Mailman 3 November 2006 - Cplusplus-sig

Re: [C++-sig] enum pickle issue in boost python
by Ralf W. Grosse-Kunstleve Nov. 28, 2006

Nov. 28, 2006

Thanks! Could you please also add pickle.dumps, pickle.loads tests to boost/libs/python/test/enum.py? If you could send the patched files (just to me, not the list) I'll test them locally and check them in. ----- Original Message ---- From: Shashank Bapat <shashankbapat(a)hotmail.com> To: c++-sig(a)python.org Sent: Monday, November 27, 2006 4:55:24 PM Subject: [C++-sig] enum pickle issue in boost python Ralf, I tried cvs HEAD and it has same problem. So I fixed it in my local copy (see diff below). This change is equivalent to class.cpp change made in version 1.50 of class.cpp: http://boost.cvs.sourceforge.net/boost/boost/libs/python/src/object/class.c… Thanks -Shashank =================================================================== RCS file: /cvsroot/boost/boost/libs/python/src/object/enum.cpp,v retrieving revision 1.9 diff -r1.9 enum.cpp 144c144 < module_name += '.'; --- > d["__module__"] = module_name; 146,147c146 < object result = (object(metatype))( < module_name + name, make_tuple(base), d); --- > object result = (object(metatype))(name, make_tuple(base), d); ____________________________________________________________________________________ Cheap talk? Check out Yahoo! Messenger's low PC-to-Phone call rates. http://voice.yahoo.com

1 0

enum pickle issue in boost python
by Shashank Bapat Nov. 28, 2006

Nov. 28, 2006

Ralf, I tried cvs HEAD and it has same problem. So I fixed it in my local copy (see diff below). This change is equivalent to class.cpp change made in version 1.50 of class.cpp: http://boost.cvs.sourceforge.net/boost/boost/libs/python/src/object/class.c… Thanks -Shashank =================================================================== RCS file: /cvsroot/boost/boost/libs/python/src/object/enum.cpp,v retrieving revision 1.9 diff -r1.9 enum.cpp 144c144 < module_name += '.'; --- > d["__module__"] = module_name; 146,147c146 < object result = (object(metatype))( < module_name + name, make_tuple(base), d); --- > object result = (object(metatype))(name, make_tuple(base), d); _________________________________________________________________ Share your latest news with your friends with the Windows Live Spaces friends module. http://clk.atdmt.com/MSN/go/msnnkwsp0070000001msn/direct/01/?href=http://sp…

1 0

enum pickle issue in boost python
by Shashank Bapat Nov. 27, 2006

Nov. 27, 2006

Ralf, Yes. I am aware of the boost pickle tutorial. It is useful if I have some complicated pickle logic in a class. But if you see enum_ has reduce and reduce_ex already defined in that case I shouldnt need to do anything special to get my enum pickled. ie providing additional methods for each enum_ to ensure correct pickling seems unnecessry. EnumTest is boost wrapper around an enum. Something like: namespace { enum Color { red = 1, green = 2, blue = 4 }; } BOOST_PYTHON_MODULE( EnumTest ) { boost::python::scope().attr("__doc__") = "Module to test Boost Python enums"; boost::python::enum_<Color>( "Color" ) .value( "red", red ) .value( "green", green ) .value( "blue", blue ) .export_values() ; } Check how class.cpp is differnt from enum.cpp in a way it processes module and name. I can fix it myself but I was wondering why is enum_ doing things differntly. May be this is just a mistake but may be there is a good reason. In that case I ll fix it and it will break something else somewhere ! Any case I ll try cvs HEAD and let you know. -Shashank _________________________________________________________________ Stay up-to-date with your friends through the Windows Live Spaces friends list. http://clk.atdmt.com/MSN/go/msnnkwsp0070000001msn/direct/01/?href=http://sp…

1 0

Re: [C++-sig] enum pickle issue in boost python
by Ralf W. Grosse-Kunstleve Nov. 27, 2006

