Building wxPython 2.5 for Development and Testing

This file describes how I build wxWindows and wxPython while doing development and testing, and is meant to help other people that want to do the same thing. I'll assume that you are using either a CVS snapshot from http://wxwindows.org/snapshots/, a checkout from CVS, or one of the released wxPythonSrc-2.5.* tarballs. I'll also assume that you know your way around your system, the compiler, etc. and that you know what you are doing! ;-)

If you want to also install the version of wxPython you build to be in your site-packages dir and be your default version of wxPython, then a few additional steps are needed, and you may want to use slightly different options. See INSTALL.txt for more details. If you only use the instructions in this BUILD.txt file then you will end up with a separate installation of wxPython and you can switch back and forth between this and the release version that you may already have installed.

If you want to make changes to any of the *.i files, (SWIG interface definition files,) or to regenerate the extension sources or renamer modules, then you will need an up to date version of SWIG. Either get and build the current CVS version, or version 1.3.20, and then apply the patches in wxPython/SWIG. See the README.txt in that dir for details about each patch and also info about those that may already have been applied to the SWIG sources. If you install this build of SWIG to a location that is not on the PATH (so it doesn't interfere with an existing SWIG install for example) then you can set a setup.py command-line variable named SWIG to be the full path name of the executable and the wxPython build will use it. See below for an example.

Building on Unix-like Systems (e.g. Linux and OS X)

These platforms are built almost the same way while in development so I'll combine the descriptions about their build process here. First we will build wxWindows and install it to an out of the way place, then do the same for wxPython.

  1. Create a build directory in the main wxWindows dir, and configure wxWindows. If you want to have multiple builds with different configure options, just use different subdirectories. I normally put the configure command in a script named ".configure" in each build dir so I can easily blow away everything in the build dir and rerun the script without having to remember the options I used before:

    mkdir bld
    cd bld
    ../configure --prefix=/opt/wx/2.5 \
                 --with-gtk \
                 --with-opengl \
                 --disable-monolithic \
                 --enable-debug \
                 --enable-geometry \
    

    On OS X of course you'll want to use --with-mac instead of --with-gtk. For GTK2 and unicode add:

    --enable-gtk2 --enable-unicode

    Notice that I used a prefix of /opt/wx/2.5. You can use whatever path you want, such as a path in your HOME dir or even one of the standard prefix paths such as /usr or /usr/local if you like, but using /opt this way lets me easily have multiple versions and ports of wxWindows "installed" and makes it easy to switch between them, without impacting any versions of wxWindows that may have been installed via an RPM or whatever. For the rest of the steps below be sure to also substitute "/opt/wx/2.5" with whatever prefix you choose for your build.

    If you want to use the image and zlib libraries included with wxWindows instead of those already installed on your system, (for example, to reduce dependencies on 3rd party libraries) then you can add these flags to the configure command:

    --with-libjpeg=builtin \
    --with-libpng=builtin \
    --with-libtiff=builtin \
    --with-zlib=builtin \
    
  2. To build and install wxWindows you could just use the "make" command but there are other libraries besides the main wxWindows libs that also need to be built so again I make a script to do it all for me so I don't forget anything. This time it is called ".make" (I use the leading ". so when I do rm -r * in my build dir I don't lose my scripts too.) This is what it looks like:

    make $* \
        && make -C contrib/src/gizmos $* \
        && make -C contrib/src/ogl CXXFLAGS="-DwxUSE_DEPRECATED=0" $* \
        && make -C contrib/src/stc $* \
        && make -C contrib/src/xrc $* 
    

    So you just use .make as if it where make, but don't forget to set the execute bit on .make first!:

    .make
    .make install
    

    When it's done you should have an installed set of files under /opt/wx/2.5 containing just wxWindows. Now to use this version of wxWindows you just need to add /opt/wx/2.5/bin to the PATH and set LD_LIBRARY_PATH (or DYLD_LIBRARY_PATH on OS X) to /opt/wx/2.5/lib.

