oinspect._getdef: bare except → except Exception so KeyboardInterrupt
and SystemExit are no longer silently swallowed during signature rendering.

interactiveshell.safe_execfile / safe_execfile_ipy: bare except on file-open
probes → except OSError, which is the only exception that opening a file can
legitimately raise. Prevents masking programmer errors and control-flow
exceptions.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

oinspect.info: obj.__class__ lookup → except AttributeError (only failure mode).
oinspect.info: str(obj) path → except Exception (user __str__ can raise anything).
completer._default_arguments_from_docstring path: bare except → except Exception
  so KeyboardInterrupt from completion is no longer swallowed.
page.page_file: two bare except → except Exception; pager env lookup →
  except KeyError (os.environ lookup can only raise KeyError).

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…ashhandler

test_history: test_hist_file_config didn't close the tempfile or the
HistoryManager's SQLite connection before attempting file deletion, causing
intermittent failures on macOS/Windows where the OS holds file locks.
Fix: close the NamedTemporaryFile immediately after creation, explicitly stop
the save_thread and close the db in the finally block, and call gc.collect()
before unlinking. Closes #15161.

application: os.getcwd() → except OSError (only OS-level errors possible).
crashhandler: open() → except OSError; config pformat → except Exception.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

oinspect: alias obj[1] subscript → except (TypeError, IndexError);
  getattr(obj, '__class__') for class docstring → except AttributeError.

shellapp: all nine startup/init bare except: → except Exception: so that
  KeyboardInterrupt and SystemExit are no longer silently swallowed during
  extension loading and exec_lines execution.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…epts

ptutils: _elide_point and _elide_typed now return the string unchanged when
min_elide <= 0, giving users a documented way to disable filename abbreviations
in tab-completion output entirely (c.TerminalInteractiveShell.min_elide = 0).
Update trait help text. Add tests in terminal/tests/test_ptutils.py.
Closes #12105.

interactiveshell: narrow three safe bare except: clauses:
- push_var_expand eval() → except Exception (can raise anything from user ns)
- SyntaxError.filename assignment (×2) → except AttributeError (only failure mode)
The other 9 bare excepts in interactiveshell.py are intentionally broad
(user-code execution funnels, custom TB handlers) and are left unchanged.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace 10 near-identical test methods across two test classes with
7 parametrized functions. The non-positive min_elide cases (0, -1, -10)
and the short-string boundary cases now run as named parameter sets,
making failures immediately identifiable and adding -10 as a third
boundary value that wasn't tested before.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace catch-all `except:` with specific exception types throughout:

- execution.py: UnicodeEncodeError|LookupError for encoding probes
- osm.py, logging.py: ValueError for int()/str.split() conversions
- logging.py, namespace.py: Exception for general failure paths
- code.py: AttributeError for displayhook.prompt_count access
- basic.py: fix dead code after `raise` in xmode handler (unreachable
  xmode_switch_err call); now properly catches and warns on failure
- script.py: OSError for process signal/terminate/kill calls
- completerlib.py: ImportError for import_module failures
- tbtools.py, ultratb.py, doctb.py: Exception in repr/str safety wrappers
- deepreload.py: Exception for reload paths, KeyError for dict lookup
- dir2.py: Exception in hasattr2 helper

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- autoreload.py: Exception in superreload and post-execute hook paths;
  Exception (re-raise) in module reload dictionary restore
- backgroundjobs.py: Exception in traceback handler lookup and job run()
- path.py: Exception for Windows registry homedir lookup; Exception
  (re-raise) in link_or_copy fallback cleanup

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Final pass of narrowing bare except: clauses:

- ipapp.py: Exception for crash report user-ns access
- custom_doctests.py: Exception for numpy array conversion
- testing/tools.py: OSError for temp file unlink on Windows
- formatters.py: Exception in formatter dispatch (FIXME re-raise)
- pickleshare.py: Exception converting pickle load errors to KeyError
- debugger.py: ValueError for int() parse in list command arg
- debugger_backport.py: Exception in exec-with-restore-context path
- demo.py: Exception in demo block execution handler

