C Wrapper Function, crashing Python?

[Java and Swing]

static PyObject *wrap_doStuff(P yObject *self, PyObject *args) {
// this will store the result in a Python object
PyObject *finalResult;

// get arguments from Python
char *result = 0;
char *in= 0;
char *aString = 0;
char *bString = 0;
MY_NUM *a;
MY_NUM *b;
int ok = PyArg_ParseTupl e(args, "sss", &in, &aString, &bString);
if (!ok) return 0;

// do work to get a and b
// count - returns an int; GetVal - returns a char *
a = GetVal(aString, count(aString, ","));
b = GetVal(bString, count(bString, ","));

// make function call, which returns a char *
result = doStuff(in, a, b);

// save result in Python string
finalResult = PyString_FromSt ring(result);

// free memory
PyMem_Free(resu lt);
PyMem_Free(a);
PyMem_Free(b);

// return the result as a Python string
return finalResult;
}

....from python I can call this function 4 times...works fine. WHen I
call it for the fifth time python.exe crashes. im thinking some memory
problem in the wrapper function perhaps...but I am not sure. The
actually C function, doStuff can be called 5, 6,7...N times without a
problem
so i know its gotta be my wrapper.

Any ideas? Thanks!

[Bernhard Herzog]

"Java and Swing" <co*******@gmai l.com> writes:

thanks for the tip, however even when I do not free aString or bString,
i'm still crashing at the malloc in the c function, not the wrapper.

Do you have any more places where you use free incorrectly? In my
experience, calling free with invalid values can corrupt the data
structures used by the memory allocator in such a way that subsequent
malloc calls crash.

Bernhard

--
Intevation GmbH http://intevation.de/
Skencil http://skencil.org/
Thuban http://thuban.intevation.org/

[Java and Swing]

As far as my C Wrapper functions are concerned...I no longer have the
need for free(...). I do use PyMem_Free, for structures I allocated by
using PyMem_New(...).

In my C code I do have things such as...

char *foo(const char *in) {
char *tmp;
tmp = (char *) malloc((strlen( in) * sizeof(char)) + 1);
strcpy(tmp, in);
...
...
free(tmp);
return someValue;
}

Is that appropriate? I was under the impression that when you malloc
memory, you free it when done.

I also have things like...

char *bar(char *in) {
char *results, *finalResults;
results = (char *) malloc(...);
finalResults = results;

while (...) { *results++ = some_val; }

return finalResults;
}

....is that correct?

As I mentioned earlier, when I run these functions from C I have no
troubles...I can run them 5, 10, 15 times, etc. From Python, using my
wrapper function is when I have trouble.

Thanks in advance.

Bernhard Herzog wrote:

"Java and Swing" <co*******@gmai l.com> writes:

thanks for the tip, however even when I do not free aString or bString,
i'm still crashing at the malloc in the c function, not the wrapper.

Do you have any more places where you use free incorrectly? In my
experience, calling free with invalid values can corrupt the data
structures used by the memory allocator in such a way that subsequent
malloc calls crash.

Bernhard

--
Intevation GmbH http://intevation.de/
Skencil http://skencil.org/
Thuban http://thuban.intevation.org/

[Bernhard Herzog]

"Java and Swing" <co*******@gmai l.com> writes:

char *foo(const char *in) {
char *tmp;
tmp = (char *) malloc((strlen( in) * sizeof(char)) + 1);
strcpy(tmp, in);
...
...
free(tmp);
return someValue;
}

Is that appropriate? I was under the impression that when you malloc
memory, you free it when done.
Looks fine. I hope someValue does not point somewhere into the tmp
buffer, though.
I also have things like...

char *bar(char *in) {
char *results, *finalResults;
results = (char *) malloc(...);
finalResults = results;

while (...) { *results++ = some_val; }

return finalResults;
}

Seems OK, too, assuming the results buffer is big enough.

