aniruddhaadak80
1007296ca0
--reasoning takes a value (metavar=LEVEL in _parser.py) but was absent from _TOP_LEVEL_VALUE_FLAGS (used by _first_positional_argv) and from _apply_profile_override's value_flags set. Every invocation like "hermes --reasoning high chat ..." therefore misclassified "high" as the first positional, and _plugin_cli_discovery_needed() forced full eager plugin CLI discovery at argparse-setup time - the documented startup cost paid on every use of the reasoning override. Add --reasoning to both sets, and add a parser-derived parity regression test so future drift between the hand-maintained sets and build_top_level_parser() fails CI instead of silently degrading startup (the exact drift class AGENTS.md bans). Fixes #93530 (cherry picked from commit 9280617ab8b308f9a8cf947f1617276a6bc8eb4f)
81 lines
3.2 KiB
Python
81 lines
3.2 KiB
Python
"""Top-level value-flag classification must match the real parser (#93530).
|
|
|
|
``_first_positional_argv`` and ``_apply_profile_override`` in ``main.py``
|
|
historically kept hand-maintained copies of "which top-level options consume
|
|
a value". Those copies drift — ``--reasoning`` was a value-taking top-level
|
|
option absent from both sets, so every ``hermes --reasoning high <cmd> …``
|
|
invocation mis-classified ``high`` as the first positional and forced eager
|
|
plugin CLI discovery (~500-650ms startup tax).
|
|
|
|
Both sites now share ``hermes_cli._parser.top_level_value_flag_sets()``,
|
|
derived from ``build_top_level_parser()`` (mirroring the
|
|
``update_cmd._holder_value_flags`` precedent). These tests pin:
|
|
|
|
1. parity — the helper covers every value-taking option the live parser
|
|
declares, so future drift fails CI instead of silently degrading
|
|
startup;
|
|
2. the ``--reasoning`` misparse regression itself.
|
|
"""
|
|
|
|
import sys
|
|
|
|
import pytest
|
|
|
|
from hermes_cli._parser import build_top_level_parser, top_level_value_flag_sets
|
|
from hermes_cli.main import _first_positional_argv, _plugin_cli_discovery_needed
|
|
|
|
|
|
def _parser_value_flags() -> tuple[set, set]:
|
|
"""(required, optional) option strings of value-taking top-level actions.
|
|
|
|
An action consumes a value when its nargs is not zero-valued: Store /
|
|
Append with nargs=None take exactly one; '?' takes an optional value.
|
|
Boolean flags in this parser expose nargs == 0.
|
|
"""
|
|
parser = build_top_level_parser()[0]
|
|
required: set = set()
|
|
optional: set = set()
|
|
for action in parser._actions:
|
|
if not action.option_strings or action.nargs == 0:
|
|
continue
|
|
target = optional if action.nargs == "?" else required
|
|
target.update(action.option_strings)
|
|
return required, optional
|
|
|
|
|
|
def test_helper_sets_match_top_level_parser():
|
|
required, optional = top_level_value_flag_sets()
|
|
parser_required, parser_optional = _parser_value_flags()
|
|
missing_required = parser_required - required
|
|
missing_optional = parser_optional - optional
|
|
assert not missing_required, (
|
|
f"value flags missing from required set: {sorted(missing_required)} "
|
|
"(they will be misread as positionals and force eager plugin discovery)"
|
|
)
|
|
assert not missing_optional, (
|
|
f"optional-value flags missing: {sorted(missing_optional)}"
|
|
)
|
|
|
|
|
|
def test_reasoning_is_classified_as_value_flag():
|
|
required, optional = top_level_value_flag_sets()
|
|
assert "--reasoning" in (required | optional)
|
|
|
|
|
|
def test_first_positional_skips_reasoning_value(monkeypatch):
|
|
monkeypatch.setattr("sys.argv", ["hermes", "--reasoning", "high", "chat", "hello"])
|
|
assert _first_positional_argv() == "chat"
|
|
assert _plugin_cli_discovery_needed() is False
|
|
|
|
|
|
@pytest.mark.parametrize("argv", [
|
|
["hermes", "--reasoning", "ultra", "--provider", "openai", "-z", "prompt here"],
|
|
["hermes", "-m", "x", "--reasoning=low", "chat"],
|
|
])
|
|
def test_reasoning_forms_do_not_shadow_the_subcommand(monkeypatch, argv):
|
|
monkeypatch.setattr("sys.argv", argv)
|
|
first = _first_positional_argv()
|
|
assert first not in {"high", "ultra", "low"}, (
|
|
f"a reasoning level leaked through as the positional: {first!r}"
|
|
)
|