Merge remote-tracking branch 'origin/dev' into jesserockz-2026-627

# Conflicts:
#	script/git-hooks/post-checkout

tests/script/test_ci_custom.py

@@ -0,0 +1,147 @@
+"""Unit tests for the ESP_LOG-needs-braces lint rule in script/ci-custom.py.
+
+The rule flags an if/else/for/while whose only body is an unbraced ESP_LOG*() call (which becomes an
+empty statement -- and a -Wempty-body warning -- once the log level compiles the macro out). These
+tests pin the comment/string/raw-string masker, the accepted control-statement shapes, and the
+NOLINT escape hatch at both placements a contributor would try.
+"""
+
+import importlib.util
+from pathlib import Path
+import sys
+
+SCRIPT_DIR = (Path(__file__).parent / ".." / ".." / "script").resolve()
+sys.path.insert(0, str(SCRIPT_DIR))
+_spec = importlib.util.spec_from_file_location("ci_custom", SCRIPT_DIR / "ci-custom.py")
+ci_custom = importlib.util.module_from_spec(_spec)
+_spec.loader.exec_module(ci_custom)
+
+mask = ci_custom._mask_cpp_comments_strings
+
+
+def _lint(content: str) -> list:
+    return ci_custom.lint_esp_log_needs_braces("test.cpp", content)
+
+
+# --- masker ---
+
+
+def test_mask_preserves_length_newlines_and_real_parens() -> None:
+    src = 'foo("bar") + baz();\nqux();\n'
+    masked = mask(src)
+    assert len(masked) == len(src)
+    assert masked.count("\n") == src.count("\n")
+    assert masked.count("(") == src.count("(")  # real parens survive for balancing
+
+
+def test_mask_blanks_line_and_block_comments() -> None:
+    assert "ESP_LOGD" not in mask("a; // if (x) ESP_LOGD(t);\n")
+    assert "ESP_LOGD" not in mask("a; /* if (x) ESP_LOGD(t); */ b;\n")
+
+
+def test_mask_blanks_string_literals() -> None:
+    assert "if" not in mask('x = "if (y) ESP_LOGD";\n')
+
+
+def test_mask_handles_raw_string_without_desync() -> None:
+    # A raw string full of quotes/parens must be consumed as one unit; code after it stays intact.
+    src = 's.print(R"(<a href="x">)");\nreturn;\n'
+    masked = mask(src)
+    assert "href" not in masked
+    assert "return;" in masked  # not swallowed by a desynced string scan
+
+
+# --- rule: flags real violations ---
+
+
+def test_flags_unbraced_if_next_line() -> None:
+    assert _lint("if (x)\n  ESP_LOGD(t);\n")
+
+
+def test_flags_unbraced_same_line() -> None:
+    assert _lint("if (x) ESP_LOGW(t);\n")
+
+
+def test_flags_c_style_for() -> None:
+    assert _lint("for (int i = 0; i < n; i++)\n  ESP_LOGD(t, i);\n")
+
+
+def test_flags_range_for_and_else() -> None:
+    assert _lint("for (auto &x : v)\n  ESP_LOGCONFIG(t);\n")
+    assert _lint("else\n  ESP_LOGE(t);\n")
+
+
+def test_flags_for_header_with_nested_call() -> None:
+    assert _lint("for (auto it = v.begin(); it != v.end(); ++it)\n  ESP_LOGD(t);\n")
+
+
+def test_for_header_does_not_reach_into_a_later_statement() -> None:
+    # The 'for' header is bounded to its own statement, so it cannot swallow the loop body and latch
+    # onto a later ')'. Without that, the '#if' line below is reported as an unbraced body even though
+    # the '#' preprocessor check should skip it.
+    assert not _lint(
+        "for (int i = 0; i < n; i++)\n  arr[i] = 0;\n#if defined(USE_X)\n  ESP_LOGD(t);\n#endif\n"
+    )
+
+
+def test_violation_after_a_for_loop_is_reported_at_its_own_line() -> None:
+    errors = _lint(
+        "for (int i = 0; i < n; i++)\n  sum += a[i];\nif (verbose)\n  ESP_LOGD(t, sum);\n"
+    )
+    lines = [line for line, _col, _msg in errors]
+    assert lines == [3]  # the 'if', not the 'for' on line 1
+
+
+def test_flags_lowercase_esph_log_family() -> None:
+    # core/log.h defines esph_log_*() alongside ESP_LOG*(); both expand to nothing below their level.
+    assert _lint('if (x)\n  esph_log_config(t, "m");\n')
+    assert _lint('if (err != ESP_OK)\n  esph_log_e(t, "m");\n')
+
+
+def test_digit_separator_does_not_disable_the_rest_of_the_file() -> None:
+    # A "'" digit separator must not be read as a char-literal opener, which blanked everything after.
+    assert _lint("uint32_t x = 1'000;\nif (y)\n  ESP_LOGD(t);\n")
+
+
+def test_mask_still_blanks_real_char_literals() -> None:
+    assert "ESP_LOGD" not in mask("char c = '\"'; // if (x) ESP_LOGD(t);\n")
+    assert not _lint("char sep = ';';\nif (x) {\n  ESP_LOGD(t);\n}\n")
+
+
+def test_flags_multiline_log_body() -> None:
+    assert _lint('if (x)\n  ESP_LOGD(t, "%d %d",\n           a, b);\n')
+
+
+def test_raw_string_before_violation_still_caught() -> None:
+    # Regression for the masker desyncing on a raw string and disabling the check for the rest.
+    assert _lint('s.print(R"(<a href="x">)");\nif (y)\n  ESP_LOGD(t);\n')
+
+
+# --- rule: ignores non-violations ---
+
+
+def test_ignores_braced_body() -> None:
+    assert not _lint("if (x) {\n  ESP_LOGD(t);\n}\n")
+
+
+def test_ignores_commented_out_code() -> None:
+    assert not _lint("// if (x) ESP_LOGD(t);\n")
+
+
+def test_ignores_preprocessor_else() -> None:
+    assert not _lint("#else\n  ESP_LOGCONFIG(t);\n#endif\n")
+
+
+def test_ignores_non_log_body() -> None:
+    assert not _lint("if (x)\n  return false;\n")
+
+
+# --- NOLINT escape hatch, both placements ---
+
+
+def test_nolint_at_end_of_log_line_suppresses() -> None:
+    assert not _lint("if (x)\n  ESP_LOGD(t);  // NOLINT\n")
+
+
+def test_nolint_on_control_line_suppresses() -> None:
+    assert not _lint("if (x)  // NOLINT\n  ESP_LOGD(t);\n")