Only remaining bare except: are in interactiveshell.py (9 intentional
user-code execution funnels) and a docstring example in ultratb.py.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…tions

Fill in type annotations that were missing from the completer module:

- get__all__entries: obj: Any -> list[str]
- _parse_tokens: c: str -> list[tokenize.TokenInfo]
- python_func_kw_matches: text: str -> list[str]
- latex_name_matcher: -> SimpleMatcherResult
- custom_completer_matcher: context: CompletionContext -> SimpleMatcherResult
- dispatch_custom_completer: text: str -> list[str] | None
- fwd_unicode_matcher: -> SimpleMatcherResult

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

The literal string \"string_form\" contains no tabs, so .expandtabs()
was a no-op. Remove it to clarify intent. The indentation width (11)
is preserved — it aligns with the display label \"String form\" (same
length) used when showing long string representations.

Also add test_pinfo_long_string_form_truncated to cover the truncation
code path (previously untested): objects whose repr exceeds 200 chars
should have <...> inserted in ? mode but not in ?? (detail_level=1).

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

New test file covering the entire display_functions public API:

- _merge(): flat/nested dict merging, non-dict inputs, mutation
- _new_id(): hex validity, uniqueness, length (32 chars = 16 bytes)
- DisplayHandle: auto/explicit ID, repr, uniqueness guarantee
- display(): fallback to print() without InteractiveShell; display_id=True
  returns handle; display_id str preserved; no display_id returns None;
  TypeError on update=True without display_id; no-objects+display_id OK
- @pytest.mark.parametrize covering int/str/list/dict types
- update_display(): verifies update=True and display_id forwarded

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…; add tests

macro.py improvements:
- Add __setstate__ so pickled Macros round-trip correctly (__getstate__
  was defined without the matching __setstate__, breaking unpickling)
- TypeError in __add__ now includes informative message with the
  offending type name instead of bare raise TypeError
- Add type hint on __init__ code: str parameter
- Add return type annotation on __add__

New tests/test_macro.py (17 tests, previously zero coverage):
- Code storage, trailing newline guarantee
- coding: / coding= declaration stripping (both PEP 263 syntaxes)
- __str__, __repr__ formatting
- @parametrize for multi-line code preservation
- __add__ with Macro, str, and invalid type (TypeError message check)
- Pickle round-trip via __getstate__/__setstate__

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

timing.py:
- Replace `assert reps >= 1` with a proper `if reps < 1: raise ValueError`
  since assert statements are silently removed with python -O, which would
  cause a ZeroDivisionError on the av_time calculation instead of the
  expected validation error.

New tests/test_error.py (17 tests, previously untested):
- Exception hierarchy: IPythonCoreError, TryNext, UsageError,
  StdinNotImplementedError, InputRejected
- Multiple inheritance: StdinNotImplementedError catches as both
  IPythonCoreError and NotImplementedError
- InputRejected is NOT a subclass of IPythonCoreError (by design)
- @parametrize across all 5 exception classes for raise/catch

New tests/test_timing.py (14 tests, previously untested):
- clock()/clock2() return non-negative floats
- timings_out() executes exactly reps times, returns last output
- ValueError on reps=0 and reps<0 (now that assert is replaced)
- @parametrize verifying per_call == total/reps identity
- timings()/timing() convenience wrappers
- kwargs forwarding via timings_out

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…ests

builtin_trap.py: Remove self._builtins_added = False from deactivate().
The flag was set but never read anywhere in the codebase — a leftover
from an older implementation. Cleanup logic already uses _orig_builtins.

