| <<<Back 1 day (to 2011/08/09) | 2011/08/10 |
mvrhel2 | hmm is the windows build broken? | 02:14.55 |
henrys | mvrhel2:my last change was rather pervasive but I was able to do the windows build locally. What's the error? | 02:26.37 |
mvrhel2 | henrys: I am cleaning up and rebuilding now | 02:31.17 |
| I was fixing my various compiler errors and had to do a full rebuild at some point and had some errors that came up about gserror.h | 02:31.49 |
| some linking issues | 02:31.57 |
| hold on a sec | 02:32.15 |
henrys | right I just removed that file. | 02:35.40 |
mvrhel2 | ok. It does not build after I reset to master and did a clean rebuild | 02:36.47 |
| two issues | 02:37.12 |
| one I can easily fix | 02:37.19 |
henrys | do you still have local changes for project files? | 02:37.26 |
mvrhel2 | one I don't understand what is going on | 02:37.28 |
| oh | 02:37.38 |
| but the header files should not be in the solution | 02:37.53 |
| oh. | 02:38.16 |
| ok. gserrors.h is in there | 02:38.28 |
| and no gserror.h | 02:38.45 |
| so this one | 02:39.38 |
| 1>gsicc_lcms.obj : error LNK2019: unresolved external symbol _gs_warn1 referenced in function _gscms_error | 02:39.44 |
| is easy to fix | 02:39.48 |
| I am not sure why this was not a problem before | 02:39.58 |
| this one | 02:40.07 |
| 1>gp_mshdl.obj : error LNK2019: unresolved external symbol _return_error referenced in function _mswin_handle_fopen | 02:40.08 |
| I don't know what is going on yet | 02:40.18 |
| oh | 02:40.34 |
| I see | 02:40.36 |
| nevermind | 02:40.46 |
| let me fix both of these | 02:40.51 |
henrys | but how did this compile for me? | 02:41.45 |
mvrhel2 | no idea. | 02:42.13 |
| it was working for me | 02:42.20 |
| until I did a clean build | 02:42.24 |
henrys | and it got through the cluster build. | 02:44.13 |
mvrhel2 | well, I don't know how the lcms one made it through | 02:45.05 |
| but I also dont see how it worked before | 02:45.30 |
| as I don't think that file had the include for the warning in the old code | 02:45.54 |
| building my fix now | 02:46.05 |
henrys | well I expect the lacking include to result in a warning. | 02:46.33 |
| gs_throw_imp is resolved to gsmisc.obj which is linked in. | 02:46.54 |
| you just added the include and it worked? | 02:47.25 |
mvrhel2 | I added the include and it is building.... | 02:47.41 |
| my machine is slow for some reason | 02:47.49 |
| it appears that it decided to do a full rebuild | 02:48.03 |
henrys | sorry to screw things up I'll check what went wrong with my windows build. | 02:49.52 |
mvrhel2 | no worries. I was worried my attempts to fix all these broken const issues in the icc code introduced this. I couldnt figure it out. | 02:51.07 |
| ok it builds now | 02:52.01 |
| That was just adding the includes | 02:52.11 |
| I don't think I need to clusterpush this to test | 02:52.27 |
| do you henrys? | 02:52.30 |
henrys | debug or release build? | 02:52.37 |
mvrhel2 | debug | 02:52.41 |
| let me build the release | 02:52.46 |
| another clean rebuild | 02:53.04 |
henrys | no clusterpush is necessary. | 02:53.07 |
| bbiab | 03:02.24 |
mvrhel2 | ok release built | 03:05.02 |
| henrys: ok I commited it. signing off now. tty tommorrow | 03:09.58 |
henrys | tor8:do you have a release date now for mupdf? | 16:48.10 |
| chrisl_away:something I never payed attention to but the gs release requires running autogen.sh - usually autoconf is used when checking the code out of the repo and just configure is needed with a tarball. But we need the autogen.sh to symbolic link Makefile.in - minor I suppose. | 17:40.23 |
