gmane.comp.programming.swig

Re: segfault on inclusion of std_vector.i

Alex — 2008-07-04T14:42:00

okay, good call. here is an example.. the code runs as planned if i comment out : %include std_vector.i in the foo.i interface file. but with that %include statement, the code builds (with the warnings below) and segfaults when I run the executable. 07:35:57 $ make swig -Wall -c++ -ruby foo.i /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): Redundant redeclaration of 'map_bang', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): previous declaration of 'map_bang'. /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): Redundant redeclaration of '__delete__', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): previous declaration of '__delete__'. /usr/share/swig1.3/ruby/rubycontainer_extended.swg:137: Warning(322): Redundant redeclaration of 'map_bang', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:137: Warning(322): previous declaration of 'map_bang'. /usr/share/swig1.3/ruby/rubycontainer_extended.swg:137: Warning(322): Redun

Re: segfault on inclusion of std_vector.i

William S Fulton — 2008-07-03T21:48:31

No idea without a small standalone test case. We don't see those problems in the numerous testcases that ship with SWIG. These are due to extra high warning levels in your compiler. They ought to be suppressed, but can possibly be ignored. ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Re: Java, typemaps, and argout?

William S Fulton — 2008-07-03T21:45:38

Java SWIG doesn't support turning a parameter into a return type. Instead you have to think of some way to pass the parameter as a Java reference, so that the JNI code can modify it and those modifications can be seen from within Java. argout can be used to do this in the scripting languages only. See the Examples/java/typemap example for ideas on how to make an output parameter possible in Java. It uses the various.i library for changing the behaviour of char *. You should be able to use the same ideas for char **. William ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Re: Strange issue with 1.3.36

William S Fulton — 2008-07-03T21:33:45

Java doesn't use Lib/std/std_list.i. Are you explicitly including it? Running swig with the -v option should print out the search paths it is considering and that should not include Lib/std. The files in std are not designed for use with java, so if you are explicitly including std/std_list.i, you are going to have to think of something else, like use an older version perhaps? William ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Inheritance Question

Connor, Adam J — 2008-07-03T16:35:33

Greetings. After an hour or so of looking at the documentation, I figured maybe it would be faster to ask. Here's my situation: we are developing a Python class with two different implementations : in some environments, it will be pure Python, making network calls to another piece (non-Python). In other environments, it will use an existing C++ class instead of making the network calls. That's where SWIG comes in. Either way, a lot of the functionality of the class can be expressed in Python, so what we'd like to do is have both implementations share a common superclass that implements the common functionality. My question: Is it possible to get a SWIG-generated class to point to a specified custom Python base class? i.e., where the generated python class says something like ------------------------ snip ------------------------ import types try: _object = types.ObjectType _newclass = 1 except AttributeError: class _object : pass _newclass = 0 del types

Strange issue with 1.3.36

2008-07-03T15:17:24

Hello All, Just download 1.3.36 for some of the Java fixes, but when I try to use it against our current rule set, mind you none of our rules have changed between 1.3.31 which works and 1.3.36 which does not, I get the following error: Running : M:\src\make\..\..\src\thirdparty\swigwin-1.3.36\swig -Fmicrosoft -Wall -c++ -java -package com.emc.cst.cstBridge -module cstBridge -IX:\include -IX:\common -IX:\C_library\include -IX:\data -IX:\services -IX:\managers -IX:\ -IM:\src\make\..\..\src\thirdparty\swigwin-1.3.36\Lib\java -IM:\src\make\..\..\src\thirdparty\swigwin-1.3.36\Lib -outdir J:\cst\cstBridge -o m:\src\com\emc\csp\common\swigInterfaces\cst_wrap.cpp -oh m:\src\com\emc\csp\common\swigInterfaces\cst.h "m:\src\com\emc\csp\common\swigInterfaces\cst.i" M:\src\make\..\..\src\thirdparty\swigwin-1.3.36\Lib\std\std_list.i(78): Error: Syntax error in input(3). SwigGen Complete! Can anyone shed any light on what might be causing this? Thanks John -----------------------------------------------------------

Re: strange naming of constants in python wrapping

anirudh vij — 2008-07-03T14:21:58

Yup...that was thhe problem. Commenting that out fixed the problem. Thanks a lot! ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Re: strange naming of constants in python wrapping

