Bug 1952351

Note that the INTERNALERROR is tracked in https://github.com/pytest-dev/pytest/pull/8227 but it IMHO only happens for tests that fail.

Error with patched pytest:


=================================== FAILURES ===================================
_ test_ambiguities[foo: bar | baz\nbar: NAME\nbaz: NAME\n-foo is ambiguous.*given a PythonTokenTypes\\.NAME.*bar or baz] _

grammar = 'foo: bar | baz\nbar: NAME\nbaz: NAME\n'
error_match = 'foo is ambiguous.*given a PythonTokenTypes\\.NAME.*bar or baz'

    @pytest.mark.parametrize(
        'grammar, error_match', [
            ['foo: bar | baz\nbar: NAME\nbaz: NAME\n',
             r"foo is ambiguous.*given a PythonTokenTypes\.NAME.*bar or baz"],
            ['''foo: bar | baz\nbar: 'x'\nbaz: "x"\n''',
             r"foo is ambiguous.*given a ReservedString\(x\).*bar or baz"],
            ['''foo: bar | 'x'\nbar: 'x'\n''',
             r"foo is ambiguous.*given a ReservedString\(x\).*bar or foo"],
            # An ambiguity with the second (not the first) child of a production
            ['outer: "a" [inner] "b" "c"\ninner: "b" "c" [inner]\n',
             r"outer is ambiguous.*given a ReservedString\(b\).*inner or outer"],
            # An ambiguity hidden by a level of indirection (middle)
            ['outer: "a" [middle] "b" "c"\nmiddle: inner\ninner: "b" "c" [inner]\n',
             r"outer is ambiguous.*given a ReservedString\(b\).*middle or outer"],
        ]
    )
    def test_ambiguities(grammar, error_match):
        with pytest.raises(ValueError, match=error_match):
>           generate_grammar(grammar, tokenize.PythonTokenTypes)

test/test_pgen2.py:357: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

bnf_grammar = 'foo: bar | baz\nbar: NAME\nbaz: NAME\n'
token_namespace = <enum 'PythonTokenTypes'>

    def generate_grammar(bnf_grammar: str, token_namespace) -> Grammar:
        """
        ``bnf_text`` is a grammar in extended BNF (using * for repetition, + for
        at-least-once repetition, [] for optional parts, | for alternatives and ()
        for grouping).
    
        It's not EBNF according to ISO/IEC 14977. It's a dialect Python uses in its
        own parser.
        """
        rule_to_dfas = {}
        start_nonterminal = None
        for nfa_a, nfa_z in GrammarParser(bnf_grammar).parse():
            # _dump_nfa(nfa_a, nfa_z)
            dfas = _make_dfas(nfa_a, nfa_z)
            # _dump_dfas(dfas)
            # oldlen = len(dfas)
            _simplify_dfas(dfas)
            # newlen = len(dfas)
            rule_to_dfas[nfa_a.from_rule] = dfas
            # print(nfa_a.from_rule, oldlen, newlen)
    
            if start_nonterminal is None:
                start_nonterminal = nfa_a.from_rule
    
        reserved_strings: Mapping[str, ReservedString] = {}
        for nonterminal, dfas in rule_to_dfas.items():
            for dfa_state in dfas:
                for terminal_or_nonterminal, next_dfa in dfa_state.arcs.items():
                    if terminal_or_nonterminal in rule_to_dfas:
                        dfa_state.nonterminal_arcs[terminal_or_nonterminal] = next_dfa
                    else:
                        transition = _make_transition(
                            token_namespace,
                            reserved_strings,
                            terminal_or_nonterminal
                        )
                        dfa_state.transitions[transition] = DFAPlan(next_dfa)
    
>       _calculate_tree_traversal(rule_to_dfas)

parso/pgen2/generator.py:278: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

nonterminal_to_dfas = {'bar': [<DFAState: bar is_final=False>, <DFAState: bar is_final=True>], 'baz': [<DFAState: baz is_final=False>, <DFAState: baz is_final=True>], 'foo': [<DFAState: foo is_final=False>, <DFAState: foo is_final=True>]}

    def _calculate_tree_traversal(nonterminal_to_dfas):
        """
        By this point we know how dfas can move around within a stack node, but we
        don't know how we can add a new stack node (nonterminal transitions).
        """
        # Map from grammar rule (nonterminal) name to a set of tokens.
        first_plans = {}
    
        nonterminals = list(nonterminal_to_dfas.keys())
        nonterminals.sort()
        for nonterminal in nonterminals:
            if nonterminal not in first_plans:
                _calculate_first_plans(nonterminal_to_dfas, first_plans, nonterminal)
    
        # Now that we have calculated the first terminals, we are sure that
        # there is no left recursion.
    
        for dfas in nonterminal_to_dfas.values():
            for dfa_state in dfas:
                transitions = dfa_state.transitions
                for nonterminal, next_dfa in dfa_state.nonterminal_arcs.items():
                    for transition, pushes in first_plans[nonterminal].items():
                        if transition in transitions:
                            prev_plan = transitions[transition]
                            # Make sure these are sorted so that error messages are
                            # at least deterministic
                            choices = sorted([
                                (
                                    prev_plan.dfa_pushes[0].from_rule
                                    if prev_plan.dfa_pushes
                                    else prev_plan.next_dfa.from_rule
                                ),
                                (
                                    pushes[0].from_rule
                                    if pushes else next_dfa.from_rule
                                ),
                            ])
>                           raise ValueError(
                                "Rule %s is ambiguous; given a %s token, we "
                                "can't determine if we should evaluate %s or %s."
                                % (
                                    (
                                        dfa_state.from_rule,
                                        transition,
                                    ) + tuple(choices)
                                )
                            )
E                           ValueError: Rule foo is ambiguous; given a NAME token, we can't determine if we should evaluate bar or baz.

parso/pgen2/generator.py:339: ValueError

During handling of the above exception, another exception occurred:

grammar = 'foo: bar | baz\nbar: NAME\nbaz: NAME\n'
error_match = 'foo is ambiguous.*given a PythonTokenTypes\\.NAME.*bar or baz'

>   ???
E   AssertionError: Regex pattern 'foo is ambiguous.*given a PythonTokenTypes\\.NAME.*bar or baz' does not match "Rule foo is ambiguous; given a NAME token, we can't determine if we should evaluate bar or baz.".

test/test_pgen2.py:-1: AssertionError
=========================== short test summary info ============================
FAILED test/test_pgen2.py::test_ambiguities[foo: bar | baz\nbar: NAME\nbaz: NAME\n-foo is ambiguous.*given a PythonTokenTypes\\.NAME.*bar or baz]

There were some changes in repr() in latest alpha of Python 3.10.
https://docs.python.org/3.10/whatsnew/changelog.html#python-3-10-0-alpha-7

bpo-40066: Enum: adjust repr() to show only enum and member name (not value,
nor angle brackets) and str() to show only member name. Update and improve
documentation to match.

bpo-40066: Enum’s repr() and str() have changed: repr() is now
EnumClass.MemberName and str() is MemberName. Additionally, stdlib Enum’s whose
contents are available as module attributes, such as RegexFlag.IGNORECASE, have
their repr() as module.name, e.g. re.IGNORECASE.
https://bugs.python.org/issue40066

Patched in copr with https://github.com/davidhalter/parso/pull/186

Will give upstream some time for feedback before backporting.