Nov. 27, 2006

What is the EnumTest? Are you aware of this page? http://www.boost.org/libs/python/doc/v2/pickle.html I wrote the Boost.Python pickle support, but I never tried it with enum types. If you think Boost.Python needs tweaking, could you please work with the cvs HEAD, if you don't already? ----- Original Message ---- From: Shashank Bapat <shashankbapat(a)hotmail.com> To: c++-sig(a)python.org Sent: Sunday, November 26, 2006 3:02:00 PM Subject: [C++-sig] enum pickle issue in boost python Both pickle and cPickle seem to have problem pickling boost wrapped enum types. Pickle protocol used here is 2. Problem is causes by enum.cpp. In there __name__ and __module__ are intentionally set in certain way (which is different than how its done in class.cpp). So I suppose there is a valid reason why its that way. In that case the __reduce__ and __reduce_ex__ implementation for enum_ should take care of this so as to ensure proper pickling. Example: import EnumTest import cPickle try: print "Lets pickle some boost enum:" f = open("tmp1.txt", "w") cPickle.dump( EnumTest.Color.red , f, cPickle.HIGHEST_PROTOCOL) except: print "FAIL\n" try: print "Why is EnumTest.Color.__name__ = '%s' and \nEnumTest.Color.__module__= '%s' ?\n" % ( EnumTest.Color.__name__, EnumTest.Color.__module__) print "Lets hack it and try again:" f = open("tmp2.txt", "w") EnumTest.Color.__name__ = "Color" EnumTest.Color.__module__ = "EnumTest" l = (EnumTest.Color.red, EnumTest.Color.green, EnumTest.Color.blue) for e in l: cPickle.dump( e , f, cPickle.HIGHEST_PROTOCOL) f.close() print "PASS\n" print "Unpickle Test:" f = open("tmp2.txt", "r") for e in l: if e != cPickle.load(f): raise "Load Failed" print "PASS\n" except: print "FAIL\n" Thanks -Shashank ____________________________________________________________________________________ Do you Yahoo!? Everyone is raving about the all-new Yahoo! Mail beta. http://new.mail.yahoo.com

1 0

enum pickle issue in boost python
by Shashank Bapat Nov. 26, 2006

Nov. 26, 2006

Both pickle and cPickle seem to have problem pickling boost wrapped enum types. Pickle protocol used here is 2. Problem is causes by enum.cpp. In there __name__ and __module__ are intentionally set in certain way (which is different than how its done in class.cpp). So I suppose there is a valid reason why its that way. In that case the __reduce__ and __reduce_ex__ implementation for enum_ should take care of this so as to ensure proper pickling. Example: import EnumTest import cPickle try: print "Lets pickle some boost enum:" f = open("tmp1.txt", "w") cPickle.dump( EnumTest.Color.red , f, cPickle.HIGHEST_PROTOCOL) except: print "FAIL\n" try: print "Why is EnumTest.Color.__name__ = '%s' and \nEnumTest.Color.__module__= '%s' ?\n" % ( EnumTest.Color.__name__, EnumTest.Color.__module__) print "Lets hack it and try again:" f = open("tmp2.txt", "w") EnumTest.Color.__name__ = "Color" EnumTest.Color.__module__ = "EnumTest" l = (EnumTest.Color.red, EnumTest.Color.green, EnumTest.Color.blue) for e in l: cPickle.dump( e , f, cPickle.HIGHEST_PROTOCOL) f.close() print "PASS\n" print "Unpickle Test:" f = open("tmp2.txt", "r") for e in l: if e != cPickle.load(f): raise "Load Failed" print "PASS\n" except: print "FAIL\n" Thanks -Shashank _________________________________________________________________ View Athletes Collections with Live Search http://sportmaps.live.com/index.html?source=hmemailtaglinenov06&FORM=MGAC01

1 0

Re: [C++-sig] Boost python in OS X, with non-framework python 2.4
by Ralf W. Grosse-Kunstleve Nov. 26, 2006

Nov. 26, 2006

> From: Lars Yencken <lars.yencken(a)gmail.com> > > My guess is that part of the python interface changed between > versions (this one type/typedef of Py_ssize_t?). I imagine that using > a more recent dev version of boost python might get around this, Yes, exactly. The current boost cvs HEAD should work with any Python >= 2.2. I tried the HEAD successfully the day before yesterday, with a whole bunch of compilers. Cheers, Ralf ____________________________________________________________________________________ Yahoo! Music Unlimited Access over 1 million songs. http://music.yahoo.com/unlimited

1 0

need 16 arguments to a function...
by anders langlands Nov. 26, 2006

Nov. 26, 2006

Can I change the max arguments limit for methods without recompiling boost::python? I need an initializer for a 4x4 matrix, hence the need for 16 floats. A

4 7

Re: [C++-sig] Boost python in OS X, with non-framework python 2.4
by Ralf W. Grosse-Kunstleve Nov. 26, 2006

Nov. 26, 2006

I successfully built a large set of Boost.Python extensions under Mac OS 10.4 Intel with Python 2.3, 2.4, and 2.5, and under Mac OS 10.3 PPC with Python 2.3 and 2.5 (I bet 2.4 will also work, but I didn't try). I only managed to build successfully against Python frameworks. I don't know if it is possible to work with a normal/plain Python build as you get e.g. from fink. You can install a Python framework from sources like this: ./configure --enable-framework make make frameworkinstall For the last step you'll need to be root. The installation will go to "/Library/Frameworks/Python.Framework/Versions/2.?/" and "/Applications/MacPython 2.?", and you'll get a new set of softlinks in /usr/local. I'm using a scons-based build system, therefore I don't know how to convince bjam to use the Python installation you want. Hopefully someone else can help out. ____________________________________________________________________________________ Do you Yahoo!? Everyone is raving about the all-new Yahoo! Mail beta. http://new.mail.yahoo.com

