boost python problem

[monstro]

Ok. So I wrote a simple boost python application. I put the exception handling code in. Now, as a test, I tried to run (purposely) a python script with a syntax error in it. The application crashes each time (Runtime Error!). How can I change the app so that it handles python script syntax errors without crashing?

[jlm699]

Could you give us the offending code and the Traceback? Your explanation is devoid of any information that could allow us to help you.

Help US help you!

[monstro]

Could you give us the offending code and the Traceback? Your explanation is devoid of any information that could allow us to help you.

Help US help you!

I was unsure whether to post this in the C++ or python forums, but here goes:

bool ScriptInterface::run_python_script(const std::string& filename, int argc, char** argv)
{    

	FILE* fp = NULL;    
	bool success = true;    
	std::string logErrors;



	if (PyImport_AppendInittab("dt_embedded_python_ext", initdt_embedded_python_ext) == -1)
	{
		throw runtime_error("could not register module __dt_embedded_python_ext__");
	}     


	Py_Initialize();    

	try{
		
	    object main_module = extract<object>(PyImport_AddModule("__main__"))();
		object main_namespace(main_module.attr("__dict__"));


		fp = fopen(filename.c_str(), "r");     
		if (!fp)
		{
			throw runtime_error(filename + ": " + strerror(errno));
		}       

		
		boost::python::exec_file(filename.c_str(), main_namespace, main_namespace);

	}    
	catch (const exception& e)
	{       
			fprintf(stderr, "Exception caught: %s\n", e.what());
			logErrors = "Exception caught: ";
			logErrors += e.what(); 
			LogMessage((char*)logErrors.c_str());
			success  = false;
	}    

	if (fp)
		fclose(fp);    


	return success;
}

Anyway, as you can see, there is a try catch exception handler in the code. The problem is that it is never executed on a simple python scipt syntax error, as the program crashes immediately after exec is called.

[monstro]

Well, I found that if I use a catch(...) I can trap the error and even keep the app from crashing by using

boost::python::handle_exception();

I am trying to redirect the stderr message from a python script syntax error to a C++ windows app. Now, I need to trap this message somehow. I changed the catch to the following:

catch (...)
 	{       
		 if(PyErr_Occurred())
		 {
			 PyErr_Print();

			 boost::python::object sys(boost::python::handle<>(PyImport_ImportModule("sys")));
			 boost::python::object err = sys.attr("stderr");
			 std::string err_text = boost::python::extract<std::string>(err.attr("getvalue")());
			 logErrors = "Exception caught: ";
			 logErrors += err_text;
			 LogMessage((char*)logErrors.c_str());
			 boost::python::handle_exception();
		 }

		 success  = false;
	}

Now I get a crash when

std::string err_text = boost::python::extract<std::string>(err.attr("getvalue")());

is called. It appears that the "getvalue" is returning a bad pointer or something...this doesn't make sense to me.

[Gribouillis]

May be you could try this

catch (...)
 	{       
		 if(PyErr_Occurred())
		 {
			 PyErr_Print();

			 boost::python::object traceback(boost::python::handle<>(PyImport_ImportModule("traceback")));
			 std::string err_text = boost::python::extract<std::string>(traceback.attr("format_exc")());
			 logErrors = "Exception caught: ";
			 logErrors += err_text;
			 LogMessage((char*)logErrors.c_str());
			 boost::python::handle_exception();
		 }

		 success  = false;
	}

I don't think the sys.stderr.getvalue exists :)

[monstro]

I tried your suggestion. It returns "None" in err_text. This isn't what is returned by the python interpreter if run from the command line.

I don't understand why it is so difficult to get the stderr message text from python...this seems like it should be a really common problem to people running embedded python.

[monstro]

I think I am giving this up. I don't know if there is a way to trap python syntax errors in boost C++, or even in the Python C API. All I get are empty strings or exceptions thrown.

I tried many different methods, but all fail. If anyone has a solution...that would be absolutely brilliant. peace out...