  3. I also have a script to help me build wxPython and it is checked in to the CVS as wxWindows/wxPython/b, but probably don't want to use it as it's very cryptic and expects that you want to run SWIG, so if you don't have the latest patched up version of SWIG then you'll probably get stuck. So I'll just give the raw commands instead.

    We're not going to install the development version of wxPython with these commands, so it won't impact your already installed version of the latest release. You'll be able test with this version when you want to, and use the installed release version the rest of the time. If do want to install the development verison please read INSTALL.txt.

    If you have more than one version of Python on your system then be sure to use the version of Python that you want to use when running wxPython programs to run the setup.py commands below. I'll be using python2.3.

    Make sure that the first wx-config found on the PATH is the one you installed above, and then change to the wxWindows/wxPython dir and run the this command:

    cd wxPython
    python2.3 setup.py build_ext --inplace --debug
    

    If your new wx-config script is not on the PATH, or there is some other version of it found first, then you can add this to the command line to ensure your new one is used instead:

    WX_CONFIG=/opt/wx/2.5/bin/wx-config
    

    If you are building with GTK2 then add the following flags to the command line:

    WXPORT=gtk2 UNICODE=1
    

    If you are wanting to have the source files regenerated with swig, then you need to turn on the USE_SWIG flag and optionally tell it where to find the new swig executable, so add these flags:

    USE_SWIG=1 SWIG=/opt/swig/bin/swig
    

    If you get errors about wxGLCanvas or being unable to find libGLU or something like that then you can add BUILD_GLCANVAS=0 to the setup.py command line to disable the building of the glcanvas module.

    When the setup.py command is done you should have fully populated wxPython and wx packages locally in wxWindows/wxPython/wxPython and .../wx, with all the extension modules (*.so files) located in the wx package.

  4. To run code with the development verison of wxPython, just set the PYTHONPATH to the wxPython dir in the CVS tree. For example:

    export LD_LIBRARY=/opt/wx/2.5/lib
    export PYTHONPATH=/myprojects/wxWindows/wxPython
    cd /myprojects/wxWindows/wxPython/demo
    python2.3 demo.py
    

    OS X NOTE: You need to use "pythonw" on the command line to run wxPython applications. This version of the Python executable is part of the Python Framework and is allowed to interact with the display. You can also Double Click on a .py or a .pyw file from the finder (assuming that PythonLauncher is still associated with these file extensions) and it will launch the Framework version of Python for you. For information about creating Applicaiton Bundles of your wxPython apps please see the wiki and the mail lists.

    SOLARIS NOTE: If you get unresolved symbol errors when importing wxPython and you are running on Solaris and building with gcc, then you may be able to work around the problem by uncommenting a bit of code in setup.py and building again. Look for 'SunOS' in setup.py and uncomment the block containing it. The problem is that Sun's ld does not automatically add libgcc to the link step.

Building on Windows

The Windows builds currently require the use of Microsoft Visual C++. Theoretically, other compilers (such as mingw32 or the Borland compilers) can also be used but I've never done the work to make that happen. If you want to try that then first you'll want to find out if there are any tricks that have to be done to make Python extension modules using that compiler, and then make a few changes to setup.py to accomodate that. (And send the patches to me.) If you plan on using VisualStudio.Net (a.k.a. MSVC 7.1) keep in mind that you'll also have to build Python and any other extension modules that you use with that compiler because a different version of the C runtime likbrary is used. The Python executable that comes from PythonLabs and the wxPython extensions that I distribute are built with MSVC 6 with all the Service Packs applied.

If you want to build a debugable version of wxWindows and wxPython you will need to have also built a debug version of Python and any other extension modules you need to use. You can tell if you have them already if there is a _d in the file names, for example python_d.exe or python23_d.dll. If you don't need to trace through the C/C++ parts of the code with the debugger then building the normal (or hybrid) version is fine, and you can use the regular python executables with it.

