See also:
- uv:
  - https://github.com/astral-sh/uv/releases/tag/0.9.17
  - https://github.com/astral-sh/uv/blob/0.9.17/CHANGELOG.md

why: Reproduce tmuxp issue #1002 where attach() raises exception if
session is killed while user is attached.
what:
- Add test that simulates session killed during attach-session
- Mark as xfail since attach() currently calls refresh() which fails
- Documents expected behavior: attach() should not raise if session gone

…illed

why: attach-session is a blocking interactive command where session state
can change arbitrarily during attachment, including being killed entirely.
Calling refresh() after such a command is semantically incorrect.
what:
- Remove self.refresh() call from Session.attach()

Fixes: tmux-python/tmuxp#1002

why: The fix in the previous commit resolves the issue, so the test
should now pass and serve as a permanent regression guard.
what:
- Remove xfail marker from test_session_attach_does_not_fail_if_session_killed_during_attach

why: Document the bug fix and behavioral change for the upcoming release.
what:
- Add breaking change note that attach() no longer calls refresh()
- Add bug fix entry for session killed during attachment

…sion killed (#616)

Fixes an issue where `Session.attach()` would raise `TmuxObjectDoesNotExist` when a user:

1. Attaches to a tmux session via `tmuxp load`
2. Works in the session
3. Kills the session (e.g., closes all windows) before detaching
4. Detaches from tmux

### User Experience

After running `tmuxp load`, users would see this traceback printed to their terminal after detaching:

```
Traceback (most recent call last):
  File "~/.local/bin/tmuxp", line 7, in <module>
    sys.exit(cli.cli())
  ...
  File ".../libtmux/session.py", line 332, in attach
    self.refresh()
  File ".../libtmux/neo.py", line 167, in _refresh
    obj = fetch_obj(...)
  File ".../libtmux/neo.py", line 242, in fetch_obj
    raise exc.TmuxObjectDoesNotExist(...)
libtmux.exc.TmuxObjectDoesNotExist: Could not find object
```

### Root Cause

`Session.attach()` called `self.refresh()` after the `attach-session` command returned. Since `attach-session` is a **blocking interactive command**, the session state can change arbitrarily during attachment - including being killed entirely.

The `refresh()` call was semantically incorrect for interactive commands:
- `attach-session` blocks until user detaches
- Session state can change during attachment  
- Refreshing after such a command makes no sense

### Timeline

- **Feb 2024**: 9a5147a - `Session.attach()` was added with the `refresh()` call
- **Nov 2025**: tmuxp tmux-python/tmuxp@fdafdd2b switched from `attach_session()` to `attach()`
- Users started experiencing the bug

### Changes

- Removes `self.refresh()` call from `Session.attach()`
- Adds regression test using NamedTuple + parametrize + test_id pattern