| | <<<Back 1 day (to 2014/01/30) | 2014/01/31 |
| ray_laptop | really early start for me today | 13:50.25 |
| kens | I see that ;-) | 13:50.35 |
| ray_laptop | Robin_Watts: thanks for the prompt review yesterday | 13:50.55 |
| Robin_Watts | ray_laptop: No worries. | 13:51.01 |
| ray_laptop | hmm... looks like I need to goose ghostbot | 13:52.02 |
| | the logs are empty | 13:52.39 |
| kens | Nobody said anythign this am | 13:52.56 |
| Robin_Watts | no one said anything. | 13:53.01 |
| ray_laptop | really ? Is it a holiday, then ? | 13:54.36 |
| | by now yu guys usually have several pages filled with complaints about your "favorite" customers ;-) | 13:55.42 |
| Robin_Watts | ray_laptop: National caffeine shortage. | 13:56.08 |
| ray_laptop | but I see that ghostbot has updated the logs now. | 13:56.28 |
| Robin_Watts | The government has sent troops into the west country to evaluate if it's possible to make tea with the floodwater. | 13:56.35 |
| ray_laptop | Robin_Watts: that sounds dire | 13:56.41 |
| Robin_Watts | ray_laptop: Yeah. They may decide to go with Earl Grey :( | 13:57.16 |
| ray_laptop | well, I'll be quiet and let you all get back to sleep ;-) | 13:58.46 |
| Robin_Watts | http://natashenka.ca/posters/ | 13:59.18 |
| jhabjan | hi there, i'm trying to force 'inkcov' device (via api) to write the output to the stdout but it seems it does not work as expected | 14:18.18 |
| | switches that i'm using are: -o- -sDEVICE=inkcov test.pdf | 14:20.58 |
| chrisl | That seems to work for me | 14:22.03 |
| jhabjan | stdio handler is set correctly by using "gsapi_set_stdio"... ( tested and it works for sure ) | 14:22.05 |
| ray_laptop | jhabjan: works for me, too | 14:22.35 |
| jhabjan | the problem i'm having is that all messages like 'Page 1', 'loading font...' are sent to stdio properly.. | 14:22.46 |
| | but the CMYK message is not | 14:22.51 |
| chrisl | Oh, but it probably won't go through the stdout handler, but direct to stdout | 14:23.07 |
| ray_laptop | jhabjan: I recommend -q or -sstdout=otherstdout.txt Your output may be lost in the noise | 14:24.02 |
| jhabjan | i did try -q | 14:24.40 |
| | and also -sOutputFile=%stdout | 14:24.44 |
| | and only -sOutputFile=test.txt works | 14:24.59 |
| | but... that would mean that i need to let everything finish and read out the txt file from the disk | 14:25.29 |
| chrisl | ray_laptop: I assume that output from the device's "write_page" method (using the FILE * parameter) won't ever get redirected to the stdout handler in the API? | 14:25.33 |
| ray_laptop | hmm something is amiss. When I do: debugbin/gswin32 -sDEVICE=inkcov -o- -sstdout=x examples/tiger.eps I get: /unknownerror in --showpage-- | 14:26.40 |
| | gswin32c works as on linux. I did this to test because gswin32 relies on the stdio hooks | 14:27.09 |
| jhabjan | i'm using windows version... | 14:27.21 |
| ray_laptop | jhabjan: gswin32c.exe or gswin32exe ? | 14:27.51 |
| jhabjan | api | 14:27.56 |
| | gswin64.dll | 14:28.07 |
| ray_laptop | jhabjan: oh. You said that, sorry | 14:28.08 |
| jhabjan | no worries | 14:28.13 |
| chrisl | ray_laptop: I suspect a problem is that the inkcov output is written directly through fprintf() | 14:28.35 |
| ray_laptop | So you must have a console mode app, or you would probably get the same error as gswin32.exe | 14:28.51 |
| jhabjan | yep... I'm trying to implement inkcov in the Ghostscript.NET library... and testing via console app | 14:29.21 |
| ray_laptop | chrisl: right. Actually, I think all devices tend to use fwrite and fprintf | 14:29.24 |
| chrisl | ray_laptop: exactly, so none of them will output through the API stdout handler | 14:29.49 |
| ray_laptop | devices are passed a FILE * | 14:29.54 |
| | jhabjan: so, I guess you can't get there with the stdio hooks. Sorry. If you want, you can open an enhancement bug, but don't hold your breath | 14:30.51 |
| jhabjan | hm... | 14:31.21 |
| | would it work via pipe ? | 14:31.28 |
| chrisl | Yeh, a pipe would work | 14:31.38 |
| jhabjan | ok... that would be an alternative solution.. | 14:31.51 |
| | thanks for your help | 14:32.54 |
| chrisl | Sorry we weren't more help.... | 14:33.07 |
| ray_laptop | jhabjan: chrisl: actually, I tried it and it failed: debugbin/gswin32 -sDEVICE=inkcov -o'| cat ' -sstdout=x examples/tiger.eps | 14:33.13 |
| | note that my system has 'cat' | 14:33.35 |
| | or did I get the syntax wrong. I don't use pipes often | 14:34.17 |
| chrisl | Huh, that command line works for me on Linux.... | 14:34.19 |
| jhabjan | on windows, for ghostscript i use pipe this way: -o%handle%1234 | 14:36.38 |
| ray_laptop | note that the error message i get is now: /ioerror in --showpage-- (not /undefinederror) | 14:37.09 |
| jhabjan | where 1234 is pipe client handle | 14:37.11 |
| ray_laptop | jhabjan: yuck! But as long as it works for you. | 14:37.45 |
| jhabjan | well... that was the only way I could make pipe communication with the ghostscript | 14:38.15 |
| | i did try other suggested ways... but only this proved to work | 14:39.09 |
| ray_laptop | jhabjan: well, thanks. But the fact that the | doesn't work as documented I _do_ consider a bug. gp_mswin has gp_open_printer that is _supposed_ to work | 14:40.33 |
| | the mswin_popen has a bunch of worrisome comments about things not working, "so we use our own" | 14:41.55 |
| chrisl | Which may, of course, no longer be true (the things not working) | 14:42.33 |
| ray_laptop | I _hate_ debugging windows code. I think I'll leave it for mvrhel :-) | 14:42.38 |
| | chrisl: right. It mentions MSVC5 and 6 | 14:43.01 |
| chrisl | I think it's fair to say those had a *lot* of problems, even for VC releases! | 14:43.29 |
| ray_laptop | hmm... looks like Robin_Watts was the last person in there (UTF8 changes) | 14:45.46 |
| | I wonder if it worked before that ? (June 2011) | 14:46.40 |
| jhabjan | well, I will give a try to output inkcov to a pipe on my machine (by handling pipe my way)... | 14:46.44 |
| | and if it will not work, I'll go with standard file output | 14:47.08 |
| | i did a test... | 14:51.56 |
| | output to a pipe works in my case | 14:52.07 |
| ray_laptop | jhabjan: thanks for the test. At least we have a workaround for folks. And it may help us fix mswin_popen | 14:52.44 |
| henrys | another 6 inches of snow, this is getting tiresome | 14:53.12 |
| jhabjan | n/p, so this actually works: '-q -o%handle%390 -sDEVICE=inkcov test.pdf' | 14:55.10 |
| kens | Robin_Watts : tor8 could you try a file for me with the current MuPDF pelase ? | 14:58.28 |
| Robin_Watts | sure. | 14:58.33 |
| kens | Bug #694991 | 14:58.40 |
| | The file displays as expected in GS and Acrobat, but MuPDF (older version) shows a black bar at the bottom | 14:59.10 |
| | I suspect this is a transparency problem | 14:59.28 |
| Robin_Watts | I see a black bar at the bottom. | 15:00.20 |
| | after row12 | 15:00.26 |
| kens | Hmm interesting. | 15:00.26 |
| | Acrobat does not show that | 15:00.31 |
| | Nor does GS | 15:00.36 |
| | I'll have to dig deeper, don;t fancy debugginb 50MB of PostScript | 15:00.58 |