Just like the unix versions I also use some scripts to help me build wxWindows, but I use some non-standard stuff to do it. So if you want to use them too you'll need to get a copy or 4DOS or 4NT from http://www.jpsoft.com/ and also a copy of unix-like cat and sed programs. You can also do by hand what my scripts are doing, but there are a lof steps involved and I won't be going into details here. There is a copy of my build scripts in wxWindowswxPythondistribmsw

  1. Set an environment variable to the root of the wxWindows source tree:

    set WXWIN=e:\projects\wxWindows
    
  2. Copy setup0.h to setup.h

    cd %WXWIN%includewxmsw copy setup0.h setup.h

  3. Edit %WXWIN%includewxmswsetup.h and change a few settings. Some of them are changed by my build scripts depending on the type of build (debug/hybrid, unicode/ansi). I change a few of the other defaults to have these values:

    wxDIALOG_UNIT_COMPATIBILITY    0
    wxUSE_DEBUG_CONTEXT            1
    wxUSE_MEMORY_TRACING           1
    wxUSE_DIALUP_MANAGER           0
    wxUSE_GLCANVAS                 1
    wxUSE_POSTSCRIPT               1
    wxUSE_AFM_FOR_POSTSCRIPT       0
    
  4. Make a %WXWIN%BIN directory and add it to the PATH. My build scripts will copy the wxWindows DLLs there.

  5. Change to the %WXWIN%buildmsw directory and copy my build scripts there.

  6. Use the .make.btm command to build wxWindows. It needs one command-line parameter which controls what kind of build(s) to do. Use one of the following:

    debug          Build debug version
    hybrid         Build hybrid version
    both           Both debug and hybrid
    debug-uni      Build a debug unicode library
    hybrid-uni     Hybrid unicode (see the pattern yet? ;-)
    both-uni       and finally both unicode libraries
    

    For example:

         .make hybrid
    
    You can also pass additional command line parameters as needed and
    they will all be passed on to the nmake commands, for example to
    clean up the build::
    
         .make hybrid clean
    
  7. When that is done it will have built the main wxWindows DLLs and also some of the contribs DLLs. There should be a ton of DLLs in %WXDIR%bin and lots of lib files and other stuff in %WXDIR%libvc_dll.

  8. Building wxPython on Windows is very similar to doing it for the unix systems. We're not going to install the development version of wxPython with these commands, so it won't impact your already installed version of the latest release. You'll be able to test with this version when you want to, and use the installed release version the rest of the time. If you ever do want to install the development verison please refer to INSTALL.txt.

    Change to the wxWindowswxPython dir and run the this command, makeing sure that you use the version of python that you want to build for (if you have more than one on your system):

    cd %WXWIN%\wxPython
    python setup.py build_ext --inplace 
    

    If you are wanting to have the source files regenerated with swig, then you need to turn on the USE_SWIG flag and optionally tell it where to find the new swig executable, so add these flags:

    USE_SWIG=1 SWIG=e:\projects\SWIG-cvs\swig.exe
    

    If you built a Unicode version of wxWindows and want to also build the Unicode version of wxPython then add this flag:

    UNICODE=1
    

    If you have a debug version of Python and wxWindows and want to build a debug version of wxPython too, add the --debug flag to the command line. You should then end up with a set of *_d.pyd files in the wx package and you'll have to run python_d.exe to use them. The debug and hybrid(release) versions can coexist.

    When the setup.py command is done you should have fully populated wxPython and wx packages locally in wxWindows/wxPython/wxPython and wxWindows/wxPython/wx, with all the extension modules (*.pyd files) located in the wx package.

  9. To run code with the development verison of wxPython, just set the PYTHONPATH to the wxPython dir in the CVS tree. For example:

    set PYTHONPATH=e:\projects\wxWindows\wxPython
    cd e:\projects\wxWindows\wxPython
    python demo.py