New tests/test_contexts.py (13 tests, previously untested):
- preserve_keys: restores existing values, deletes keys that didn't exist
  on entry, preserves unrelated keys, multiple keys at once
- Restoration after exception (context managers must clean up)
- @parametrize for falsy values (None, 0, False, '', [], {}) — guards
  against truthiness bugs in restoration logic
- No-keys edge case is a no-op

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…sted)

New tests/test_encoding.py (12 tests):
- get_stream_enc: stream with encoding, without encoding attr, empty/None
  encoding, custom default arg, real sys.stdin
- getdefaultencoding: returns valid encodable string, DeprecationWarning
  on prefer_stream= kwarg, cp0→cp1252 fallback with RuntimeWarning
- DEFAULT_ENCODING: non-empty string that's a valid codec

New tests/test_syspathcontext.py (9 tests):
- prepended_to_syspath: adds at front, removes on exit, returns self
- No-op (and no-removal) when dir already in sys.path
- Removal on exception, exception propagation (returns False)
- Handles concurrent removal (ValueError suppressed on exit)

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

New tests/test_sentinel.py (7 tests):
- repr format is 'module.name'
- Optional docstring set/not-set behavior
- Two Sentinels with same name are distinct objects (identity check)
- name and module attributes accessible

New tests/test_strdispatch.py (12 tests):
- add_s / s_matches: exact match, no-match returns empty
- add_re: pattern match, no match
- flat_matches combining string and regex matches
- Priority ordering (lower number = higher priority = comes first)
- dispatch() generator yields chains, empty for no match
- Multiple values for same key preserved
- Docstring example from the class produces expected [123, 34, 686]

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

New tests/test_display_trap.py (7 tests):
- Context manager sets/restores sys.displayhook correctly
- is_active property true inside context, false outside
- Returns self from __enter__
- Nested traps: outer restored when inner exits
- Exception propagates (returns False from __exit__)
- _nested_level increments/decrements correctly through nesting

New tests/test_payload.py (11 tests):
- write_payload with dict stored; TypeError for non-dict inputs
- single=True updates existing payload with same source (replaces)
- single=False appends regardless of source
- No-source payloads always append
- Different sources coexist
- read_payload returns all, clear_payload empties, initial state empty

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- Corrects comment that mistakenly referenced BuiltinTrap instead of DisplayTrap
- Adds 13 tests for IPython.utils.data (chop and deprecated uniq_stable)
  covering edge cases: empty sequences, uneven splits, string chops,
  deprecation warning emission, and order preservation

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

builtin_trap.py:
- Add from __future__ import annotations
- Annotate __init__, __enter__, __exit__, add_builtin, remove_builtin,
  activate, deactivate with precise types

hooks.py:
- Add from __future__ import annotations
- Annotate CommandChainDispatcher.__init__, __call__, __str__, add

tests/test_hooks.py (8 new tests):
- Empty chain raises TryNext
- __str__ and __iter__ coverage
- Parametrized priority-sort verification (3 cases)
- First successful func stops the chain (later funcs not called)
- Argument forwarding to registered callables

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

py3compat.py:
- Add from __future__ import annotations
- Remove typing.Optional in favor of X | None
- Add type annotations to safe_unicode, input, execfile

logger.py:
- Add from __future__ import annotations
- Replace old-style _get_mode/_set_mode + property() with @property/@Setter
- Add type annotations to all public methods
- Fix logfile attribute type to IO[str] | None

tests/test_logger_unit.py (new, 25 tests):
- logmode property validation (all 5 valid modes, invalid raises ValueError)
- Initial state assertions
- logstart/logstop lifecycle (over, append, error if already active)
- log_write for input/output, suppression when inactive
- switch_log toggling and edge cases
- logstate output when active/inactive

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

capture.py:
- Add from __future__ import annotations
- Replace typing.List/Optional/Type with built-in list/X|None/type[X]
- Add return type to CapturedIO.outputs property

compilerop.py:
- Add from __future__ import annotations
- Add parameter and return types to code_name, ast_parse, reset_compiler_flags,
  compiler_flags, get_code_name, cache, extra_flags
- Fix format_code_name return type: was str (wrong), now tuple[str,str]|None
- Remove unused import time

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace typing.Optional/Type/Tuple with built-in X|None/type[X]/tuple[...]
Add from __future__ import annotations where missing