| Robin_Watts | Acrobat shows a dead fly in the left hand corner. | 15:01.04 |
| kens | I don;'t see that either, which left corner top ort bottom ? | 15:01.26 |
| Robin_Watts | bottom corner. | 15:01.35 |
| kens | Not for me on Acrobat X | 15:01.43 |
| Robin_Watts | I suspect that there is a transparency thing in there. | 15:02.00 |
| henrys | Robin_Watts, paulgardiner how difficult is it to build an x86 api? | 15:02.16 |
| Robin_Watts | a plane of all transparent, and that is corrupt or truncated. | 15:02.17 |
| kens | Ah Acrobat XI does shwo squashed thing | 15:02.20 |
| henrys | s/api/apk | 15:02.21 |
| Robin_Watts | henrys: For android? | 15:02.36 |
| kens | Its hte word Wafte | 15:02.37 |
| | ZOom in an its visible | 15:02.44 |
| Robin_Watts | or of the ghostdocs demo ? | 15:03.00 |
| | Doing an x86 build of the android app is easy. | 15:03.21 |
| henrys | Robin_Watts: yes, I ask because genymotion might be a good way for us to test. Apps seem to run very fast, full screen on a pc, etc. | 15:03.26 |
| | Robin_Watts, paulgardiner I'm using it now it looks good for the built in apps, have you guys used that? | 15:06.22 |
| Robin_Watts | Never heard of it. | 15:06.32 |
| paulgardiner | Is that easier than a plain Windows exe? | 15:06.40 |
| | If you want a PC viewer, there was a build called unidisp-test that produced one. | 15:07.50 |
| | Not sure it was kept in a working condition. | 15:08.09 |
| Robin_Watts | What problem do we think this is solving? | 15:08.16 |
| henrys | Robin_Watts, paulgardiner :teasing on the platform it will probably run on. | 15:09.25 |
| | s/teating/testing | 15:09.33 |
| paulgardiner | Visual testing? | 15:10.04 |
| Robin_Watts | Sorry, this is still not clear to me. | 15:10.04 |
| | Are you talking about testing the MuPDF android app? | 15:10.15 |
| | Or are you talking about testing the GhostDocs converter? | 15:10.30 |
| henrys | GhostDocs converter | 15:10.47 |
| Robin_Watts | henrys: Right, so why can't we build that for windows and test it there? | 15:11.05 |
| | hell of a lot easier to debug it like that. | 15:11.11 |
| | The third option I was going to give was "MuPDF-Office", the thing where we integrate the converter properly into MuPDF. | 15:12.03 |
| paulgardiner | henrys: when I added the pdfexport feature to the ghostdocs test-shell, I was using the Windows build | 15:12.12 |
| henrys | paulgardiner: oh I had no idea any of that was available, if that is all set up then fine. | 15:12.54 |
| paulgardiner | The test-shell can give us pdf or pre-export bitmaps | 15:12.58 |
| Robin_Watts | henrys: The FBS is a pain in the ass, but it can at least build most things for a wide range of platforms. | 15:13.26 |
| paulgardiner | henrys: oh okay. I understand now. | 15:13.41 |
| chrisl | ray_laptop: http://git.ghostscript.com/?p=user/chrisl/ghostpdl.git;a=commitdiff;h=c9e52b18 | 15:13.51 |
| henrys | Robin_Watts: okay I am stuck why genymotion or I have to go buy a device. The tiny nexus I have is useless. | 15:14.11 |
| paulgardiner | henrys: it should build for linux too | 15:14.15 |
| henrys | paulgardiner: so can you release something I can build. | 15:14.32 |
| | ? | 15:14.33 |
| Robin_Watts | henrys: I've made the FBS know how to build executables using the ndk targeted for arm on windows. | 15:15.29 |
| henrys | paulgardiner: you only released an apk so I assumed under time pressure you had cobbled just enough together to do an android build. Sorry I should have paid more attention. | 15:15.34 |
| paulgardiner | henrys: np. I didn't make it clear | 15:15.56 |
| Robin_Watts | if you want me to make something that will build executables using the ndk targetted for x86 on windows, that's a bit more work, but not too bad. | 15:16.17 |
| henrys | Robin_Watts: yes I'll follow those instructions, but I also want to build paulgardiner stuff | 15:16.35 |
| paulgardiner | The version from my and Robin's repos on casper will build the pdf exporting version using the build command from Robin's email with the extra -define=PDF_EXPORT flag | 15:17.05 |
| Robin_Watts | If you want me to make something that will build executables using the ndk targetted for x86 on macos, that's potentially a lot more work :) | 15:17.29 |
| paulgardiner | Robin_Watts: sounds like we don't need an x86-targetted android build | 15:17.36 |
| Robin_Watts | henrys: It's not just a question of following the instructions. | 15:17.46 |
| | Currently we cannot build x86 targetted android exes | 15:18.16 |
| | paulgardiner: I'm confused again then. | 15:18.28 |
| paulgardiner | Maybe its me that's confused. | 15:18.45 |
| henrys | Robin_Watts, paulgardiner so maybe it is best for laypeople of the staff to have a windows exe of smart office proper and an exe of paul's stuff. | 15:18.46 |
| | laypeople like me. | 15:18.56 |
| Robin_Watts | henrys: Step 1: Everyone should buy Smart Office from the app store. | 15:19.15 |
| henrys | Robin_Watts, paulgardiner I'd like to have something to run on a large screen. | 15:19.23 |
| Robin_Watts | That way people can stop asking me "does it do XXXX?" | 15:19.25 |
| paulgardiner | henrys: you want a viewer or convertor? | 15:19.54 |
| Robin_Watts | henrys: buy a Nexus 10, or something akin to it. | 15:19.58 |
| henrys | Robin_Watts:okay fair enough | 15:20.20 |
| Robin_Watts | henrys: I've been talking to people about the Office mismatches. | 15:20.41 |
| | Apparently it's something that customers always complain out. | 15:20.59 |
| henrys | Robin_Watts: you mean the code not working ? ;-) | 15:21.20 |
| paulgardiner | henrys: we definitely have the convertor building for windows. The FBS used to support building a Windows viewer, but I haven't checked that still works | 15:21.36 |
| Robin_Watts | And their standard answer is "Office can't agree what the layout of these files should be between different versions. How can we be expected to?" | 15:21.41 |
| | seriously, if it was going to be easy to fix, it would have been fixed. | 15:22.38 |
| paulgardiner | Only problem: I think we have a competitor that advertises how closely their product's output matches Office. | 15:22.47 |
| henrys | Robin_Watts: with the buyout I would have expected a document describing this stuff or at least a bug database record like we have. | 15:22.50 |
| Robin_Watts | There was a bugzilla at Picsel. | 15:23.08 |
| | but we didn't get a copy of the database for that with our source license. | 15:23.24 |
| | There was similarly a wiki, but we didn't get that. | 15:23.54 |
| henrys | Robin_Watts: might that tell us a lot, or do you think the signal to noise ratio too low to to be useful | 15:23.54 |
| Robin_Watts | honestly, I don't know. | 15:24.11 |
| henrys | Robin_Watts: we can certainly ask for that stuff. | 15:24.14 |
| Robin_Watts | The build commands for all the different things would have been buried in the ATS database, but again, we didn't get a copy of that. | 15:25.05 |
| | The liquidators cited customer secrets when we asked for that. | 15:25.29 |
