Yes, libtool is a bitch, that's why I wrote a patch (it's in the mapnik trac) to not use it, and use ::LoadLibrary() directly (on Windows). Mixing MingGW and VS binaries won't work. It's a simple patch, have a look (search for my username roel_v on the trac) and let me know if you need further help getting it to work.
Regarding the boost version, apparently it's because mapnik uses a deprecated API of a specific library (I forgot the details), but it's header-only so as long as you list the boost header directory that comes with mapnik before any newer boost versions, you'll be good. cheers, roel On Wed, Apr 13, 2011 at 10:16 PM, William Johnston <[email protected]> wrote: > Well, I said I would report back. Unfortunately, I wasn't successful. > > I got all the prerequisites and compiled them. > > These compile with little or no trouble in VS: > http://site.icu-project.org/download/46#ICU4C-Download > http://download.savannah.gnu.org/releases/freetype/ > http://trac.osgeo.org/proj/ > ftp://ftp.remotesensing.org/pub/libtiff > http://www.libpng.org/pub/png/libpng.html > http://zlib.net/ > http://www.ijg.org/ > ftp://xmlsoft.org/libxml2/ > > But this one is a show stopper: > http://www.gnu.org/software/libtool/ > > What I ended up doing for libtool was compiling it in MinGW. And converting > the .a file to a .lib. It's supposed to work with Microsofts Visual C++, but > this is one place where I feel things may have gone wrong. > > I made some edits to the Mapnik code to get it to work with the latest Boost > files. And this is the second place where I may have messed things up. I > didn't use the Boost header files that came with the Mapnik source at all. > I'm kind of confused why Boost header files are distributed with Mapnik, but > there is probably a good reason. So that's another place things may have > gone wrong. > > Anyway, I got the whole thing to compile, link, and run without crashing... > but it didn't work. > It throws and exception at line 190: > lyr.set_datasource(datasource_cache::instance()->create(p)); > saying that is doesn't have a plugin for type "shape." Which probably means > it failed to load the shape.input file earlier. Which could be a problem > with libtool, or the shape.input file, or even a messed up Boost. > > Anyway, I don't need this to run in Visual Studios. All I need are some > tiles to test something. And to that end this morning I learned enough > Python and Mapnik to generate a good enough tile set for my purposes. I > should have done that a week ago, but I'm stubborn. And since I'm stubborn > I'll probably continue to tinker with it for a while, but I'm doubtful that > I'll make much progress. > > > On 4/6/2011 10:47 AM, William Johnston wrote: >> >> LOL I didn't know it was fashionable to dislike VS. I just know that every >> time I use it it's like getting kicked in the head. >> >> Anyway, I'm pretty sure the Boost files are the ones I need. I'll >> recompile all the others and let you know how it goes. >> >> On 4/5/2011 2:48 PM, Roel Vanhout wrote: >>> >>> Are you sure these libraries were all compiled with VS10? I'm quite >>> sure they're not. For example the freetype page mentions the msvc 6 >>> runtime dll's. There is no guarantee in C++ that object files from >>> different compilers will work together, so mixing vs6 and vs8 object >>> files is almost sure to not work. >>> >>> So, compile everything from source, using my instructions on the wiki >>> (I posted the link yesterday or so - or maybe it was on -devel). It's >>> virtually impossible to get C++ programs to work from binary >>> libraries. Don't bother trying until you know about all the details >>> and how the compilers work. >>> >>> When you compile all these things from source, make sure to use the >>> same settings for each project: runtime libary, usage of SECURE_CRT, >>> what version of the crt to use, etc. Again, I don't recommend trying >>> unless you already know how to work with VS. >>> >>> Also, this has nothing to do with VS. It's just how C++ works. I know >>> it's fashionable to dislike VS, but the fact of the matter is that >>> it's universal that things only work when one knows how to work with >>> them. >>> >>> >>> cheers, >>> >>> roel >>> >>> >>> On Tue, Apr 5, 2011 at 7:18 PM, William Johnston<[email protected]> >>> wrote: >>>> >>>> Well, I didn't compile any of the libraries. I got precompiled versions >>>> for >>>> Windows: >>>> >>>> Mapnik 0.7.1: >>>> http://prdownload.berlios.de/mapnik/mapnik-0.7.1-win32-py25_26.zip >>>> Boost 1.46.1: http://www.boostpro.com/download/ >>>> ICU4C 4.6.1: http://site.icu-project.org/download >>>> GNU Libtool 1.5.26: http://gnuwin32.sourceforge.net/packages/libtool.htm >>>> Freetype 2.3.6: http://gnuwin32.sourceforge.net/packages/freetype.htm >>>> >>>> >>>> In Visual Studios I started with an empty project. I changed only the >>>> project settings having to do with finding the right libraries and >>>> include >>>> files. >>>> Here are some of the project settings that might be relevant: >>>> Character Set ------- Use Multi-Byte Character Set >>>> Runtime library ----- Multi-threaded DLL >>>> >>>> Changing the "Character set" to "Use Unicode Character Set" didn't >>>> change >>>> anything. >>>> Changing the "Runtime library" to "Multi-threaded Debug DLL" caused the >>>> program to crash at a later point: >>>> m.insert_style("provinces",provpoly_style); >>>> >>>> I fiddled with a bunch of other setting, but nothing worked. >>>> >>>> I would like to take this moment to say that I have always hated Visual >>>> Studios. >>>> >>>> >>>> >>>> On 4/5/2011 4:55 AM, Roel Vanhout wrote: >>>>> >>>>> Looks like an issue between ANSI and Unicode strings (in the MSVC >>>>> sense, i.e. single- vs multibyte strings). >>>>> >>>>> How did you compile things? Build everything (including boost) from >>>>> source? What VS project files do you use? >>>>> >>>>> >>>>> On Tue, Apr 5, 2011 at 2:51 AM, William Johnston<[email protected]> >>>>> wrote: >>>>>> >>>>>> I'm using VS 2010 and Mapnik 0.7.1 and compiling to 32 bits. >>>>>> >>>>>> When I run rundemo.exe I get this: >>>>>> >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\Release>a >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\ma >>>>>> pnik-0.7.1\lib\mapnik >>>>>> running demo ... >>>>>> looking for 'shape.input' plugin in... >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\mapnik >>>>>> -0.7.1\lib\mapnik/input/ >>>>>> looking for DejaVuSans font in... >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\mapnik-0.7. >>>>>> 1\lib\mapnik/fonts/DejaVuSans.ttf >>>>>> ### Configuration error: Failed to parse filter expression: >>>>>> rundemo Z:\SoftwareDevelopment >>>>>> Parsing aborted at 'r' >>>>>> >>>>>> It's trying to parse the command line. If I change rundemo.exe to >>>>>> a.exe >>>>>> and >>>>>> rerun it I get: >>>>>> >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\Release>a >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\ma >>>>>> pnik-0.7.1\lib\mapnik >>>>>> running demo ... >>>>>> looking for 'shape.input' plugin in... >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\mapnik >>>>>> -0.7.1\lib\mapnik/input/ >>>>>> looking for DejaVuSans font in... >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\mapnik-0.7. >>>>>> 1\lib\mapnik/fonts/DejaVuSans.ttf >>>>>> here### Configuration error: Failed to parse filter expression: >>>>>> a Z:\SoftwareDevelopment\c++\M >>>>>> Parsing aborted at 'a' >>>>>> >>>>>> The error is caused by this line: >>>>>> provpoly_rule_on.set_filter("[NAME_EN] = 'Ontario'"); >>>>>> >>>>>> My understanding is that it is supposed to be parsing "[NAME_EN] = >>>>>> 'Ontario'" rather than the run command. >>>>>> >>>>>> If I change the line to be these two lines: >>>>>> >>>>>> std::string test("[NAME_EN] = 'Ontario'"); >>>>>> provpoly_rule_on.set_filter(create_filter(test)); >>>>>> >>>>>> I get: >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\Release>Mapnik >>>>>> C:\mapnik-0.7.1\lib\mapnik >>>>>> running demo ... >>>>>> looking for 'shape.input' plugin in... >>>>>> Z:\SoftwareDevelopment\c++\Mapnik\mapnik >>>>>> -0.7.1\lib\mapnik/input/ >>>>>> looking for DejaVuSans font in... >>>>>> C:\mapnik-0.7.1\lib\mapnik/fonts/DejaVuSans.t >>>>>> tf >>>>>> ### Configuration error: Failed to parse filter expression: >>>>>> ،n >>>>>> >>>>>> Even weirder. >>>>>> >>>>>> The peer line in rundemo.py works fine. >>>>>> >>>>>> This error make no logical sense. My current thinking is that there is >>>>>> something wrong with mapnik.dll >>>>>> or mapnik.lib. >>>>>> >>>>>> Does anyone have any experience with this? >>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> Mapnik-users mailing list >>>>>> [email protected] >>>>>> https://lists.berlios.de/mailman/listinfo/mapnik-users >>>>>> > _______________________________________________ Mapnik-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/mapnik-users
