test: report cwd in order-sensitivity runner

tests/README.md

@@ -97,11 +97,41 @@ python3 tests/run_order_report.py --seed 123 -- tests/cli/ -q
 python3 tests/run_order_report.py -- tests/cli/ -q   # generates and prints a seed
 ```
 
-The same seed reproduces the same order, so to reproduce a failing run, re-run
-with the seed it printed (the runner prints an exact reproduction command at
-the top of each run). Failures discovered this way are real isolation bugs:
-fix them in separate scoped PRs - do not silence them with `skip`/`xfail`, and
-do not "fix" them by depending on a particular order.
+The same seed reproduces the same order when the reported working directory,
+pytest target arguments, and test environment are also the same. The runner
+prints all command arguments with shell-safe POSIX quoting and uses the
+invoking Python interpreter.
+
+A generated-seed run starts with output like:
+
+```text
+[order-report] working directory: /path/to/odysseus
+[order-report] shuffling test order with seed 284734921
+[order-report] reproduce from this working directory with the same test environment:
+[order-report] reproduce with: /path/to/odysseus/.venv/bin/python /path/to/odysseus/tests/run_order_report.py --seed 284734921 -- tests/cli/ -q
+```
+
+Run the printed command from the reported working directory to reproduce the
+same fixed-seed order:
+
+```text
+[order-report] working directory: /path/to/odysseus
+[order-report] shuffling test order with seed 284734921
+[order-report] reproduce from this working directory with the same test environment:
+[order-report] reproduce with: /path/to/odysseus/.venv/bin/python /path/to/odysseus/tests/run_order_report.py --seed 284734921 -- tests/cli/ -q
+```
+
+Pytest output remains visible between the report header and footer. A failing
+run ends with pytest's normal failure report followed by:
+
+```text
+FAILED tests/example_test.py::test_example - AssertionError
+[order-report] seed 284734921: pytest exit code 1 (report-only; fix order-sensitive failures in separate scoped PRs)
+```
+
+Failures discovered this way are real isolation bugs: fix them in separate
+scoped PRs - do not silence them with `skip`/`xfail`, and do not "fix" them by
+depending on a particular order.
 
 The runner propagates pytest's exit code, so it composes with normal local
 workflows; "report-only" means it is not a CI gate, not that failures are

tests/run_order_report.py

@@ -16,7 +16,8 @@ Examples:
 
 The shuffle is applied through a local ``pytest_collection_modifyitems`` hook
 passed to ``pytest.main`` as an in-process plugin; no conftest or global
-plugin is involved. The exit code is pytest's own.
+plugin is involved. Reproduction requires the reported working directory,
+seed, pytest arguments, and test environment. The exit code is pytest's own.
 """
 from __future__ import annotations
 
@@ -105,7 +106,12 @@ def print_report_header(seed: int, pytest_args: Sequence[str]) -> None:
         "--",
         *pytest_args,
     ]
+    print(f"[order-report] working directory: {Path.cwd()}")
     print(f"[order-report] shuffling test order with seed {seed}")
+    print(
+        "[order-report] reproduce from this working directory with the same "
+        "test environment:"
+    )
     print(f"[order-report] reproduce with: {shlex.join(repro)}")
 
 

tests/test_run_order_report.py

@@ -125,6 +125,13 @@ def test_explicit_seed_is_printed_with_repro_command(capsys):
     assert f"reproduce with: {repro}" in out
 
 
+def test_working_directory_is_reported(capsys, monkeypatch, tmp_path):
+    monkeypatch.chdir(tmp_path)
+    run(["--seed", "123", "--", "-q"], pytest_main=_FakePytestMain())
+    out = capsys.readouterr().out
+    assert f"[order-report] working directory: {tmp_path}" in out
+
+
 def test_footer_repeats_seed_and_outcome(capsys):
     run(["--seed", "123", "--", "-q"], pytest_main=_FakePytestMain(returncode=1))
     out = capsys.readouterr().out