| | so I suspect they will say the same about the bugs. | 15:25.39 |
| henrys | paulgardiner, Robin_Watts we really do need the app running on a PC I believe for marcos to do extensive testing. It will take ages to look at this stuff on a tablet. | 15:25.45 |
| Robin_Watts | henrys: Why would he be testing using the app? | 15:26.06 |
| kens | Robin_Watts : dumb question, does MuPDF handle transfer functions yet ? | 15:26.08 |
| paulgardiner | henrys: a viewer, or to-pdf/bitmap convertor? | 15:26.10 |
| Robin_Watts | kens: no. | 15:26.14 |
| kens | OK then that's what the problem is | 15:26.22 |
| | THe file sets transfer functions | 15:26.30 |
| Robin_Watts | henrys: I'm sure we've been through this before. | 15:26.45 |
| kens | Looks like Acrobat applies htem | 15:26.47 |
| Robin_Watts | I can give him a windows exe that he can call that will 1) Load files in and spit bmps out. | 15:27.09 |
| | 2) Load files in and spit pdfs out. | 15:27.19 |
| | That should be enough for him to do compares etc, right? | 15:27.36 |
| | No need for him to mess with the app. | 15:27.42 |
| paulgardiner | Possibly 3) shows document on screen | 15:27.45 |
| henrys | paulgardiner: well both because we need to understand the source of bugs - but Robin_Watts has recently discovered it looks like the bugs are not pdf generation | 15:27.49 |
| | Robin_Watts: yes that will be fine. | 15:28.16 |
| paulgardiner | Some of the bugs are pdf generation. The odd things missing like underlining | 15:28.25 |
| henrys | Robin_Watts: send a link to tech | 15:28.35 |
| Robin_Watts | henrys: yeah. But people can see whether the problems are pdf generation or not by buying a copy of SmartOffice and trying it for themselves. | 15:28.54 |
| | paulgardiner: Ah. I hadn't twigged about the underlining, but you may be right. | 15:29.35 |
| | That kind of thing might be fairly simple to solve. | 15:29.42 |
| | henrys: Send what link to tech? | 15:30.07 |
| paulgardiner | henrys: I'm still a bit confused as to what we need, but if we need a windows viewer I can look at whether unidisp-test still builds. | 15:30.12 |
| Robin_Watts | paulgardiner: For marcos to test, I'm sure he'd rather have something scriptable. | 15:30.49 |
| paulgardiner | Robin_Watts: not if he's doing loads of visual testing | 15:31.16 |
| henrys | I don't know if I agree with Robin_Watts about viewing on a tablet. There is a huge difference to me looking at a desktop size screen vs. a tablet. | 15:31.34 |
| ray_laptop | can't easily script getting output from the reference Office to compare to | 15:32.03 |
| Robin_Watts | paulgardiner: I bet he will. He is well set up for quickly comparing bitmaps that come out of the other tests. | 15:32.25 |
| | ray_laptop: I thought he'd said he had that scripted already. | 15:32.36 |
| henrys | but maybe I need glasses. | 15:32.38 |
| paulgardiner | Robin_Watts: oh yeah I guess | 15:32.46 |
| henrys | ray_laptop: yes marcosw has that end of it fixed AIUI | 15:33.05 |
| Robin_Watts | henrys: If you want a version of 'a smart office based document viewer' on the desktop, then paulgardiner can probably get you a unidisp viewer build. | 15:33.30 |
| chrisl | ray_laptop: this should resolve the i/o thing with inkcov, if you think it's okay: http://git.ghostscript.com/?p=user/chrisl/ghostpdl.git;a=commitdiff;h=1598652e | 15:34.13 |
| henrys | Robin_Watts: the link I was referring to was the windows exe you are going to provide marcos. | 15:35.19 |
| Robin_Watts | right. | 15:35.27 |
| henrys | paulgardiner: not a high priory thing I can use the app on my iPhone for now. | 15:36.03 |
| | paulgardiner: but it does seem it would be easier for everyone to engage with the code. | 15:36.39 |
| paulgardiner | henrys: okay. I'll have a look when I get the chance. It may just work | 15:36.41 |
| ray_laptop | chrisl: this looks like a typo "that way, it can be been by" | 15:37.20 |
| chrisl | Right: "...seen by...." | 15:37.41 |
| ray_laptop | chrisl: so this is intended to be an example for everyone to use in devices ? | 15:37.47 |
| henrys | bad snow day ⦠I have cars in my yard | 15:38.22 |
| chrisl | ray_laptop: not really - I doubt most people want actual raster data dumped out through stdout/stderr handlers | 15:38.25 |
| ray_laptop | because it looks like you have to do it in any device that might want to go to stdout when the stdio handlers are set | 15:38.53 |
| Robin_Watts | henrys: unwanted ones? | 15:38.53 |
| henrys | yes 2 apparently lost control didn't hit the house thank god | 15:39.17 |
| ray_laptop | henrys: put prices on them. You might make some money ;-) | 15:39.23 |
| kens | chrisl gsprint dumps rasters via a pipe :-) | 15:39.26 |
| chrisl | ray_laptop: yes we would do that - but like I say, I don't think it makes sense for raster data. Maybe txtwrite is the other one that might want it | 15:39.41 |
| Robin_Watts | henrys: urk. | 15:39.42 |
| kens | But stdout wouldn't be good as it would get all the other stdout stuff in there | 15:39.42 |
| ray_laptop | kens: that's what -sstdout is for | 15:39.54 |
| paulgardiner | henrys: Not sure if you get episodes of Horizon in the US, but they've shown another very interesting dietry one. I could compress it and upload it somewhere if you are interested. | 15:40.09 |
| ray_laptop | all of the "other" stdout then goes the the file | 15:40.20 |
| henrys | paulgardiner: more fasting? | 15:40.28 |
| kens | ray_laptop : well, then there's a use case for rasters on stdout :-) | 15:40.45 |
| paulgardiner | No, sugar vs fat | 15:40.48 |
| kens | I watched that one | 15:41.00 |
| ray_laptop | kens: right. It's ofen done in uniz scripts to have gs work as a filter | 15:41.18 |
| henrys | paulgardiner: ah yes I'd love to see that⦠I've always had a theory about that. fat's fine sugar's not but I'd like to see it. | 15:41.21 |
| kens | henrys not what the program says :-) | 15:41.34 |
| chrisl | kens, ray_laptop: but would any integrator using the API stdout/stderr handlers really want raster data via those? | 15:41.49 |
| kens | but its a pop science program so don't expect too much | 15:41.52 |
| paulgardiner | henrys: yes as ken says results are very surprising | 15:42.05 |
| kens | chrisl I have no idea. But given that Russell used a pipe it wouldn't surprise me if he ha dopted for stdout instead | 15:42.17 |
| ray_laptop | chrisl: yes. We've had customers doing that so that they can operate without writing anything to a file | 15:42.28 |
| paulgardiner | henrys: I'll set it up to transcode over night | 15:42.30 |
| henrys | paulgardiner: oh thanks | 15:42.43 |
| ray_laptop | (paranoid about security, or something) | 15:42.52 |
| kens | paulgardiner : I recall form somewhere (xkcd ?) '"eat food*, not too much, variety" *if it has to tell you how wonderful for you it is, its not food. | 15:43.11 |
| chrisl | ray_laptop: but surely there would be a better way to do that? | 15:43.13 |
| kens | A pipe springs to mind | 15:43.32 |
| chrisl | Or just write a simple device that keeps the raster in memory | 15:43.52 |
