we are effectively stripping the password off the file as we save it. | 17:46.10 |
sporadicsoftwre | That as the idea but mupdf does not accept the owner password. The same code works for a user password | 17:46.31 |
| I added checking PDF permissions and requesting the owner password as required but pdf_authenticate_password dos not accept the owner password | 17:47.14 |
Robin_Watts | sporadicsoftwre: Sorry, maybe I'm being thick here. | 17:47.32 |
| So open the file in MuPDF. It asks for a password. You give it the owner password and it's not accepted? | 17:47.53 |
pedro | goes for the train - biab | 17:48.12 |
Robin_Watts | This is a file before you've made any changes to it ? | 17:48.18 |
sporadicsoftwre | When I say mupdf I just mean the library. So, yes, I check and see the user does not have permissions to make changes or notes so I request the owner password. I use pdf_authenticate_password but that does not return successfully and the permissions have not changed. This is before any changes are attempted | 17:49.27 |
| I am using fz_meta to check permission FWIW | 17:50.08 |
Robin_Watts | sporadicsoftwre: This is a file that mupdf has had nothing to do with writing ? | 17:51.58 |
sporadicsoftwre | Robin_Watts: Correct | 17:52.14 |
Robin_Watts | So what does this have to do with annotations? | 17:52.24 |
sporadicsoftwre | I am allowing the user to add annotations but per the PDF permissions they cannot without the PDF owner password | 17:53.05 |
Robin_Watts | Right. So the bug report is actually "MuPDF will not accept the owner password on a file". | 17:53.52 |
sporadicsoftwre | I think so | 17:54.03 |
Robin_Watts | Sorry, I'd got sidetracked by the mention of annotations. | 17:54.08 |
sporadicsoftwre | My apologies, I was just trying the explain the whole situation | 17:54.22 |
Robin_Watts | That does sound like a bug. Can you go to bugs.ghostscript.com please, and open a bug. | 17:54.26 |
| Attach a file that shows the problem, and give us the user/owner passwords in the bug description. | 17:54.55 |
| Then we have have a look. | 17:55.04 |
sporadicsoftwre | Will do. Thank you for your assistance! | 17:55.08 |
Robin_Watts | No problem. | 17:55.15 |
| IIRC the exact encoding in which we expect passwords differs a bit from version to version of PDF. | 17:55.54 |
| do user and owner passwords have any 'strange' chars in them ? | 17:56.16 |
sporadicsoftwre | No, the PDF I am testing as no user password (its hat the issue?) and the owner pass is test | 17:56.46 |
Robin_Watts | sporadicsoftwre: I can't say until I (or one of my colleagues) has had a chance to look. | 17:57.27 |
sporadicsoftwre | FYI I opened bug 695101. Thanks again | 18:01.46 |
Robin_Watts | Thanks. | 18:01.59 |
sebras | 2/msg Robin_Watts hm... | 18:33.37 |
Jogux | marcosw : sorry, I've been away from my computer - let me catch up | 18:36.51 |
| marcosw : okay, that sounds awesome. I think there are two things to do now | 18:40.20 |
| 1) carry on and get all the other vms to boot | 18:40.27 |
| 2) let me and mace know how we can access these systems | 18:40.38 |
Robin_Watts | ooh. It seems we have mujs.com | 18:43.59 |
sebras | Robin_Watts: are you sure? whosis seems not be updated. | 18:47.46 |
Robin_Watts | sebras: No, Miles just mailed to say he's agreed the deal. | 18:48.03 |
| It'll take a while to actually come through. | 18:48.12 |
Jogux | robin_watts : awesome :) | 18:52.58 |
marcosw | Jogux: giving someone access to the systems is going to take some reconfiguration on my part. My home machines are all behind a firewall but there is little (none, really) security behind it. | 18:54.53 |
Jogux | marcosw : ah. oh. er. | 18:55.15 |
Robin_Watts | marcosw: We can ssh into the linux personas of those machines, right? | 18:55.57 |
marcosw | also the debian64-ats01 system did not request an ip address from my dhcp server, so presumably I have to login via the remote console and change the ip address to system in my local space (or figure out why networking isn't working). | 18:56.15 |
