gh-95778: Use a note for the max digits error message

[vstinner]

Add _PyErr_AddNote() function to the internal C API.

• Issue: CVE-2020-10735: Prevent DoS by large int<->str conversions #95778

[vstinner]

@iritkatriel @Zac-HD: This PR adds the _PyErr_AddNote() function to the internal C API. It's a convenient function to add a note to the currently raised exception. The implementation is not trivial: it has to normalize the current exception to get an exception object and then call BaseException_add_note() which is a static function (not available in other C files).

I prefer to call directly BaseException_add_note() rather than calling the add_note() method, in case if a custom exception override the add_note() method. What I really want here is to add a note to the __notes__ attribute (and create it if it doesn't exist).

I'm not sure about the API in case of error. Currently, the function ignores any exception raised while trying to add a note to the exception :-( Do you think that an unraisable exception should be logged here? Or should we replace the currently raised exception with this new one? Maybe it would be acceptable if it's chained with the previous exception?

[vstinner]

I'm looking at which existing code might benefit from _PyErr_Note(). "Suggestions" added by PyErr_Display() with _Py_Offer_Suggestions() (issue #82711) is not a good fit: this code doesn't modify the exception on purpose. Or do you want to add suggestions in notes? If yes, we should maybe make sure that the note is not added multiple times if PyErr_Display() is called multiple times on the same exception ;-)

[vstinner]

Maybe _PyPegen_number_token() can use _PyErr_AddNote() for this code?

            // This acts as PyErr_Clear() as we're replacing curexc.
            PyErr_Fetch(&type, &value, &tb);
            Py_XDECREF(tb);
            Py_DECREF(type);
            /* Intentionally omitting columns to avoid a wall of 1000s of '^'s
             * on the error message. Nobody is going to overlook their huge
             * numeric literal once given the line. */
            RAISE_ERROR_KNOWN_LOCATION(
                p, PyExc_SyntaxError,
                t->lineno, -1 /* col_offset */,
                t->end_lineno, -1 /* end_col_offset */,
                "%S - Consider hexadecimal for huge integer literals "
                "to avoid decimal conversion limits.",
                value);
            Py_DECREF(value);

[iritkatriel]

Might be good to mention in the comment that there must be a current exception. I think otherwise it crashes or some assertion fails, not sure.

[iritkatriel]

The intention with exception notes was that you use it to add information that is not available at the time the exception is created and raised, but only after you caught it. This doesn't follow that pattern.

Why not just add one or two newlines to the msg?

[vstinner]

The intention with exception notes was that you use it to add information that is not available at the time the exception is created and raised, but only after you caught it. This doesn't follow that pattern. Why not just add one or two newlines to the msg?

Oh, that's far from issue #95778 scope, so I created the new issue #96958 to discuss that instead. I propose to continue the discussion there.

[vstinner]

I close the issue to continue discussing it in #96958.