| ray_laptop | kens: right, but pipes appear to currently be broken in windoze | 15:43.55 |
| kens | ray_laptop : surely not! gsprint works that way | 15:44.06 |
| | and it works for me | 15:44.13 |
| | admittedly on 9.10 not current code | 15:44.25 |
| ray_laptop | chrisl: most customers are daunted by writing devices | 15:44.28 |
| kens | is not surprised, I am too | 15:44.41 |
| ray_laptop | kens: what did you try ? | 15:44.46 |
| chrisl | ray_laptop: but you said it was really easy..... ;-) | 15:44.53 |
| kens | ray_laptop : gsprint | 15:44.53 |
| | It uses a pipe to write the data form GS and pick up via gsprint | 15:45.05 |
| | to render to a DIB | 15:45.12 |
| ray_laptop | chrisl: easy for me. Not easy for customers | 15:45.16 |
| henrys | I'd really like to be done with the entire business and just eat kibble I store in my pocket but I understand that doesn't work for humans. We get sick and die or worse. | 15:45.22 |
| chrisl | ray_laptop: possibly any indication we need to rejig the API a bit?? | 15:45.35 |
| kens | henrys blanaced diet is beer (real ale that is) and cheese | 15:45.45 |
| ray_laptop | kens: what ghostscript comman line to a pipe works for you ? | 15:45.48 |
| kens | apparently it has all the nutrients you need :-)_ | 15:45.55 |
| | ray I'd have to fire up gsprint | 15:46.10 |
| | I'd prefer not to do that..... | 15:46.16 |
| henrys | kens:yeah but we've got to keep the economy going | 15:46.17 |
| ray_laptop | kens: not gsprint. just plain gswin32 or gswin32c | 15:46.32 |
| kens | Possibly the cheese needs ot be on crackers | 15:46.38 |
| chrisl | ray_laptop: anyway, if you think a more general (i.e. non-inkcov specific) solution is called for, then I'll look at doing that. | 15:46.47 |
| kens | ray_laptop : gpsrint uses a pipe, that's all I'm saying | 15:46.49 |
| | -sOutptuFile=%handle%0x00000154 | 15:48.38 |
| ray_laptop | kens: I'm not saying that pipes don't work. I was trying to say that ghostscript's mswin_popen appears broken | 15:48.40 |
| kens | AH that's not what I understood. | 15:48.53 |
| ray_laptop | kens: this is what fails for me: bin/gswin32c -sDEVICE=bbox -o'|cat' examples/tiger.eps | 15:49.20 |
| henrys | paulgardiner: I was under the impression horizon, because of it's association with the bbc, was real science, as kens says it does seem more entertainment oriented. Shouldn't be taken as peer reviewed quality stuff. | 15:49.37 |
| kens | OK that's not the same as using a pipe :-) | 15:49.41 |
| | Horixon *sued* to be a decent science documentary, but not these days | 15:50.06 |
| | used to be.... | 15:50.48 |
| chrisl | ray_laptop: Actually, I'm confused: you say a customer previously required raster output via the API stdio handlers, but the device API doesn't actually support that - so did they customise the code to do it? | 15:50.53 |
| ray_laptop | kens: actually it fails with gswin32.exe It sends the bbox output to stdout regardless: -o'|cat > y' still writes to stdout and the file "y" is not created | 15:52.07 |
| paulgardiner | henrys: yeah, I think they've moved towards trying to make it more main-stream entertaining, but I think it's still honest, in that they try to make it clear that some of the things they report need proper studies. They often present things a "Dr SomebodyOrOther believes..." | 15:52.16 |
| Robin_Watts | Everything that's on the BBC is 100% factual and correct. I recommend QI. And Top Gear. | 15:54.31 |
| chrisl | "Everything that's on the BBC is 100% factual and correct" - well, except the News, of course...... | 15:55.58 |
| jhabjan | chrisl: just to clear something out.... I did not requested raster output via API stdio, I was just trying to catch ink coverage values via stdout as I expected I could redirect output to the stdout easily. | 16:00.15 |
| | the idea was to have this working 'in memory' without a need to write something to disk.. | 16:00.34 |
| chrisl | jhabjan: that's what my patch does, but there's some debate about the form it should take, and whether it's something that should be available to other device (ones that output raster rather than text) | 16:01.29 |
| jhabjan | also, not to cunfuse, I'm not the customer... I'm author of the Ghostscript.NET :-) | 16:01.34 |
| ray_laptop | jhabjan: right. That's why I used the 'bbox' output as another example. It is also often used to get info about a file without ever generating rasters | 16:01.36 |
| chrisl | ray_laptop: like I said, if you think a more general solution is warranted, then I'm happy to do that | 16:04.19 |
| mvrhel_laptop | Robin_Watts: you there? | 16:05.58 |
| Robin_Watts | mvrhel_laptop: I am. | 16:06.04 |
| ray_laptop | chrisl: A more general solution _would_ be nice, but I'm not sure how ATM. I guess if -o- was used, we'd have to create a pip with the reader calling the stdio callback or something. | 16:06.14 |
| Robin_Watts | mvrhel_laptop: just reducing that file. | 16:06.27 |
| ray_laptop | chrisl: you may have a better approach. | 16:06.35 |
| jhabjan | anyway.. i found a temporary solution for this just to suit some user need: https://ghostscriptnet.codeplex.com/discussions/528389 | 16:06.46 |
| mvrhel_laptop | Robin_Watts: cool thanks | 16:06.49 |
| | Robin_Watts: so shouldn't I be able to take the patch from http://git.ghostscript.com/?p=user/robin/ghostpdl.git;a=commitdiff;h=24672833660644f1e6ffb78c2ffd90ffb2d65674 and just apply it to my local check out here to test | 16:07.58 |
| | for some reason that does not work for me | 16:08.08 |
| ray_laptop | chrisl:and the API is supposed to work on linux as well, it's just not used as often as on winodze | 16:08.18 |
| Robin_Watts | mvrhel_laptop: You should be able to. | 16:08.27 |
| | Try: git fetch robin | 16:08.33 |
| | then: git cherry-pick lcms2_thread | 16:08.49 |
| mvrhel_laptop | ok | 16:10.26 |
| Robin_Watts | OK. I shrank the file from 120 Meg to 3k :) | 16:12.31 |
| | mvrhel_laptop: Alternatively, I can send you a patch you can git am ? | 16:13.17 |
| mvrhel_laptop | that would be fine | 16:13.36 |
| | Robin_Watts: nice on the 120M to 3K | 16:13.42 |
| ray_laptop | hmm... bbox is *not* a good example. It doesn't use fwrite and doesn't even call 'print_page' but has it's own output_page function. So 'txtwrite' or 'inkcov' is a better example for what fails: | 16:15.23 |
| | e.g.: gswin32 -sDEVICE=txtwrite -sOutputFile=- -sstdout=x examples/colorcir.ps gives: /unknownerror in showpage | 16:15.25 |
| | and this does nothing: gswin32c -sDEVICE=txtwrite -sOutputFile='!cat' -sstdout=x examples/colorcir.ps | 16:17.09 |
| Robin_Watts | mvrhel_laptop: http://ghostscript.com/~robin/0001-LCMS.patch | 16:17.33 |
| mvrhel_laptop | thanks Robin_Watts | 16:17.56 |
| ray_laptop | oops. typo: This works: gswin32c -sDEVICE=txtwrite -sOutputFile='|cat' -sstdout=x examples/colorcir.ps | 16:18.19 |
| mvrhel_laptop | I have to run car to shop. will bbiaw | 16:18.25 |
| ray_laptop | chrisl: BTW, I have no objection to your posting that commit until we can come up with a more general solution | 16:44.14 |