chrisl | henrys: it should be correct in the archive, isn't it - let me double check | 17:43.00 |
henrys | and the luratech compile failed for me on the mac. | 17:43.47 |
chrisl | In what what did it fail? | 17:44.12 |
henrys | I am checking commercial ghostscript 9.04 | 17:44.22 |
| jp2_common.h:51 - can't find malloc.h | 17:44.52 |
chrisl | henrys: so, didn't anyone test Luratech on Mac before? | 17:46.52 |
henrys | I sort of thought it would be tested by doing a clusterpush. | 17:47.22 |
alexcher | henrys: Luratech run on the cluster before. | 17:47.46 |
henrys | it's okay on linux? | 17:47.48 |
chrisl | It worked on Linux for me, yes. I'll double check the commercial release archive | 17:48.22 |
henrys | if by chance the mac pro was down at the time. | 17:48.27 |
| that would obviously be a problem | 17:48.37 |
| but the mac pro's uptime is fairly good. | 17:49.54 |
| if it is just a mac problem I think we can safely just do a patch. | 17:51.10 |
chrisl | henrys: I'm slightly concerned that we were a bit quick in opting for Luratech always in the commercial release - it seems it have never been tested on Win64 nor on Mac - that's a bit worrying. | 17:52.49 |
henrys | Win64 is a hole - but Mac should have been tested with alexcher and you doing clusterpushes. Why didn't we see this? | 17:55.07 |
chrisl | I did change the build so that both luratech source dirs are under "luratech" and I only tested that locally, but I can't see how that would cause the problem on Mac | 17:58.06 |
henrys | the -I is wrong for jp2_adt_band_array.c it is different than other jp2 files do you see that on linux? | 17:59.24 |
chrisl | GS with luratech just built fine on Linux | 18:00.25 |
henrys | yes but if you look at the -I stuff is it correct for that file. | 18:00.54 |
chrisl | I don't see how the -I setting can be different, they are include in the LWF_JP2_CC used for all those files | 18:02.53 |
henrys | sending my compiler output | 18:05.26 |
| by email | 18:05.29 |
| I just noticed that one jp2 file has 2 includes and the one that fails has one. | 18:06.30 |
| but that is not the problem malloc.h needs special treatment. | 18:06.43 |
chrisl | So, on the Mac we should be #defining MAC | 18:07.31 |
| henrys: what does "uname" return, on the Mac? | 18:09.35 |
henrys | Darwin | 18:10.30 |
| still bugs me the clusterpush didn't fall over. | 18:11.00 |
chrisl | Hmm, that ought to work. It looks like the test in configure is failing to spot it's a Mac. I'll fire up my Mac Mini tomorrow, and sort it out. | 18:11.32 |
henrys | or I can make you a login if you like. | 18:11.59 |
chrisl | Okay, that would work better, my Mac Mini is a G4 model. I take it I need an IP address, too? | 18:13.08 |
henrys | yeah I'll send you an email when I have it set up it will use your keys from casper. | 18:14.09 |
chrisl | Great, thanks. | 18:14.21 |
henrys | the stuff in malloc_.h does work so I wonder if it is something we want to change in luratech. | 18:15.31 |
| probably best not to fork their code if we can avoid it. | 18:16.13 |
chrisl | We've already forked it | 18:16.27 |
henrys | XCFLAGS=-DMAC does fix the problem. | 18:17.00 |
chrisl | I guess the XCFLAGS aren't getting to Luratech, then? | 18:17.56 |
henrys | no if I specifiy XCFLAGS=-DMAC on the command line the code compiles, now I have another error sigh... | 18:18.35 |
chrisl | You probably also need -DMAC_OS_X_BUILD | 18:19.00 |
henrys | oh nvm let ma start again | 18:19.09 |
| I knew the opaque commercial release was going to be a problem, I see it pulling in all sorts of GPL code. | 18:20.54 |