2 2

Crash during exception translation using Intel Compiler 9.1 and optimized builds
by Pierre Baillargeon Nov. 26, 2006

Nov. 26, 2006

We're using the framework to translate C++ exception to Python exceptions. The code works in debug builds but crashes in optimized builds. Were using Intel Compiler 9.1 on Windows. Has anyone experienced this? Any possible workaround? The symptoms are: - a crash under _is_exception() in MSVCR80.DLL, called from handle_exception_impl(). - The bottom of the stack is obviously bad: below _is_exception_typeof() we find tanhf (float's hyperbolic tangent? Microsoft obviously found a new type-system theorem involving high-level maths), then a long boost template and 0x0. - The arguments received in the long boost::python type (which I cannot see because the name is too long for windbg) are a0, a1 and f, and f is 0 (which might explain to jump to 0x0). Currently the only work-around we've found is to disable the exception translation. From the look of the stack, it seems that the problem is in the linked list of translators. Could it be an order-of-initialization issue? Or a temporary value that should stick around or that the compiler mistakenly optimizes away?

3 3

Boost python in OS X, with non-framework python 2.4
by Lars Yencken Nov. 26, 2006

Nov. 26, 2006

Hello! [apologies for long post ahead] I've been using boost python with some of my research code for a little while on linux, and have really enjoyed it. I've been having trouble on my new mac laptop though, since (a) I have a large codebase that needs python2.4, and (b) I can't seem to compile and link boost python extensions against python2.4 as provided by fink or darwin ports. No matter what I try, I can only get extensions successfully compiled for the framework python2.3 which comes with OS X. I know this because with python 2.4 (from fink) my code gives Fatal Python error: Interpreter not initialized (version mismatch?) Abort trap whereas this doesn't occur with 2.3, where the extension works fine. If I use boost.python from fink or darwin ports, the problem is the same. I've tried installing my own version, using configure arguments --with-python=/sw/bin/python2.4 --with-python-root=/sw --with-python-version=2.4 --with-libraries=python. Here's the output from the previous compilation stage for my extension, which appears to be successful: g++ -o dataType.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost dataType.cpp g++ -o entry.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost entry.cpp g++ -o alignment.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost alignment.cpp g++ -o ambiguousAlignment.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost ambiguousAlignment.cpp g++ -o utils.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost utils.cpp g++ -o dataType.dylib -dylib -bundle -L/sw/lib -lpython2.4 -lboost_python -lm dataType.o entry.o alignment.o ambiguousAlignment.o utils.o I did see some earlier posts on this, but they seemed a bit unresolved. From my understanding, regardless of command line options you give to configure, if you compile boost.python using the darwin toolset, it links against the framework python. I tried building using the gcc toolset, and specifying --with-python=/sw/bin/python2.4 --with-python-root=/sw --with-python-version=2.4 --with-libraries=python. If you do this, the g++ arguments are actually wrong for OS X, and you have to manually perform the final linking step yourself. This works, if you change c++ -s -fPIC -shared -o <target dylib> <long list of object files> to c++ -dynamiclib -dynamic -o <target dylib> <long list of object files> -L/sw/lib -lpython2.4 for the targets libboost_python-gcc-1_33_1.dylib and libboost_python-gcc-d-1_33_1.dylib (leaving in the -g flag for the debug library). I can then install it. After all this, I try recompiling. g++ -o dataType.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost dataType.cpp g++ -o entry.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost entry.cpp g++ -o alignment.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost alignment.cpp g++ -o ambiguousAlignment.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost ambiguousAlignment.cpp g++ -o utils.o -c -Wall -O3 -fPIC -I/sw/include/python2.4 -I/usr/local/include/boost utils.cpp g++ -o dataType.so -dylib -bundle -L/sw/lib -lpython2.4 -lboost_python-gcc-1_33_1 -lm dataType.o entry.o alignment.o ambiguousAlignment.o utils.o Now, when I test, I get same error, but with python2.3, python2.4 and python2.5. This to me indicates that either my library, or boost's library, has been compiled with a mixture of python flavours, acceptable to none. Does anyone have any ideas on what else I could try, or suggestions in the case I'm barking up the wrong tree? Many thanks, Lars

1 0