| chrisl | ray_laptop: okay, thanks | 16:44.29 |
| ray_laptop | since a general solution will probably take a while. | 16:45.30 |
| | chrisl: do you want to open an enhancement bug for that, or should I ? | 16:47.03 |
| chrisl | ray_laptop: I'll do it later | 16:47.22 |
| ray_laptop | chrisl: OK. Thanks. | 16:47.29 |
| marcosw | chrisl: last week you mentioned a change in the ufst version we were using; can you remind me what you said? | 16:47.57 |
| chrisl | marcosw: I changed the directory name in svn-private from ufst-6.2 to ufst-6.x | 16:48.30 |
| | It now has 6.3 in it | 16:48.41 |
| henrys | marcosw: did miles get with you about ats tests I told him to follow up with you or maybe I've lost track of what we're doing. | 16:48.52 |
| | ? | 16:48.53 |
| jhabjan | is there any device that can output the result like tool_bin/pdf_info.ps does? | 16:50.26 |
| marcosw | chrisl: thanks. | 16:50.45 |
| chrisl | np | 16:50.50 |
| marcosw | henrys: my fault for not following up with you. I discussed it with miles but there was a question of which tests to get. they have a release for every version of office going back 15 years and seperate tests for every office component. | 16:52.14 |
| | I was thinking office 2010 and 2013. | 16:52.31 |
| | and word and powerpoint (and maybe excel). | 16:52.51 |
| henrys | that seems a reasonable start to me paulgardiner Robin_Watts did you want to weigh in on the purchase decision. | 16:53.29 |
| | ? | 16:53.29 |
| Robin_Watts | henrys: Miles needs to choose what versions we should be supporting. | 16:54.03 |
| henrys | Robin_Watts: okay go ask miles and then my phone will ring⦠;-) | 16:54.41 |
| Robin_Watts | In an ideal world, we'd know what versions the code got 'right' then we could pick those versions. | 16:55.00 |
| marcosw | each product/release is ~1200, so office 2010/2013 for word/powerpoint would be $4800. | 16:55.34 |
| paulgardiner | Perhaps it would be useful to compare with several versions. | 16:55.51 |
| marcosw | actually that's not true for office 2013, they don't list a price and appear to only sell them in a suite (luckily excel/powerpoint/word is one suite): http://storefront.qualitylogic.com/c-69-windows-8.aspx | 16:56.28 |
| Robin_Watts | I suspect that before we purchase test suites, we'd be better off purchasing old versions of office. | 16:56.45 |
| marcosw | Robin_Watts: I'm sure previous versions of office can be had for <$100 on eBay. | 16:57.31 |
| Robin_Watts | That someone can install all the old versions of office (possibly 1 per VM or something), and we can make a simple test document that we can load into each version in turn and compare results. | 16:57.47 |
| marcosw | Robin_Watts: someone === marcosw ? | 16:58.02 |
| Robin_Watts | someone = not me :) | 16:58.11 |
| marcosw | puts finger to nose :-) | 16:58.22 |
| henrys | marcosw: maybe we are being too ambitious and should just stick with 2013 for now - everyone can subscribe to the cloud thing and we are all set. | 16:58.27 |
| Robin_Watts | too slow :) | 16:58.27 |
| | I would imagine that later versions are more important to get exactly right. | 16:58.51 |
| henrys | marcosw: that's my vote 2013 only is a good start we can always go back. | 16:59.54 |
| marcosw | I did download the top 50 hits for .docx and .doc from the google, but I'm not sure how to determine which versions of word generated each of them. | 16:59.58 |
| Robin_Watts | marcosw: You might want to have a footle in /home/testdocs/testdocs on peeved :) | 17:00.41 |
| henrys | marcosw: oh that's interesing | 17:00.45 |
| Robin_Watts | I think the picsel spider may know how to classify by version. | 17:01.15 |
| ray_laptop | jhabjan: toolbin/pdf_info.ps just writes to stdout. It doesn't need a 'device' | 17:02.36 |
| henrys | marcosw: but in terms of testing basic functionality coverage a 2013 suite works, then we can use wild files to test conversion, does that seems reasonable? | 17:02.49 |
| ray_laptop | jhabjan: and since it writes using PostScript, its output _does_ go through the stdout callback | 17:03.27 |
| jhabjan | ray_laptop: yep, but the thing is that I can't ship pdf_info.ps with my library... | 17:03.44 |
| ray_laptop | jhabjan: why not ? | 17:03.57 |
| jhabjan | ray_laptop: Cannot mix MPL and GPL code | 17:04.27 |
| marcosw | henrys: I'm fine with starting with just 2013, I'll tell miles to order the test files. | 17:04.51 |
| ray_laptop | jhabjan: if you are calling the API functions, you are already GPL infected | 17:05.02 |
| jhabjan | ray_laptop: actually... I'm not | 17:05.16 |
| marcosw | Robin_Watts: any idea what the ole directory is? It appears to be the only one that has .doc files in it (and they are in v6 and v8 subdirectories, which I don't understand either). | 17:05.28 |
| ray_laptop | jhabjan: you must be if you are setting a stdout callback | 17:05.42 |
| Robin_Watts | ole = object linking and embedding. | 17:05.47 |
| jhabjan | ray_laptop: I spoke with Miles, he checked with his lawyer.... I'm fine | 17:05.57 |
| henrys | marcosw: and I guess we need to figure out bug reporting. I guess this should be privately tracked for now, is there some way in bugzilla to make an entire component private? | 17:06.07 |
| Robin_Watts | I believe that pre-XML word/excel/ppt files were all OLE files. | 17:06.12 |
| | v6 = from v6 of word. | 17:06.40 |
| ray_laptop | jhabjan: I'm not sure I understand how. | 17:06.42 |
| Robin_Watts | or office, rather. | 17:06.47 |
| henrys | wonders what on earth we've gotten ourselves into... | 17:07.19 |
| marcosw | henrys: I'm not sure if bugzilla has that ability, I'll check. If not we could would just run a seperate bugzilla behind a login. | 17:07.46 |
| ray_laptop | jhabjan: So, your Ghostscript.NET is MPL, but it calls the API functions ??? | 17:07.57 |
| Robin_Watts | henrys: It's a huge cess pit. But we've stolen someone elses raft. :) | 17:07.58 |
| jhabjan | ray_laptop: In my case, as long i don't ship any part of Ghostscript with Ghostscript.NET library, I'm fine | 17:08.04 |
| | ray_laptop: correct | 17:08.22 |
| ray_laptop | jhabjan: well, I guess it satisfies the legal terms of 'distribution', and it _does_ allow someone to substitute a new gsdll64 if they want to download a new one. | 17:10.17 |
| jhabjan | ray_laptop: Actually, two projects which I'm running at the moment are MPL-2... and as long as I'm not shipping Ghostscript library with my projects I'm fine with MPL-2. Projects are not statically linked to the Ghostsctip library ( which GPL disallows ), they call exposed Ghostscript API functions which is a same as calling .exe directly. In this | 17:11.01 |
| | particular case, there are different thoughts if this is also considered as derivative work (i figured out that U.S. court is not clear with this also). | 17:11.01 |
| henrys | marcosw: okay we aren't stuck with bugzilla either if you think it will be slow. I wonder if a text file in a private repo is faster if there are many reports, whatever you think | 17:11.09 |