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace typing.Optional/Tuple/List with built-in X|None/tuple[...]/list[...]
Add from __future__ import annotations to both files

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

doctb.py:
- Add from __future__ import annotations
- Replace Optional[X] with X | None

crashhandler.py:
- Add from __future__ import annotations
- Replace Dict/Optional with dict[...]/X | None
- Move types import with stdlib imports

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace deprecated typing.Optional/List/Dict/Union with built-in equivalents
using from __future__ import annotations.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace typing.List/Optional/Tuple/Type with built-in equivalents
Add from __future__ import annotations to both files

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Add from __future__ import annotations; replace List/Dict/Tuple/Optional/Union
with built-in generics. Use builtins.list in methods of LSString and SList
where 'list = property(...)' shadows the built-in list in annotations.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace deprecated typing.Optional/List/Dict/Tuple/Union with built-in
equivalents using from __future__ import annotations.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Add from __future__ import annotations. Remove Optional from typing imports
and replace Optional[X] with X | None. Replace Union[X, Y] type aliases and
annotations with X | Y syntax. Keep Union import since it is used at runtime
in the origin is Union identity check.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Add from __future__ import annotations. Replace Optional[X] with X | None,
Union[X, Y] with X | Y, Dict/List/Tuple with dict/list/tuple. Remove string
quotes from forward references that are now valid due to deferred evaluation.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Covers skipdoctest.py, displaypub.py (also remove typing as t / use dict),
ptutils.py, paths.py, page.py, tips.py, magics/basic.py, magics/auto.py,
_process_posix.py, and extensions/autoreload.py.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Set self._closed = True before channel validation in Tee.__init__ so that
__del__ does not raise AttributeError if __init__ raises ValueError for an
invalid channel. Convert test_io.py from unittest.TestCase to plain pytest
functions with @pytest.mark.parametrize.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…ytest

Fix buggy assertions in test_full_path_posix/win32 (were checking truthiness
not actual values). Convert test_parser to @pytest.mark.parametrize. Convert
TestAssertPrints from unittest.TestCase to plain pytest functions.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…cess_win32_controller.py

Combine test_code_name and test_code_name2 into parametrized test covering
multiple cell numbers. Add parametrized test for cache with various source
types including Unicode. Add from __future__ import annotations to
_process_win32_controller.py.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace unittest.TestCase setUp with @pytest.fixture for the EventManager,
allowing plain function-based tests that are easier to read and compose.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace unittest.TestCase and manual for-loops with @pytest.mark.parametrize,
turning each pattern/expected pair into a distinct named test case.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…etrize

Replace unittest.TestCase setUp/tearDown with pytest tmp_path + monkeypatch
fixtures, and parametrize the magic_run_completer pattern tests.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- shellapp.py: add from __future__ import annotations, replace `import typing as t` with `from typing import Any`
- test_path.py: replace TestShellGlob and TestLinkOrCopy unittest.TestCase classes with
  pytest fixtures (tmp_path, monkeypatch) and parametrized shellglob tests

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace unittest.TestCase with plain functions; parametrize the four
long-collection tests into a single test_long_collection_repr.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…est.raises

In TestAudioDataWithNumpy, replace all self.assertRaises calls with pytest.raises.
Keep the class inheritance structure intact since TestAudioDataWithoutNumpy
reuses all tests via inheritance.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Convert for-loop test patterns to @pytest.mark.parametrize, turning each
input/expected pair into a distinct named test case. Also unwrap single-item
loops into direct assignments.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Add from __future__ import annotations, remove deprecated List/Union
typing imports, use list[...] and X | Y | Z union syntax.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- test_oinspect: parametrize test_calldef_none over 4 callable objects
- test_inputtransformer2: parametrize test_find_autocalls and test_transform_autocall
  over the three autocall syntax cases

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Convert for-loops in test_select_figure_formats_set and test_no_gui_backends
to @pytest.mark.parametrize for clearer per-case failure reporting.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Remove all unittest.TestCase classes:
- TestAutoSuggest → test_changing_provider()
- TestContextAwareCompletion → parametrized test_adjust_completion_text_based_on_context
- InteractiveShellTestCase → plain functions (update mock_input to not need self)
- TerminalMagicsTestCase → test_paste_magics_blankline()