Robin_Watts | Or into at least 1 of those machines? | 18:56.22 |
Jogux | marcosw : oh. damn. | 18:56.40 |
| marcows : security wise, your worry is essentially that if you give me/mace access to this system, we also end up with access to all your personal systems? | 18:57.38 |
Robin_Watts | If we can get into one of the linux boxes, then we can run an ssh tunnel to get into anything behind the firewall. But it obviously does require you to get networking working :) | 18:57.43 |
marcosw | robin_watts: in theory yes, but I think they have my .ssh private key on them (and currently I can only get qemu to run under sudo). Both these issues are fixable. | 18:58.08 |
Jogux | ah | 18:58.43 |
Robin_Watts | marcosw: I wonder if it's worth configuring all the VMs onto a different subnet. | 18:58.45 |
marcosw | Jogux: yeah. if I remove the my private key and setup a whole in my firewall that should be okay. | 18:58.45 |
| ^whole^hole | 18:58.54 |
Jogux | marcosw : okay, great, that sounds good | 18:59.07 |
marcosw | also, I'm running qemu under curses, since I couldn't get it to work with X (the error message was "Could not initialize SDL(No available video device) - exiting"). | 19:00.22 |
| google suggests it's a missing library. | 19:00.37 |
Jogux | that sounds absolutely fine - we don't need anything graphical | 19:00.43 |
marcosw | that's what I figured. | 19:00.54 |
Jogux | afaik these boxes won't even have gdm/x11/installed. | 19:01.14 |
| oh. except windows. gah. | 19:01.23 |
marcosw | what about the windows? | 19:01.27 |
Jogux | I hate windows. | 19:01.32 |
marcosw | I knew I liked you for a reason :-) | 19:01.42 |
Jogux | it'll have rdp setup, so if we can figure out the ips we should be okay. | 19:01.51 |
Robin_Watts | Jogux: Presumably you can remote desktop into the windows boxen? | 19:01.53 |
Jogux | marcosw : hehe | 19:02.12 |
Robin_Watts | If we put all the picsel stuff onto a different subnet, we can probably open an ssh tunnel to henrys macpro. | 19:02.37 |
marcosw | Does windows boot without a display? I guess it must, since amazon aws has windows images. | 19:02.53 |
Jogux | marcosw> good question. actually I believe kvm will expose the raw console via vnc too. | 19:03.18 |
| so it must be faking up displays | 19:03.34 |
marcosw | how fast a mac do we need? I have a spare imac that isn't doing anything. I also have a macpro, but that's my personal machine and so I'm hesitant to use it as an ats machine (it's currently a cluster node, but that's pretty low impact). | 19:04.22 |
Jogux | marcosw : I think anything will be fine for now. | 19:04.43 |
| I assume it can run mavericks? | 19:04.49 |
marcosw | Jogux: I think it's running Mountain Lion, so Mavericks will work (afaik Mavericks runs on all the machines that Mountain Lion supports). | 19:06.40 |
Jogux | sounds right. actually ML is probably fine; iirc xcode 5.1 runs on it. | 19:07.07 |
ray_laptop | Jogux: I have a mac mini as well (I'd have to upgrade the OS). It's 5+ years old, but it belongs to Artifex | 19:07.23 |
Robin_Watts | 5+ years old won't run mavericks, I believe. | 19:08.29 |
| (or ML) | 19:08.44 |
| Actually, maybe I'm being unfair. I have a (late) 2009 macbook and that runs mavericks. | 19:09.21 |
marcosw | robin_watts: apple orphaned some machines with Mountain Lion that weren't very old. | 19:13.19 |
Robin_Watts | mac mini Early 2009 or later required | 19:14.57 |
marcosw | according to everymac.com the MacMini Early 2009 is the latest one that will run Mountain Lion (or Mavericks). | 19:14.58 |
Robin_Watts | So 5 years old is OK, 5+ is not :) | 19:15.06 |
| tor8: 2 commits on robin/master | 19:15.56 |
| henrys, paulgardiner, tor8: So, assuming tor8 OKs these commits, I think I'm in a position to park the mupdf work I'm doing. | 19:16.48 |
Jogux | stops for food. biab. | 19:16.50 |
