"Fredrik Lundh" <[EMAIL PROTECTED]> wrote in message news:[EMAIL PROTECTED] > "Saravanan D" wrote: > > > > > 03 023ffaa4 1e013182 00000000 055b1250 00637470 python23!cmp_outcome(int > > op > > > > = 88026108, struct _object * v = 0x00000001, struct _object * w = > > > > 0x00000000)+0xa9 [F:\Python-2.3.3\Python-2.3.3\Python\ceval.c @ 3880] > > > > 04 023ffb18 1e016ba4 014f3318 00000002 0099f170 > > python23!eval_frame(struct > > > > _frame * f = 0x053f2bfc)+0x542 > > [F:\Python-2.3.3\Python-2.3.3\Python\ceval.c > > > > @ 1965] > > > > 05 023ffb30 1e016a69 0099f170 023ffb7c 00000002 > > > > python23!fast_function(struct _object * func = <Memory access error>, > > struct > > > > _object *** pp_stack = <Memory access error>, int n = <Memory access > > error>, > > > > int na = <Memory access error>, int nk = <Memory access error>)+0x94 > > > > > > the above doesn't look very healthy. does your application involve > > > non-standard extensions (including extensions you've written your- > > > self)? does the extensions contain callbacks into Python code? > > > > > > > My application uses extension modules which devliers the callbacks to Python > > Code. Im using SWIG 1.3.19 version to generate relevant python interfaces > > for the C++ code. > > how are you accessing the callbacks? if you're doing that from threads > created at the C++ level, you must make sure that the thread state is > properly set up. see:
Callbacks have been called from the threads which are created in C++ code. Before calling the callback, the thread state has been properly set (as per python documentation). The application runs 3 to 4 days continuously with out having any issues. But, crash happens sparadically. (once in a week / once in 4 days ) > do you keep track of the thread state, so that callbacks run in the > same state as they were registered in? > > </F> > > > -- http://mail.python.org/mailman/listinfo/python-list