[capi-sig] getting popen error after long term use

[Simon Kiertscher]

Hi all,

I hope I post on the correct list, sorry if not.

I have written a C-daemon and I use the C/Python API for calling bash
commands via python using popen.

My daemon seems to run without errors, but after a unknown period of
time (sometimes the daemon runs a week without problems, sometimes a
month) the program isn't able to run the python scripts any more.

I get the following error:
Exception exceptions.AttributeError: "'Popen' object has no attribute
'_child_created'" in <bound method Popen.__del__ of <subprocess.Popen
object at 0x2aaaaf1fa910>> ignored

I googled the error but didn't get smarter.

So I have no idea if the error is on the python site or on the c site.
And I have no idea why the error occurs after a long time of run. I
would think (because of the unknown time fact) there is a memory leak or
something like that, what else could it be? Has anyone an idea would I
could do wrong?

My c calls look like that:

int shutdown(const char *node_address, struct content *content)
{
	assert(node_address != NULL);
	PyObject *funk, *prm, *ret;

	funk = PyObject_GetAttrString(content->modul, "shutdown");
	if(funk == NULL)
	{
		syslog(LOG_ERR, "[MASTER] [shutdown] CANT LOAD SHUTDOWN FROM
%s.PY",content->scriptname);
		return(EXIT_FAILURE);
	}
	
	prm = Py_BuildValue("(s)", node_address);
	if(prm == NULL)
	{
		Py_CLEAR(funk);
		syslog(LOG_ERR, "[MASTER] [shutdown] CANT BUILD VALUE");
		return(EXIT_FAILURE);
	}
	
	ret = PyObject_CallObject(funk, prm);
	if(ret == NULL)
	{
		Py_CLEAR(funk);
		Py_CLEAR(ret);
		syslog(LOG_ERR, "[MASTER] [shutdown] CANT CALL FUNCTION OBJECT");
		return(EXIT_FAILURE);
	}
	
	if(PyInt_AsLong(ret)!=0)
	{
		Py_CLEAR(prm);
		Py_CLEAR(funk);
		Py_CLEAR(ret);
		syslog(LOG_ERR, "[MASTER] [shutdown] ERROR ON PyInt_AsLong");
		return(EXIT_FAILURE);
	}
	
	Py_CLEAR(ret);
	Py_CLEAR(prm);
	Py_CLEAR(funk);

	return EXIT_SUCCESS;
}

And my Python calls look like that:

def shutdown(node_address):
	p = subprocess.Popen(["ipmitool","-l", "LAN", "-H",node_address,"-U",
"root", "-P", config.IPMIPassword, "power",
"soft"],stdout=subprocess.PIPE, stderr=subprocess.PIPE)
	p.communicate()
	return p.returncode

When I kill the daemon after noticing that it can't call he functions
any more, I get the following errors in addition:

Exception exceptions.IndexError: 'list index out of range' in 'garbage
collection' ignored
Fatal Python error: unexpected exception during garbage collection

I strongly hope someone has an idea about that.
Sorry for the bad English.

Best Regards,
Simon