Re: [PATCH v1] gdb/python: fix memory leak in gdb_py_tp_name

[Tom Tromey <tom@tromey.com>]

>>>>> "Matthieu" == Matthieu Longo <matthieu.longo@arm.com> writes:

Matthieu> The proposed approach consists in changing gdb_py_tp_name() to return a
Matthieu> std::string

That seems fine.

Matthieu> A unfortunate side effect of this fix is that every call sites where the
Matthieu> tp_name is printed, must now use `.c_str()' because PyErr_Format()
Matthieu> and its siblings cannot handle std::string.

Yeah, that's unfortunate but normal.

Matthieu>  /* Return the type's fully qualified name from a PyTypeObject.  */
Matthieu> -const char *
Matthieu> +std::string
Matthieu>  gdb_py_tp_name (PyTypeObject *py_type) noexcept
Matthieu>  {
Matthieu>  #if PY_VERSION_HEX >= 0x030d0000
Matthieu>    /* Note: PyType_GetFullyQualifiedName() was added in version 3.13, and is
Matthieu>       part of the stable ABI since version 3.13.  */
Matthieu> -  PyObject *fully_qualified_name = PyType_GetFullyQualifiedName (py_type);
Matthieu> +  gdbpy_ref<> fully_qualified_name (PyType_GetFullyQualifiedName (py_type));
Matthieu>    if (fully_qualified_name == nullptr)
Matthieu>      return nullptr;

First, returning nullptr here seems really weird.  I think this should
cause a crash?  In C++23 this constructor is even deleted.

Second, if PyType_GetFullyQualifiedName fails, this will violate the
invariants gdb normally uses for error handling.

Matthieu> -  return PyUnicode_AsUTF8AndSize (fully_qualified_name, nullptr);
Matthieu> +  return PyUnicode_AsUTF8AndSize (fully_qualified_name.get (), nullptr);
 
I think this can also fail.

Matthieu> +  gdbpy_ref<> qualname (PyType_GetQualName (py_type));
Matthieu>    if (qualname == nullptr)
Matthieu>      return nullptr;
 
Matthieu> -  return PyUnicode_AsUTF8AndSize (qualname, nullptr);
Matthieu> +  return PyUnicode_AsUTF8AndSize (qualname.get (), nullptr);
 
Similar stuff in these spots.

Tom