Josh Cherry — 2008-07-03T13:51:31

On Thu, 3 Jul 2008, anirudh vij wrote: swig -help -python will help you see what's wrong with this. I know nothing about cmake, but set(CMAKE_SWIG_FLAGS -globals) looks like the culprit. Josh ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Re: strange naming of constants in python wrapping

anirudh vij — 2008-07-03T12:31:26

Well, actually cmake handles swig for me. the relevant lines in CMakeLists.txt are ------ if (SWIG_FOUND AND BUILD_PYTHON_WRAPPING) FIND_PACKAGE (PythonLibs REQUIRED) if (NOT_PythonLibs_FOUND) message("python libs or headers not found. Python Wrapping will not get built. Install python-dev and try again!") else (NOT_PythonLibs_FOUND) set(BUILD_PYTHON_WRAPPING_DEPENDENCIES_CORRECT yes) endif (NOT_PythonLibs_FOUND) endif (SWIG_FOUND AND BUILD_PYTHON_WRAPPING) ############################################################################### ## PYTHON WRAPPING if (BUILD_PYTHON_WRAPPING AND BUILD_PYTHON_WRAPPING_DEPENDENCIES_CORRECT) set(CMAKE_SWIG_FLAGS -globals) set(CMAKE_SWIG_OUTDIR ${SWIG_DIR}/bbp_sdk_python) make_directory(${CMAKE_SWIG_OUTDIR}) set(SWIG_MODULE_libbbp_sdk_python_EXTRA_DEPS ${SWIG_DIR}/array_wrapper.i ${SWIG_DIR}/Cell_Target.i ${SWIG_DIR}/Common.i ${SWIG_DIR}/Compartment_Report_Frame.i ${SWIG_DIR}/Compartment_Repo

Re: strange naming of constants in python wrapping

Josh Cherry — 2008-07-02T21:24:12

On Wed, 2 Jul 2008, anirudh vij wrote: -outdir is a SWIG command-line option. Normally cvar would be the name of the object called -outdir here. It looks like a messed-up command line is being used that contains "-globals -outdir". Josh ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

strange naming of constants in python wrapping

anirudh vij — 2008-07-02T21:11:03

Hi, I was wrapping some C++ code with python. The relevant part of the code goes like ----- #ifdef SWIG #define const %constant #endif const int foo=; #ifdef SWIG #undef const #endif ----- In the wrapper python file(not the cxx one), the constants appear as members of a type "-outdir" the exact lines of the file are: ---- -outdir = _Laboratory.-outdir foo = -outdir.foo ---- _Laboratory.so is the shared library. I am doing all this on Linux(Ubuntu 8.04) with swig 1.3.33(straight from ubuntu repos). Python version is 2.5 Now python flags the declaration of "-oudir" as an error(because of the minus sign). The funny thing is that the symbol "-outdir" actually exists in the shared _Laboratory.so. I verified this by doing a -- from _Laboratory import * --- then tab completion in ipython verifies that --outdir exists and has the member foo. Any ideas why swig chooses such a strange name ? thanks, Anirudh. ------------------------------------------------------------------------- Sponsored by: Sour

Java, typemaps, and argout?

J. Tang — 2008-07-02T20:30:18

I have read the SWIG manual, but am stumped on how to do the following: I have the C function: void recv_data(char **data, int *len); that reads stuff and puts data into the location pointed to by '*data' and the read size to '*len'. I want to write a Java wrapper that returns an instance of byte[] instead of java.lang.String. The best I could come up with is: jByteArray fake_recv_data() { char *data = NULL; int len = 0; jobject jresult = NULL; recv_data(self->sipc, &data, &len); jresult = (*jenv)->NewByteArray(jenv, len); (*jenv)->SetByteArrayRegion(jenv, jresult, 0, len, data); return jresult; }; I believe that I need a native function that will change the output from a parameter to a function return value. However, I do not know how to access jenv from within the native function. What is the recommended SWIG way of doing what I need? I am sure that there is some magic via %typemaps possible, but I am thoroughly stumped. '%

Re: unsigned char* manipulations

Christopher Barker — 2008-07-02T19:14:26

well, despite its name, a python string is really just an array of bytes, so it is an appropriate type for a char pointer. This is a pointer, how do you know how long the data block is? SWIG is going to need to know that to create a python string. -Chris

Re: unsigned char* manipulations

Josh Cherry — 2008-07-02T14:03:04

On Wed, 2 Jul 2008, hannit wrote: I would use %extend to add a function that returns a std::string. Josh ------------------------------------------------------------------------- Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW! Studies have shown that voting for your favorite open source project, along with a healthy diet, reduces your potential for chronic lameness and boredom. Vote Now at http://www.sourceforge.net/community/cca08

