[Python-checkins] r46671 - in python/trunk: Lib/bsddb/dbobj.py Lib/bsddb/test/test_all.py Lib/test/test_bsddb3.py Misc/NEWS Modules/_bsddb.c

[Tim Peters]

> Author: gregory.p.smith
> Date: Mon Jun  5 19:38:04 2006
> New Revision: 46671
>
> Modified:
>    python/trunk/Lib/bsddb/dbobj.py
>    python/trunk/Lib/bsddb/test/test_all.py
>    python/trunk/Lib/test/test_bsddb3.py
>    python/trunk/Misc/NEWS
>    python/trunk/Modules/_bsddb.c
> Log:
>   * add support for DBSequence objects [patch #1466734]

Coverity generated what appear to be legitimate complaints about this part:

> +#if (DBVER >= 43)
> +static PyObject*
> +DBSequence_construct(PyObject* self, PyObject* args, PyObject* kwargs)
> +{
> +    PyObject* dbobj = NULL;
> +    int flags = 0;
> +    static char* kwnames[] = { "db", "flags", NULL};
> +
> +    if (!PyArg_ParseTupleAndKeywords(args, kwargs, "O|i:DBSequence", kwnames, &dbobj, &flags))
> +        return NULL;
> +    if (dbobj == Py_None)
> +        dbobj = NULL;
> +    else if (dbobj && !DBObject_Check(dbobj)) {
> +        makeTypeError("DB", dbobj);
> +        return NULL;
> +    }
> +    return (PyObject* )newDBSequenceObject((DBObject*)dbobj, flags);
> +}
> +#endif

1. It shouldn't be possible for a successful PyArg_ParseTupleAndKeywords()
   call to leave dbobj == NULL, so the "dbobj &&" part of the last test
   appears to be pointless.

2. If dbobj == Py_None, then dbobj is forced to NULL, and
   newDBSequenceObject() will then be called with a NULL first argument.  This
   is the one Coverity spotted:  newDBSequenceObject() will segfault if it's
   called with a NULL first argument.

I don't know what this function _should_ do if dbobj==Py_None, but
segfaulting probably isn't the intent ;-)