| | marcosw: I know bugzilla has nicer record keeping but it is so clunky sometimes | 17:11.46 |
| marcosw | henrys: okay. | 17:14.56 |
| ray_laptop | jhabjan: right. So, the thing is toolbin/pdf_info.ps is a program that is AGPL, and is in source form. So it _might_ be possible for you to distribute that program, as a separate download, then have your users download that, as they do with the gsdll64.dll. | 17:16.18 |
| jhabjan | ray_laptop: yep, that's one solution | 17:17.15 |
| marcosw | is still finding features in bugzilla that he didn't know about, i.e. the patch viewer: http://bugs.ghostscript.com/attachment.cgi?id=10372&action=diff | 17:17.27 |
| ray_laptop | jhabjan: Or you could write your own pdf_info.ps to get the info you actually want in the format you want. The PS operators and procedures used are part of the Ghostscript PDF interpreter and are no different to the API functions, as far as I can see. Your app can send the PS using run_string | 17:19.24 |
| jhabjan | ray_laptop: yes, I think I will go with that | 17:20.35 |
| ray_laptop | your PS script to get the info from the PDF can be MPL, AFAICT | 17:21.04 |
| | jhabjan: obviously, you want to avoid copyright infringement, but there are MANY ways to code PS differently and avoid that. | 17:21.57 |
| mvrhel_laptop | Robin_Watts: ok that patch applied cleanly thank you. Don't know why the one that I extracted from your repos myself did not work. | 17:22.00 |
| marcosw | henrys: it does appear that bugzilla supports private bugs on a product basis. I'm not sure how recent that feature is, so I may have to update our bugzilla install (we are currently one revision behind). | 17:22.12 |
| Robin_Watts | mvrhel_laptop: the git fairy hadn't blessed it :( | 17:22.23 |
| ray_laptop | pdf_info.ps was intended to be an example anyway for people wanting to 'crack' PDF's to extract information | 17:22.32 |
| mvrhel_laptop | apparently. | 17:22.33 |
| ray_laptop | I have to run an errand. BBIAB | 17:22.50 |
| jhabjan | ray_laptop: yep, i was thinking a same thing | 17:23.08 |
| Robin_Watts | marcosw: I have an exe here for you. | 17:37.49 |
| | I'm uploading it to casper now. | 17:38.02 |
| | Have you got a windows box (or a WINE thing) to hand to give it a quick test? | 17:38.25 |
| marcosw | Robin_Watts: yes, I have a win7 virtual machine. | 17:38.39 |
| Robin_Watts | (hmm, a WINE box. I like that) | 17:38.42 |
| | ok. I *think* I've got it self contained now. | 17:38.58 |
| | 5 mins left on the upload. | 17:39.21 |
| | but essentially, you want to make an output directory, and you'll run: | 17:39.51 |
| | windows-da-test.exe -save <savedir> -hq in.doc out.txt msword rom:libraries/image/resource rom:libraries/font/library | 17:41.06 |
| | and that should give you a series of .bmp files. | 17:41.22 |
| | Likewise: | 17:41.31 |
| | windows-da-test.exe -pdfexport out.pdf in.doc out.txt msword rom:libraries/image/resource rom:libraries/font/library | 17:41.54 |
| | should give you an "out.pdf" | 17:42.07 |
| marcosw | Robin_Watts: there was some discussion of using it for pdf -> bmp as well. I'm not sure I understood the need for that ability but does that work? | 17:44.09 |
| Robin_Watts | marcosw: Yes. | 17:44.22 |
| | you use "in.pdf" instead of "in.doc" and "pdf" instead of "msword" | 17:44.47 |
| | The full list of input types are: html, msword, pdf, ppt,wmf, txt, csv, xml, excel, image, mhtml, drawfile, hangul, ooxml, decomail | 17:45.36 |
| marcosw | doesn't even now what some of those are :-) | 17:46.08 |
| | is ooxml Object Oriented XML? Does the world need such a thing? | 17:46.41 |
| | no, phew. it's open office xml. | 17:46.59 |
| Robin_Watts | marcosw: me either :( | 17:47.32 |
| | I have a nicer standalone tool now to do pdf export where it will figure out the types for itself, but that doesn't do bitmap output. | 17:48.06 |
| marcosw | henrys: okay, I think I've done it. There is now a ghostdocs product on bugs.ghostscript.com that should only be visible if you are a member of the ghostdocs group (which currently only artifex staff are). GhostDocs bugs are not cc'd to gs-bugs@ghostscript.com since I believe that is world-readable. | 17:49.14 |
| | Robin_Watts: NP, I don't mind specifying the input file type. | 17:51.19 |
| Robin_Watts | OK, it's uploaded. windows-da-test.exe in my casper home dir | 17:52.56 |
| marcosw | the GhostDocs bugs won't show up on the weekly bug aging and bug status emails either, but I'll fix that once we get some actual bugs entered (probably the bug aging email isn't urgent, since that only shows customer bugs). | 17:53.25 |
| Robin_Watts | marcosw: I'm trying for a linux build of the converter now. | 17:54.15 |
| marcosw | Robin_Watts: it's not a big deal, since I have to run Windows to convert from .doc(x) -> pdf via Word I might as well have the GhostDocs tool in Windows as well (btw, are we calling it GhostDocs?). | 17:55.43 |
| Robin_Watts | marcosw: I believe GhostDocs is the term that Miles is selling this under, yes. | 17:56.09 |
| | Miles seems very keen on the current hacked together workflow we have (if it's not PDF, convert it to PDF, then load it into MuPDF). | 17:57.00 |
| henrys | Robin_Watts: me too, I did kind of sell it to him. | 18:11.44 |
| Robin_Watts | henry: Right, but it has meant our previous "hacked together, good enough for a quick and dirty proof of concept" converter has needed to be refined. | 18:12.25 |
| henrys | Robin_Watts: I am trying to talk to Miles now, we need to slow down here and go into data collection mode. I hope he listens | 18:13.20 |
| kens | goodnight all | 18:14.13 |
| henrys | I did ask on IRC either you didn't see it or dismissed it as "idiotic" (entirely possible) the idea of replacing the edr layer with a pdf layer. We could at least preserve layout with tags, did you see that Robin_Watts ? | 18:15.06 |
| Robin_Watts | I did see that. | 18:15.27 |
| | and I mentioned it to paulgardiner | 18:15.40 |
| | and we both updated our CVs :) | 18:15.49 |
| henrys | Robin_Watts: hah | 18:15.58 |
| Robin_Watts | Whenever you edit a character in a file, do you want to have to rewrite an entire PDF? | 18:16.39 |
| henrys | Robin_Watts: the way you described editing in the app it was only capable of very small edits anyway | 18:17.26 |
| Robin_Watts | I may have been being unduly harsh. | 18:17.47 |
| | I wasn't involved in the editing push, and I left to go and work for some other company before that was shipped. | 18:18.21 |
| | The UI of smart office 2 is vastly better than that of smart office 1. | 18:19.08 |
| | probably because of the need to allow for such editing. | 18:19.44 |
| | It does look like you can create docs from scratch now. | 18:19.55 |
| | I suspect there are still limitations, like I wouldn't want to bet that you can edit tab positions etc, but.... | 18:20.15 |
| | but people should buy smart office and try it for themselves. I may have mentioned this before :) | 18:21.41 |
| | At the moment the Smart Office code loads everything into a single uber format, Edr. This is then laid out to give displaylists. | 18:22.46 |
