Re: [PATCH v3 0/4] Python safety initial work

[Tom de Vries <tdevries@suse.de>]

On 8/7/26 9:48 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> I found this code:
> Tom> ...
> Tom>       /* Note this returns a borrowed reference.  */
> Tom>       PyObject *arg = PyTuple_GetItem (args, i);
> Tom> ...
> Tom> and decided to try to convert all PyTuple_GetItem calls.  The result
> Tom> of that exercise is attached.
> 
> Tom> I ended up also touching the wrapper function:
> Tom> ...
> Tom>  static inline gdbpy_borrowed_ref<>
> Tom>  gdbpy_tuple_get_item (gdbpy_borrowed_ref<> tuple, Py_ssize_t pos)
> Tom>  {
> Tom> -  PyObject *result = PyTuple_GetItem (tuple, pos);
> Tom> +  gdbpy_opt_borrowed_ref<> result = PyTuple_GetItem (tuple, pos);
> Tom>    if (result == nullptr)
> Tom>      throw gdb_python_exception ();
> Tom> -  return result;
> Tom> +  return (PyObject *)result;
> Tom>  }
> Tom> ...
> Tom> but perhaps you left that out intentionally?
> 
> Yeah, in this spot it didn't seem to really be necessary.
> The idea of the wrapper file is to eventually isolate all "raw" calls to
> the Python API.
> 
> Tom> I haven't used the gdbpy_tuple_get_item wrapper, AFAICT it was not
> Tom> applicable.
> 
> The long term goal is to convert everything to call the wrappers.
> This way exception handling will be enforced.

Ok, I've started over and converted valpy_call, now indeed using 
gdbpy_tuple_get_item.

Thanks,
- Tom