Bernhard

--
Intevation GmbH http://intevation.de/
Skencil http://skencil.org/
Thuban http://thuban.intevation.org/

[Java and Swing]

Bernhard Herzog wrote:

"Java and Swing" <co*******@gmai l.com> writes:

char *foo(const char *in) {
char *tmp;
tmp = (char *) malloc((strlen( in) * sizeof(char)) + 1);
strcpy(tmp, in);
...
...
free(tmp);
return someValue;
}

Is that appropriate? I was under the impression that when you malloc
memory, you free it when done.
Looks fine. I hope someValue does not point somewhere into the tmp
buffer, though.

someValue doesn't.

I also have things like...

char *bar(char *in) {
char *results, *finalResults;
results = (char *) malloc(...);
finalResults = results;

while (...) { *results++ = some_val; }

return finalResults;
}

Seems OK, too, assuming the results buffer is big enough.

results = (char *) malloc((sizeof( char) * strlen(in) + 1);
..should be big enough. At any rate, it works four times in a row
before it fails.

I wonder what else or how else to solve this. Why is malloc'ing
the memory crashing it!?

[Tony Nelson]

In article <11************ ********@o13g20 00cwo.googlegro ups.com>,
"Java and Swing" <co*******@gmai l.com> wrote:

one more update...

if I remove PyMem_Free and free(...) ...so no memory clean up...I can
still only call doStuff 4 times, the 5th attemp crashes Python.

Java and Swing wrote:

update:
if I use C's free(result), free(a) free(b) instead of PyMem_Free...I
only get one successfuly use/call of doStuff.

i.e.
// this works
doStuff(...)

// python crashes here
doStuff(...)

Java and Swing wrote:

static PyObject *wrap_doStuff(P yObject *self, PyObject *args) {
// this will store the result in a Python object
PyObject *finalResult;

// get arguments from Python
char *result = 0;
char *in= 0;
char *aString = 0;
char *bString = 0;
MY_NUM *a;
MY_NUM *b;
int ok = PyArg_ParseTupl e(args, "sss", &in, &aString, &bString);
if (!ok) return 0;

// do work to get a and b
// count - returns an int; GetVal - returns a char *
a = GetVal(aString, count(aString, ","));
b = GetVal(bString, count(bString, ","));

// make function call, which returns a char *
result = doStuff(in, a, b);

// save result in Python string
finalResult = PyString_FromSt ring(result);

// free memory
PyMem_Free(resu lt);
PyMem_Free(a);
PyMem_Free(b);

// return the result as a Python string
return finalResult;
}

...from python I can call this function 4 times...works fine. WHen I
call it for the fifth time python.exe crashes. im thinking some memory
problem in the wrapper function perhaps...but I am not sure. The
actually C function, doStuff can be called 5, 6,7...N times without a
problem
so i know its gotta be my wrapper.

Any ideas? Thanks!

I think your wrapper should look something like:

static PyObject *wrap_doStuff(P yObject *self, PyObject *args)
{
// this will store the result in a Python object
PyObject *finalResult;

// get arguments from Python
char *result = 0;
char *in= 0;
char *aString = 0;
char *bString = 0;
MY_NUM *a;
MY_NUM *b;

int ok = PyArg_ParseTupl e(args, "sss", &in, &aString, &bString);
if (!ok) return 0;

// do work to get a and b
// count - returns an int; GetVal - returns a char *
a = GetVal(aString, count(aString, ","));
b = GetVal(bString, count(bString, ","));

// make function call, which returns a char *
result = doStuff(in, a, b);

// save result in Python string
finalResult = PyString_FromSt ring(result);

// free memory
free(result);
free(a);
free(b);

// return the result as a Python string
return finalResult;
}

You must match malloc() with free(), and PyMem_Malloc() with
PyMem_Free(). Malloc() and free() usually crash /after/ the call that
did the damage. You may wish to avail yourself of your platform's
malloc debugging facilities.

Note that I don't do this stuff in C, I do it in pyrex, and I'm new to
it anyway, so there may still be something wrong. Unless you are
determined to learn how to do this in C, I think you should switch to
pyrex.
_______________ _______________ _______________ _______________ ____________
TonyN.:' *firstname*nlsn ews@georgea*las tname*.com
' <http://www.georgeanels on.com/>