chrisl | I know, but I didn't have time to sort it out. It is on my to-do list for the next release. | 18:23.56 |
henrys | I'm going to have some lunch than I'll send you the login info thanks chrisl. | 18:28.30 |
tkamppeter | I have taken libicc into the Ubuntu distro and now I want to make Ghostscript using it instead of its built-in icclib. How do I proceed? | 18:34.08 |
chrisl | tkamppeter: I don't think we use icclib anymore, that has been supplanted by lcms - AIUI | 18:36.40 |
tkamppeter | chrisl, so the icclib directory is unused, obsolete? | 18:37.11 |
chrisl | As far as I know, yes. It's still there only for reference. Best double check with henrys or mvrhel2 (when he's around) to be sure. | 18:40.06 |
tkamppeter | chrisl, I have grepped through the code now and the only file including icc.h is imdi/cctiff.c | 18:40.47 |
chrisl | tkamppeter: I did an "out of the box" build with the icclib directory removed, and it built just fine, so it does like that's the only use of it. | 18:43.35 |
tkamppeter | For what is the file imdi/cctiff.c good for? It is written by libiccc author Graeme Gill. | 18:44.25 |
henrys | imdi should be removed. | 18:45.58 |
| should be out by 9.05 | 18:46.27 |
chrisl | henrys: I thought we had to keep it for some customer uses? | 18:46.32 |
henrys | are you think of wts? | 18:47.00 |
tkamppeter | Then I know imdi is not used. Looked for me like that, too. No *.mak file bt a Jamfile. | 18:47.19 |
chrisl | henrys: could be, I confess I don't what imdi is or is used for | 18:48.08 |
| s/don't/don't know | 18:48.20 |
henrys | imdi is color conversion - and was replaced by lcms. | 18:48.40 |
chrisl | tkamppeter: well there's you answer - you can ignore imdi and icclib. | 18:49.55 |
tkamppeter | chrisl, thank you very much. | 18:52.31 |
henrys | not to sound like a broken record but this shared library business is simply too fragile, I do wish you'd just use the static libraries whenever possible. I see on the mac many shared libraries are in the app bundle. | 18:52.32 |
tkamppeter | Going on to next problem: https://bugs.launchpad.net/ubuntu/+source/ghostscript/+bug/659398 | 18:52.48 |
henrys | mvrhel2 plans to replace well tempered screening so I'd rather not fool with this if we can avoid it. | 18:54.06 |
tkamppeter | henrys, distro policies. We must use shared libraries as much as possible. So that if a bug is in a shared lib, we only need to replace the lib and not also a bunch of statically linked programs. | 18:54.22 |
| henrys, so best is to simply drop this option in foomatic-db? | 18:54.49 |
chrisl | tkamppeter: I'd also like to make the point that an error exit condition is *not* a crash - perhaps you could pass that on to the reporter....... | 18:55.49 |
henrys | I understand the philosophy and in theory it makes sense in practice I believe is it hasn't worked and is one of the huge problem the open source community faces competing with windows and the mac. I'll leave it with that. | 18:56.15 |
tkamppeter | chrisl, all in all, the job did not come out ehrn the WTS option was set to anything other than "Off". | 18:56.32 |
henrys | yes if we can drop the option that would be great. | 18:57.12 |
chrisl | tkamppeter: but it's not a crash (in GS, at least), not by any stretch of the imagination - crash is a seg fault, divising by zero, etc. | 18:57.42 |
tkamppeter | chrisl, I know, the bug title is from the OP not from me. | 18:59.56 |
henrys | any other bugs tkamppeter? | 19:00.13 |
tkamppeter | chrisl, title corrected. | 19:00.36 |