| | Talking to various people over the past few days, I've had some feedback that ties in with what paulgardiner said the other day. | 18:23.11 |
| | Using a single format and then having a generic layout stage that works from that makes it very hard to cope with all the different idiosyncrasies of the different apps. | 18:24.12 |
| | it makes the layout stage much more complex. | 18:24.18 |
| | It has been suggested by various people (including paulgardiner) that we'd be better off having an 'interface' for loading, and an 'interface' for doing layout. | 18:24.55 |
| | That way each layout engine can be much simpler and we don't get the huge cross product of bugs. | 18:25.28 |
| mvrhel_laptop | Robin_Watts: I will get smart office this weekend and play around with it so I can see what it can do | 18:35.48 |
| Robin_Watts | On Android there is a Smart Office Lite version that's free. | 18:36.06 |
| | Or it's only 10 bucks or so. | 18:36.13 |
| mvrhel_laptop | I think miles can afford the $10 version | 18:36.26 |
| Robin_Watts | I worry when we rely on my crap memory about what features it has etc. | 18:36.53 |
| | Any firsthand information I have is at least 3 years old now. | 18:37.46 |
| henrys | Robin_Watts: I'll but it too. | 18:38.22 |
| mvrhel_laptop | np. I understand your frustration. I am more worried about letting customers see something that is less then steller. With the current thing that we have (and I understand that the issues are likely in the office to pdf conversion) there are a lot of issues just in the files that I have fooled with | 18:38.27 |
| | I worry that Miles is going to want to get it to a customer to check out before it is ready | 18:38.49 |
| Robin_Watts | mvrhel_laptop: yeah. paulgardiner pointed out that underlining doesn't work properly. | 18:39.04 |
| mvrhel_laptop | I did talk with him about this yesterday | 18:39.04 |
| henrys | mvrhel_laptop: I am trying to slow miles down | 18:39.09 |
| | mvrhel_laptop: the song is: we are vetting this technology here is a demo we made with it, we have to stick with something like that for now | 18:39.42 |
| mvrhel_laptop | yes | 18:39.54 |
| henrys | and it probably isn't conversion last word from Robin_Watts | 18:40.22 |
| mvrhel_laptop | at some point, it would be intersesting to see how hard it is to fix a bug in this code | 18:40.28 |
| Robin_Watts | absolutely. The technology gives us a good leg up into a new space, but it's not a fully formed solution. | 18:41.01 |
| | mvrhel_laptop: "interesting", yes. | 18:41.10 |
| mvrhel_laptop | considering I can't even begin to get a build in windoze | 18:41.28 |
| Robin_Watts | One thing... I'd forgotten just how horrible it is not to be able to right click on things in MSVC and have it jump me to the right file :) | 18:42.06 |
| mvrhel_laptop | when I first stepped into gs, I was up and running in a reasonable amount o time | 18:42.09 |
| Robin_Watts | mvrhel_laptop: Getting your first build out is a steep learning curve, yes :( | 18:42.28 |
| henrys | mvrhel_laptop: yes but just think what happened after than in gs | 18:42.58 |
| Robin_Watts | Did you try following the instructions I sent out? | 18:42.58 |
| | If you get stuck while following those, please let me know, and I can try to refine them. | 18:43.18 |
| | I may have missed a step. | 18:43.30 |
| mvrhel_laptop | Robin_Watts: ok. I forget where I got hung up. I got frustrated and moved on | 18:43.32 |
| | I will try again this weekend and see | 18:43.38 |
| Robin_Watts | Ok. It would be good for me to get those instructions into shape as others will have the same problems. | 18:44.15 |
| mvrhel_laptop | ok. I will make it a priority to do | 18:44.28 |
| henrys | mvrhel_laptop: it would be good to find some things that will work in Japan, is that possible or has everything you've looked at been bad? | 18:46.45 |
| mvrhel_laptop | henrys: this simple excel sheet was ok | 18:47.07 |
| | the word doc that I grabbed which was a MS driver document had all sorts of issues, as did my power point that I gave at the Linux meeting | 18:47.36 |
| | henrys: I agree, I need to find something that looks good. | 18:47.51 |
| Robin_Watts | mvrhel_laptop: What sort of issues? | 18:47.56 |
| mvrhel_laptop | and has figures, images and text | 18:47.59 |
| | Robin_Watts: I will have to go back and look. I can try to isolate to single pages and then provide them to you (or someone) | 18:48.38 |
| Robin_Watts | It would be really good to know if that's pdf export or document loading. | 18:48.44 |
| mvrhel_laptop | Robin_Watts: also, the one document (which was a simple one) crashed | 18:49.03 |
| | so out of the 5 docs that I have tried only one looked ok and it was a 1 page excel doc | 18:49.27 |
| | To me, it looks like we have quite a bit of work to do. But, we are way way way beyond where we would be if we did not have this as a starting point | 18:50.53 |
| | I will open the same files with the SmartOffice app and see what it makes of them too | 18:51.32 |
| | Robin_Watts: ^^ | 18:51.34 |
| Robin_Watts | cool. | 18:51.43 |
| mvrhel_laptop | that should answer some questions for us (or for you) | 18:51.54 |
| henrys | mvrhel_laptop: I can believe that folks using this on mobile would be okay with a lot of problems, but printing which will be our target audience is much less forgiving. | 18:52.20 |
| Robin_Watts | yes. | 18:52.37 |
| | My understanding was that code did a better job of matching than this :( | 18:53.02 |
| | I think things that are "missing" are likely to be the pdf export. | 18:53.34 |
| | Things that are badly positioned etc is going to be loading :( | 18:53.48 |
| henrys | Everyting crumbled after Robin_Watts and paulgardiner left | 18:53.57 |
| | ;-) | 18:53.59 |
| mvrhel_laptop | :) | 18:54.22 |
| | they thought they could run away from this code, but it has followed them like a stray cat | 18:54.46 |
| Robin_Watts | I worked on the core rendering and the PDF agent. I never touched the office agents. | 18:54.49 |
| henrys | Robin_Watts: that may change | 18:55.05 |
| mvrhel_laptop | hehe | 18:55.09 |
| Robin_Watts | yeah... | 18:55.12 |
| mvrhel_laptop | Robin_Watts: we are all in there with you dont worry! | 18:55.26 |
| Robin_Watts | If this does become a huge priority, there are at least others we can call on to help. | 18:55.51 |
| mvrhel_laptop | When I get this windoze stuff wrapped up, I will jump in. | 18:55.52 |
| Robin_Watts | so when Miles promises a customer a release in a fortnight... | 18:56.16 |
| mvrhel_laptop | I will still be working on the windows stuff | 18:56.30 |
| | ;) | 18:56.38 |
| Robin_Watts | :) | 18:57.03 |
| | I have a linux build working. | 19:08.14 |
| henrys | Robin_Watts: woohoo - I'll take that, I won't have to set up windows hell | 19:08.52 |
| Robin_Watts | Note, that's linux, not necessarily macos. I haven't tested it there. | 19:09.34 |
| henrys | no I want the linux stuff. | 19:09.49 |