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace unittest.TestCase setUp/tearDown with a pytest fixture using tmp_path
and monkeypatch.syspath_prepend. Combine the 6 run-submodule test methods into
two parametrized functions test_run_submodule and test_debug_run_submodule.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…and parametrize

- Remove all unittest.TestCase subclasses
- Add @pytest.fixture recursion_setup to replace setUp
- Move class-level code string constants to module level
- Convert conditional MemoryErrorTest to @pytest.mark.skipif
- Add functools.wraps to recursionlimit decorator so pytest can
  resolve fixtures through the wrapper
- Remove duplicate import sys and import unittest

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace TempFileMixin+TestCase with a tmp_pyfile pytest fixture that
creates and cleans up temporary Python/IPython files.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- Replace self.assert* with plain assert statements throughout
- Convert setUp/tearDown to pytest fixtures with autouse=True
- Replace tt.TempFileMixin with local tmp_pyfile fixture helpers
- Remove import unittest; keep class hierarchy for ExitCodeChecks/
  TestSystemRaw/TestSystemPipedExitCode (inherited test methods)
- Fix trailing whitespace in usage.py, dtexample.py, test_ipdoctest.py

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

…e TestCase classes

test_naked_string_cells: The test was failing when run after test_displayhook.py
because test_interactivehooks_ast_modes_semi_suppress leaves a semicolon-terminated
cell as the last entry in input_hist_parsed. When the next run_cell call uses the
default store_history=False, quiet() checks that stale semicolon cell and silences
the display hook, so _ is never updated. Fix: use store_history=True and also pre-
clear _, __ and ___ from user_ns to avoid the separate update_unders=False issue.

Also convert two more test files:
- tests/test_tools.py: replace Test_ipexec_validate(TempFileMixin) with
  ipexec_tmpfile pytest fixture
- tests/test_process.py: replace SubProcessTestCase(TempFileMixin) with
  subprocess_tmpfile pytest fixture; replace self.assert* with plain assert

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Remove unittest.TestCase classes and convert to top-level functions with
pytest fixtures. Use pytest.raises instead of self.assertRaises, plain
assert instead of self.assert*, and autouse fixtures for teardown.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Remove unittest.TestCase classes and convert to top-level functions with
pytest fixtures. Use pytest.raises, plain assert, and caplog instead of
assertLogs for log capture. Flatten TestAudioDataWithoutNumpy inheritance
into separate _no_numpy functions with skipif_not_numpy guards.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Remove unittest.TestCase base from Fixture and TestAutoreload classes.
Convert Fixture setUp/tearDown to an autoreload_fixture pytest fixture
using types.SimpleNamespace. All test methods become top-level functions
accepting autoreload_fixture. Use pytest.raises and plain assert
throughout.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Remove all unittest.TestCase classes. Convert setUp/tearDown to pytest
fixtures. Convert all self.assert* to plain assert / pytest.raises.
Top-level test functions replace class methods throughout both files.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Replace TempFileMixin-based classes with a run_tmpfile pytest fixture.
Convert all test methods to top-level functions. Remove the context
manager usage of TempFileMixin by calling the fixture directly.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

- test_ultratb.py: collapse 4-mode test_nonascii_msg loop into parametrize
- test_magic.py: convert test_extract_symbols for-loop to parametrize
- test_deduperreload.py: collapse 6 identical test_deduperreloader_basic*
  tests (3 pairs of duplicates) into 3 parametrized cases
- test_display.py: fix orphaned test_audio_raises_for_nested_list method
  left inside a function after class conversion

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj

Consolidate 4 test_has_open_quotes* functions into a single parametrized
test. Convert test_all_completions_dups internal jedi_status loop to
parametrize.

https://claude.ai/code/session_01VPnGTgifZ8HMAfiqEVftjj