chrisl | tkamppeter: thanks - I wasn't having a rant at you, just at end users who cause confusion by claiming that anything that doesn't behave as they want is a "crash" ;-) | 19:01.54 |
tkamppeter | henrys, yes, will tell about it later. | 19:12.21 |
SteveP_ | How do I install Ghostscript? There's no exe file. | 19:34.43 |
chrisl | SteveP_: I assume from the question you are using Windows? | 19:35.40 |
SteveP_ | Yes. I use Irvanview and to open PDF files they say I need Ghostscript | 19:36.31 |
| I meant Irfanview | 19:37.33 |
chrisl | Okay, so go http://www.ghostscript.com/download/gsdnld.html and use the "Ghostscript GPL Release" for Windows (32/64 bit, whichever you use), and that should get you an installer, | 19:38.43 |
SteveP_ | thank you. | 19:39.10 |
chrisl | thinks Irfanview isn't really in the spirit of the GPL :-( | 19:39.53 |
| henrys: I'm going finish now.I've just tested a fix for configure problem on the cluster and it built and ran on the macpro, so I'll commit tomorrow morning. Should I replace the archives for the commercial release with ones with a fixed configure(.ac) in them? | 19:44.09 |
henrys | oh you found the login? | 19:45.02 |
| I was just about to send it to you. | 19:45.12 |
chrisl | henrys: no, but I can see from the cluster result that the luratech code was used, and that the Mac pro ran tests. | 19:45.49 |
henrys | we assumed that before ;-) | 19:46.14 |
| I had another failure in a contrib driver. | 19:46.36 |
| compile failure that is. | 19:46.43 |
| anyway henry-artifex.homeip.net should work for ya. | 19:47.05 |
chrisl | I know contrib is a problem, it's always been that way - again, on my list to sort out, but as it's not new, I didn't want to hold up the release | 19:47.35 |
henrys | but the system doesn't build. | 19:47.59 |
chrisl | do configure --disable-contrib | 19:48.40 |
henrys | okay | 19:48.56 |
| I can't believe we expect customers to know that ... | 19:49.19 |
mvrhel2 | alexcher: are you around? | 19:50.15 |
chrisl | Well, as I said, I don't like it, but it would have meant a few days delay to the release to sort it out, and (again) it isn't new | 19:50.30 |
mvrhel2 | so the Resource/Halftone/Default did not seem to work for me | 19:50.38 |
henrys | chrisl:right I understand I think a patch for the malloc stuff is fine, when you get a chance. | 19:51.11 |
mvrhel2 | alexcher: and the option of putting it on the command line is not really an option, since it is a large threshold array (unless there is a way to do that on the command line) | 19:51.16 |
chrisl | henrys: I'll do that tomorrow morning - like you, I want to test my fix "manually" on the macpro before I consider it fixed. | 19:56.09 |
henrys | okay | 19:56.32 |
chrisl | I take it the login and IP will still be valid in 11 or 12 hours? | 19:56.55 |
henrys | right | 19:57.07 |
chrisl | Good, thanks | 19:57.16 |
| okay - 'night all! | 19:57.43 |
alexcher | mvrhel2: gs has an include i.e. run operator. | 20:16.29 |
tkamppeter | Ghostscript in Ubuntu uses lcms, is there an advantage to switch to lcms2? | 20:16.48 |
alexcher | chrisl: I cannot push to thirdparty/openjpeg.git : fatal: The remote end hung up unexpectedly | 20:19.43 |
henrys | tkamppeter:it is on the todo list, AFAIK it is not backward compatible. | 20:23.27 |
chrisl_away | tkamppeter: lcms2 does not current work fully, it should not be used at the moment. | 20:24.22 |
| alexcher: that repos is really tor8's for mupdf, but I suspect it's a read only mirror - just guessing. I'm going to pull in the openjpeg source like the other libs in GS. | 20:25.46 |