| Robin_Watts | henrys: Well, you can sanity check me if you want... | 19:10.53 |
| | If you do: git clone henrys@ghostscript.com:/home/robin/sauce/epage.git epage.git | 19:11.36 |
| | Then: cd epage.git | 19:11.43 |
| | Then: scripts/linuxbuild.py da/pdf-exporter -vs2005 -nonet -romfss2 -define=ALL_DAS -define=PDF_EXPORT -debug | 19:12.10 |
| | Hopefully you'll get a build out. | 19:12.20 |
| | In all probability you'll need to install some 32bit build packages. | 19:12.32 |
| henrys | cloning now | 19:13.50 |
| | wow that's big | 19:17.04 |
| | actually the transfer speed git is reporting is only 2 Mb/s I don't know what's up with that. | 19:18.03 |
| Robin_Watts | henrys: It's a gig. | 19:19.31 |
| henrys | Robin_Watts: parallel compilation will eventually be desirable ;-) | 19:35.13 |
| Robin_Watts | henrys: Someone else can add that to the fbs :( | 19:35.38 |
| henrys | but it is buiding... | 19:36.14 |
| | Robin_Watts: I have a build and too few arguments, do you have a sample command line? | 19:44.24 |
| Robin_Watts | I do. | 19:44.31 |
| | da-pdf-exporter /home/robin/sauce/epage.git/agents/msword/test/l208.doc l208.txt /home/robin/sauce/epage.git/l208.pdf | 19:45.27 |
| | Currently you need to give full paths for the input and output docs. | 19:45.45 |
| henrys | ah I am using ~ notation perhaps that is the issue | 19:46.21 |
| Robin_Watts | paths are turned into file:// urls internally. | 19:46.40 |
| | I don't know whether ~ will play nice. | 19:46.50 |
| henrys | what's the intermediate txt file about? | 19:49.34 |
| Robin_Watts | henrys: that used to get various things like a layout dump | 19:50.24 |
| henrys | Robin_Watts: it's working !!! | 19:50.40 |
| Robin_Watts | I can probably remove that intermediate one now. | 19:50.50 |
| | henrys: woo hoo! | 19:50.53 |
| henrys | Robin_Watts: I ran it out of some long winded /tmp directory I don't know if I was supposed to do that but the pdf looks okay | 19:55.06 |
| Robin_Watts | yeah, thats expected. | 19:58.44 |
| henrys | Robin_Watts: now I can step through this in the debugger and see what you've gotten us into ;-) part of the reason I'm not so satisfied with just getting the app ... | 19:59.33 |
| Robin_Watts | ok, was on phone, sorry. | 20:00.02 |
| | The build puts everything into a 'genroot' directory. | 20:00.30 |
| | and all the options etc for the build go into the path within that genroot. | 20:00.41 |
| | Thus you can do lots of different builds with different options and not have to worry about them colliding. | 20:00.59 |
| henrys | Robin_Watts: okay makes sense. | 20:01.26 |
| Robin_Watts | you can probably set an environment var to say where you want it to be, but /tmp works for me on linux. | 20:02.04 |
| | tor8: ping | 20:02.12 |
| | Might be late for tor8. I'll burble here for the logs... | 20:02.27 |
| | I've just had a complaint in about the SVG text looking a bit crap when zoomed in, and looking at it now I can see that it's true. | 20:03.00 |
| | The problem is that if we ask for the outline of text from freetype and give it an identity matrix, it comes back quantised. | 20:03.42 |
| henrys | mvrhel_laptop: the linux build just worked didn't have to do anything - at least for pdf export. | 20:03.47 |
| Robin_Watts | If I give it a 1000,0,0,1000,0,0 matrix and the divide the path down on return I get much nicer results. | 20:04.27 |
| | except the edges of the glyph are slightly cut off. | 20:04.45 |
| | Again, I think this is because the initial call to fz_bound_glyph has a small matrix in it, which means we get quantised results back. | 20:05.20 |
| mvrhel_laptop | henrys: fair enough. I will go ahead and build under linux too. it would be nice (for me) to have a windows version building and running to though | 20:10.50 |
| | lunch timr | 20:12.31 |
| | time | 20:12.33 |
| henrys | mvrhel_laptop: me too | 20:12.56 |
| Robin_Watts | mvrhel_laptop: And it would be good for me to have someone try the instructions out. | 20:13.06 |
| ray_laptop | running regression on the fix for pxlmono. I had the wrong color_info (all RGB) for the Gray and CMYK pdf14 accumulator device prototype initializers. | 20:30.12 |
| mvrhel_laptop | Robin_Watts: well I will push forward with it all | 20:30.57 |
| ray_laptop | bug 694994 should be fixed by the commit I pushed. Anyone care to review the patch at: http://git.ghostscript.com/?p=user/ray/ghostpdl.git;a=commitdiff;h=34bd584f338f1b17607f45d2119b16365c61cc8c | 20:33.52 |
| | mvrhel_laptop: since it is color related, you are the logical choice (and it's a nice easy patch) :-) | 20:34.38 |
| Robin_Watts | ray_laptop: It looks plausible to me. | 20:35.02 |
| mvrhel_laptop | me too | 20:35.14 |
| Robin_Watts | Looks like they were all setup for rgb before? | 20:35.17 |
| ray_laptop | Robin_Watts: mvrhel_laptop: thanks. | 20:35.27 |
| Robin_Watts | mvrhel_laptop: I should be here this weekend, so I'll be around if you have problems. | 20:36.04 |
| ray_laptop | Robin_Watts: right. I did a cut and paste of the RGB and forgot to edit the color info parts. Somehow having different names didn't just work :-) | 20:36.10 |
| mvrhel_laptop | ok thanks | 20:36.11 |
| Robin_Watts | ah! I thought it was a long standing thing. | 20:42.22 |
| | henrys: I quite like the idea of trying to get hold of old versions of office for comparison. | 20:42.52 |
| | Do you reckon that's worth pursuing? | 20:43.23 |
| | Looks like there is 95, 97, 2000, 2003, 2007, 2010, 2013 | 20:50.46 |
| | 95 is v7, 97 v8, etc. Oh, there is a 2002 too. | 20:51.25 |
| | We can run everything from 95 to 2010 on an XP VM, so we could set ourselves up a VM for testing. | 20:53.18 |
| marcosw | Robin_Watts: the windows executable you uploaded to casper gives an error message when I try to run it on win7: The program can't start because MSVCR80D.dll is missing from your computer. | 21:29.26 |
| | apparently this is because the program was compiled with debugging enabled and I don't have the same version of Visual Studio you do. | 21:33.55 |
| henrys | Robin_Watts: no problem buying the software, it seems like a lot of work. marcosw would probably not like this plan much ;-) | 22:28.37 |
| mvrhel_laptop | henrys: making some nice progress on gsview. I have the pages all displaying nicely with a nice option of thumbnails displayed on the left from which you can click to get to a particular page. going to add in zooming and some other page display options (e.g. 2 up etc). then I will see about hooking in gs for distilling. | 22:48.39 |
| | need to head out for a bit | 22:52.18 |
| tor8 | Robin_Watts: yes, we've run into that before. see the comments in do_ft_render_glyph (should be 65536, should be 64) | 23:17.24 |
| | freetype truncates the coordinates between doing the transform and char-size scaling | 23:17.42 |
| Robin_Watts | tor8: So, do we have any better scheme in mind for fixing it? | 23:55.56 |
| | marcosw: It was built with VS2005 in debug mode. | 23:56.32 |
| | I can do a release build if you want. | 23:56.37 |
| | Forward 1 day (to 2014/02/01)>>> | |
IRC Logs