Resolve the socket path in Server's repr instead of hard-coding /tmp

[tony]

Summary

• Fix repr(Server()) naming a socket the object is not talking to. The fall-through branch hard-coded /tmp/tmux-<euid>/default, but tmux resolves its socket directory from $TMUX_TMPDIR, so under a test harness, a sandbox, a container, or anyone with TMUX_TMPDIR exported the repr was simply wrong.
• Add libtmux._internal.env.resolve_socket_path(), which computes tmux-<euid>/<socket_name or "default"> under $TMUX_TMPDIR, falling back to /tmp.
• Replace the inline copy of that computation in pytest_plugin._reap_test_server, which carried a comment explaining it had to recompute what Server would not tell it. Both callers now share one resolver.
• Remove a getattr(self, 'socket_name', 'default') in the socket_name branch that ran after self.socket_name is not None had already been established.

Display-only: which socket the object talks to is unchanged.

Why the repr matters

A repr is what appears in tracebacks, --showlocals banners, logs, and print(server) — precisely where someone is trying to work out which server they are looking at. Naming the wrong socket there is worse than naming none.

Before / After

Under a TMUX_TMPDIR:

repr(server)       # Server(socket_path=/tmp/tmux-1000/default)
server.cmd("display-message", "-p", "#{socket_path}").stdout[0]
                   # /tmp/repro-JxuS/tmux-1000/default

repr(server)       # Server(socket_path=/tmp/repro-JxuS/tmux-1000/default)
server.cmd("display-message", "-p", "#{socket_path}").stdout[0]
                   # /tmp/repro-JxuS/tmux-1000/default

With TMUX_TMPDIR unset the output is byte-identical to before, which is why the three shipped doctests that assert this repr needed no edit.

Design decisions

$TMUX_TMPDIR or /tmp, and nothing else. TMPDIR is deliberately not consulted, because tmux does not consult it either — man tmux: "the sockets are all created in a directory tmux-UID under the directory given by TMUX_TMPDIR or in /tmp". Confirmed empirically on tmux 3.7b: TMPDIR=<dir> tmux -L probe still lands in /tmp/tmux-1000/probe. Consulting TMPDIR would have broken the existing doctests on any machine that exports it.

Computed, not observed. The helper says where tmux would put the socket, not that a daemon is listening. Its docstring says so and points live callers at the #{socket_path} format instead. Server.__repr__ is the only non-test consumer, and it is display-only.

Lives in _internal. libtmux._internal.env already owns the tmux-environment readers (resolve_env, socket_path_from_env, pane_id_from_env), so this adds no public API surface.

Empty TMUX_TMPDIR is treated as unset. Verified against tmux 3.7b, where TMUX_TMPDIR= tmux -L probe also lands in /tmp.

Scope

Repairing the unreachable socket_path auto-derivation in Server.__init__ is deliberately not in this PR — that is a behaviour change with wider blast radius, tracked separately. This resolver would make it a small change, though.

Test plan

• test_repr_socket_path_honors_tmux_tmpdir — sets TMUX_TMPDIR via monkeypatch and asserts the repr names a path inside it; fails on master
• test_repr_socket_name — covers the getattr removal (behaviour-preserving, so it passes either way)
• Doctests on Server.__repr__ and resolve_socket_path, which are collected and run
• The three shipped doctests asserting the old output still pass unedited: README.md, docs/quickstart.md, docs/api/libtmux.neo.md
• uv run ruff check . / uv run ruff format . / uv run mypy clean
• uv run py.test --reruns 0 clean
• just build-docs clean

Note for the merger

The CHANGES entry cites the issue number; swap it for this PR's number on merge. CHANGES will conflict with the sibling PRs for #724, #725 and #726, which all add to the same block — keep both sides.

[codecov]

Codecov Report

❌ Patch coverage is 30.00000% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 52.28%. Comparing base (be7c8c1) to head (8f437fa).

Files with missing lines	Patch %	Lines
src/libtmux/_internal/env.py	22.22%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #727      +/-   ##
==========================================
- Coverage   52.45%   52.28%   -0.17%     
==========================================
  Files          26       26              
  Lines        3729     3737       +8     
  Branches      747      747              
==========================================
- Hits         1956     1954       -2     
- Misses       1469     1479      +10     
  Partials      304      304              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.