| alexcher: while on that subject, for the autoconf openjpeg stuff: I take it we still want to default to Jasper, and have OpenJPEG as an option, for now? | 20:26.40 |
alexcher | That's fine for me. OpenJPEG is about 3 times faster and uses 3 times less memory but the library has bugs and omissions. | 20:28.01 |
chrisl_away | alexcher: okay, good. It's easy to switch the default/option when we're ready. | 20:29.22 |
| henrys: I've committed the luratech on Mac configure fix - thanks for the login. | 20:29.53 |
henrys | great. | 20:31.39 |
chrisl_away | really is finishing now :-) | 20:32.53 |
henrys | good night | 20:33.15 |
ljsdofuynsdfufuh | hey guys, I'm trying to get a pdf that is about 120k down to <100k | 22:30.13 |
| It is a resume, and the job's uploader function doesn't allow >100k resumes | 22:30.36 |
| everytime I try to do stuff to reduce the size, it gets about 20k larger | 22:30.56 |
alexcher | ljsdofuynsdfufuh: you can try to run your file through ghostscript as: | 22:51.25 |
ljsdofuynsdfufuh | as? | 22:51.42 |
alexcher | gswin32c -sDEVICE=pdfwrite -o OUT.PDF IN.PDF | 22:52.07 |
ljsdofuynsdfufuh | I'm on linux, if that matters | 22:52.26 |
ghostgum2 | If you have any images in the file, try adding -dPDFSETTINGS=/ebook after the -sDEVICE-pdfwrite. This will resample high-res images down to 150dpi. | 22:53.26 |
ljsdofuynsdfufuh | yeah alexcher that makes it 2k larger | 22:53.42 |
| no it is only font, and it was made (exported) by openoffice from an odt file | 22:54.06 |
| Also, it is only 3 pages | 22:54.23 |
alexcher | You can also take out some of your experience points. Perhaps PDP-11 or OS/360 are of litte interest now. | 22:54.45 |
| Seriously, post the file sonewhere. | 22:55.51 |
ljsdofuynsdfufuh | has a lot of personal info on it | 22:56.22 |
| it is a resume | 22:56.26 |
| Isn't there only a few things it could be? | 22:57.20 |
alexcher | The resume is created to be viewable for everybody, isn't it? | 22:57.25 |
ljsdofuynsdfufuh | sometimes | 22:57.44 |
tor8 | alexcher: the thirdparty git repos are readonly mirrors | 22:58.29 |
ljsdofuynsdfufuh | how can I see what extra stuff it might have? | 22:58.52 |
alexcher | ljsdofuynsdfufuh: I use emacs for this. | 22:59.45 |
| tor8: The stock OpenJPEG doesn't support everything that is needed for PDF. | 23:01.11 |
ljsdofuynsdfufuh | alexcher: it just shows the pdf | 23:02.16 |
tor8 | alexcher: right. so we need to maintain our own set of patches or a branch then. | 23:02.39 |
alexcher | ljsdofuynsdfufuh: Look at the objects that have large /Length attribute, or a large number of small objects. | 23:03.55 |
ljsdofuynsdfufuh | I'm not sure how... not an emacs guy | 23:05.26 |
| What's the name of the tool? | 23:05.38 |
alexcher | ljsdofuynsdfufuh: any text editor that can show a mix of text and binary will do. | 23:07.40 |
ljsdofuynsdfufuh | heh, I don't have any experience hacking pdfs | 23:13.43 |
| can I send it to you alexcher? | 23:13.53 |
alexcher | ljsdofuynsdfufuh: yes, alex.cherepanov@artifex.com | 23:16.52 |
| ljsdofuynsdfufuh: I'm a nice guy, not a recruiter or something like yhis. | 23:18.25 |
ljsdofuynsdfufuh | sent | 23:20.27 |
alexcher | ljsdofuynsdfufuh: Most of the space is consumed by fonts. Thy to use different fonts. if possible, use Type 1 fonts instead of TrueType. | 23:28.14 |
| ljsdofuynsdfufuh: Helvetica looks very close to Arial but takes no space at all. It's a build-n font. | 23:30.38 |
ljsdofuynsdfufuh | ok thanks | 23:40.37 |
| gotta either go find the original odt or convert it to one | 23:40.57 |
| Forward 1 day (to 2011/08/11)>>> | |