Robin_Watts | This will leave Raed without his desired Bates numbering. | 19:17.05 |
| This will leave Christophe without his watermarking. | 19:17.19 |
| but the step on the way to those things has been done. | 19:17.35 |
marcosw | my imac is the Mid-2007 model and runs Mavericks (as does my 2008 MacPro). | 19:17.41 |
Robin_Watts | marcosw: Have you considered asking Miles to pay for your garage to be fortified? :) | 19:18.16 |
marcosw | agrees with Jogux that lunch sounds good | 19:18.24 |
| robin_watts: you mean for security? :-) If someone breaks into my garage a bunch of 4 year old dell servers is not going to be what they take. | 19:19.12 |
Robin_Watts | henrys, paulgardiner, tor8: The one thing that I'm a little unhappy about leaving in the current state is that we never compress streams before we write them out. | 19:19.18 |
| marcosw: Oh, Porsche boy, right. | 19:19.27 |
| This means that whenever we clean files and write them out, they get much larger. | 19:20.29 |
| Also, whenever we make appearance streams they end up written uncompressed too. | 19:20.50 |
| Is it worth me spending a day or so extending the pdf-write.c stuff so that it can optionally deflate streams on writing? | 19:21.24 |
| tor8: Oh, ass. Something I forgot. My fz_printf changes had deliberately made it so that output to both buffer and file went through the new printf stuff so we got equivalent results for both. | 19:23.10 |
| I fear that your new commit doesn't do that. | 19:23.24 |
| oh, ignore me, it does. | 19:24.28 |
Jogux | is bakc. | 19:50.43 |
Jogux | is back. | 19:50.46 |
henrys_ | Robin_Watts: can tor8 take that over? (compression) | 20:21.14 |
Robin_Watts | absolutely. | 20:21.32 |
henrys_ | or it's a good bounty right - fairly isolated. | 20:21.36 |
| ? | 20:21.48 |
Robin_Watts | also, there is a bug of mvrhel's about bouncy text. | 20:21.56 |
| but basically I can move to SmartOffice for tomorrow. | 20:22.57 |
| I am tempted to finish the kerning fixes to the fonts. | 20:23.09 |
| It's not going to make a huge difference, but it means we'll be able to be sure that we have got equivalent metrics in all cases. | 20:23.53 |
| and hopefully by the time I've finished that, we'll have ATS up and running, and I can work with paulgardiner/Jogu/pedro to figure out what to do next where we won't all be tripping over each other. | 20:24.35 |
henrys_ | Robin_Watts: sounds good to me. | 20:39.38 |
djcreedy | Hey, was wondering if any of the MuPDF devs were around? | 20:40.45 |
Jogux | djcreedy : there's a couple hanging around | 20:50.17 |
Jogux | finally gets this VM to boot. yay. | 20:50.32 |
Jogux | has no idea what the logins for the windows boxes are. | 20:55.48 |
Robin_Watts | djcreedy: On irc, it's always best to ask your question, not ask to ask :) | 21:07.10 |
djcreedy | Robin, this is true. | 21:07.45 |
| I just wanted to ask about MuPDF commercial licensing, and whether it was a per-app bases or perpetual single license. And also if someone could give a ball park price for it. | 21:08.27 |
Robin_Watts | We don't deal with commercial licensing on here, we're just engineers, so we point people to Scott Sackett (sales@artifex.com) for that. | 21:09.24 |
| But, I am 99% sure that we don't do perpetual single licenses. It's always a per year (or quarter) figure. | 21:09.49 |
| There might be revenue sharing options - you'd have to ask Scott. | 21:10.14 |
| I can't give a ballpark figure, cos our licenses are tailor made to what people want. | 21:10.44 |
djcreedy | Awesome. Appreciate you taking the time to respond mate. I'll reach out to Scott. | 21:12.25 |
Robin_Watts | no worries. | 21:12.32 |
tor8 | Robin_Watts: two commits on robin/master LGTM | 21:50.05 |
| do you want the third one in as well? | 21:50.21 |
zeniko | Robin_Watts: are rendering changes to be expected from the "fix operator buffering of inline images" patch? | 22:00.07 |
| Forward 1 day (to 2014/03/20)>>> | |