Re: unsigned char* manipulations

hannit — 2008-07-02T11:23:59

some more data: Seems like the problem is related to the fact that the original data is binary ans python (of course) converts it to a string. I assume that if I could read it into a StringIO or cStringIO buffers things would work fine. any idea how to do that ? hannit wrote:

segfault on inclusion of std_vector.i

Alex — 2008-07-02T08:08:09

I am using swig to generate a ruby interface for some c++ code I've written. I want to include std_vector so that (hopefully) I input ruby arrays into this method I have that accepts std::vector. anyways, if I build without.. %include std_vector.i ..my application builds fine, I can run my it but I cannot use this method in ruby. If I do include that I get this output: $ swig -Wall -c++ -ruby datajockey.i /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): Redundant redeclaration of 'map_bang', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): previous declaration of 'map_bang'. /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): Redundant redeclaration of '__delete__', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:136: Warning(322): previous declaration of '__delete__'. /usr/share/swig1.3/ruby/rubycontainer_extended.swg:137: Warning(322): Redundant redeclaration of 'map_bang', /usr/share/swig1.3/ruby/rubycontainer_extended.swg:137: Warnin

unsigned char* manipulations

hannit — 2008-07-02T07:32:56

Hello, I have a c++ class that has the following data functions: unsigned char* getSentDataP(); uint32_t getSentDataSize() const; I've managed wrapping them with python using swig by converting unsigned char* to char * and vice versa. unfortunately when checking the received data I get incorrect CRC. When dug deeper I figured out that I'm not getting the right amount of data. the c++ function declares on a certain size while len(p.getSentDataP()) returns a different size (the difference is inconsistent for different data packets). Would appreciate any advice Hannit

Re: sys.setdlopenflags(dl.RTLD_NOW|dl.RTLD_GLOBAL)

Andy Buckley — 2008-07-01T17:43:14

I also had a problem with this and would welcome a SWIG approach to either specifically handle the dlopen global issue, or to allow prepending of aribtrary Python code at the start of the generated module. But for now I do some ugly stuff with this autotools Makefile.am fragment and script: In Makefile(.am): ------ all-local: < at >sh fixpyrivet ------ NB. this is for a Python module called pyrivet. and here's fixpyrivet ------ if test -f pyrivet.py; then if test -z "$( grep RTLD_GLOBAL pyrivet.py )"; then cp pyrivet.py pyrivet.py.bak echo "## dlopen global workaround" > pyrivet.py echo "import sys, dl" >> pyrivet.py echo "flags = sys.getdlopenflags()" >> pyrivet.py echo "sys.setdlopenflags(flags | dl.RTLD_GLOBAL)" >> pyrivet.py echo "" >> pyrivet.py cat pyrivet.py.bak >> pyrivet.py rm -f pyrivet.py.bak fi fi ------ Very ugly, but it works on POSIX systems, which is all that I need for now. I may change it to use a slightly more careful import as in your example. A SW

Re: sys.setdlopenflags(dl.RTLD_NOW|dl.RTLD_GLOBAL)

Mathieu Malaterre — 2008-07-01T09:46:23

Done: http://gdcm.svn.sourceforge.net/viewvc/gdcm/trunk/Wrapping/Python/gdcm.py?view=markup I have a hand written gdcm module, that simply load the python swig-generated module... -Mathieu On Tue, Jul 1, 2008 at 11:25 AM, Mathieu Malaterre gmail.com> wrote:

Re: sys.setdlopenflags(dl.RTLD_NOW|dl.RTLD_GLOBAL)

Mathieu Malaterre — 2008-07-01T09:25:32

[retrying] So far I could only see one solution: http://patches.ubuntu.com/by-release/extracted/ubuntu/o/openbabel/2.1.1-4/some_platform_have_no_python_dl.patch So they ship a pregenerated swig .py file in the distribution that they manually modified. Is this the ony way of patch the swig generated file ? Isn't there an automated ? Thanks -Mathieu On Sat, Jun 28, 2008 at 1:35 AM, Mathieu Malaterre gmail.com> wrote:

Re: Swig-lua internals

Gedalia Pasternak — 2008-07-01T02:17:29

------------------------------------------------------------------------- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://sourceforge.net/services/buy/index.php_______________________________________________ Swig-user mailing list Swig-user< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/swig-user