tests/script/test_clang_tidy_hash.py

@@ -81,3 +81,40 @@ def test_read_file_bytes(tmp_path: Path) -> None:
     result = clang_tidy_hash.read_file_bytes(test_file)
 
     assert result == test_content
+
+
+def test_calculate_idedata_cache_hash_changes_with_infra_code(tmp_path: Path) -> None:
+    _populate(tmp_path)
+    infra = tmp_path / "esphome" / "espidf" / "clang_tidy.py"
+    infra.parent.mkdir(parents=True)
+    infra.write_text("a")
+    before = clang_tidy_hash.calculate_idedata_cache_hash(repo_root=tmp_path)
+    assert before == clang_tidy_hash.calculate_idedata_cache_hash(repo_root=tmp_path)
+    infra.write_text("b")
+    assert clang_tidy_hash.calculate_idedata_cache_hash(repo_root=tmp_path) != before
+
+
+def test_calculate_idedata_cache_hash_includes_listed_files(tmp_path: Path) -> None:
+    _populate(tmp_path)
+    before = clang_tidy_hash.calculate_idedata_cache_hash(repo_root=tmp_path)
+    listed = tmp_path / "esphome" / "platformio" / "library.py"
+    listed.parent.mkdir(parents=True)
+    listed.write_text("x")
+    assert clang_tidy_hash.calculate_idedata_cache_hash(repo_root=tmp_path) != before
+
+
+def test_idedata_cache_hash_only_widens_for_esp32(tmp_path: Path) -> None:
+    _populate(tmp_path)
+    infra = tmp_path / "esphome" / "espidf" / "clang_tidy.py"
+    infra.parent.mkdir(parents=True)
+    infra.write_text("a")
+    esp32_before = clang_tidy_hash.idedata_cache_hash("esp32-idf-tidy", tmp_path)
+    other_before = clang_tidy_hash.idedata_cache_hash("esp8266-arduino-tidy", tmp_path)
+    infra.write_text("b")
+    assert (
+        clang_tidy_hash.idedata_cache_hash("esp32-idf-tidy", tmp_path) != esp32_before
+    )
+    assert (
+        clang_tidy_hash.idedata_cache_hash("esp8266-arduino-tidy", tmp_path)
+        == other_before
+    )

tests/script/test_determine_jobs.py

@@ -151,9 +151,14 @@ def test_main_all_tests_should_run(
         patch.object(determine_jobs, "_is_clang_tidy_full_scan", return_value=False),
         patch.object(
             determine_jobs,
-            "_all_integration_test_files",
+            "all_integration_test_files",
             return_value=fake_test_files,
         ),
+        patch.object(
+            determine_jobs,
+            "load_integration_durations",
+            return_value=dict.fromkeys(fake_test_files, 200.0),
+        ),
         patch.object(
             determine_jobs,
             "get_changed_components",
@@ -189,24 +194,12 @@ def test_main_all_tests_should_run(
     output = json.loads(captured.out)
 
     assert output["integration_tests"] is True
-    # run_all=True expands to the full glob and pre-buckets into 3 parts.
-    # Each bucket's `tests` is a JSON list of file paths.
+    assert output["integration_run_all"] is True
+    # run_all=True expands to the full glob; balance and naming are pinned
+    # by the unit tests, main() only needs to round-trip the structure
     assert isinstance(output["integration_test_buckets"], list)
-    assert len(output["integration_test_buckets"]) == 3
-    assert [b["name"] for b in output["integration_test_buckets"]] == [
-        "1/3",
-        "2/3",
-        "3/3",
-    ]
-    for bucket in output["integration_test_buckets"]:
-        assert isinstance(bucket["tests"], list)
-        for path in bucket["tests"]:
-            assert isinstance(path, str)
     bucket_files = [f for b in output["integration_test_buckets"] for f in b["tests"]]
-    assert bucket_files == fake_test_files
-    # Bucket sizes are balanced (max-min difference at most 1).
-    sizes = [len(b["tests"]) for b in output["integration_test_buckets"]]
-    assert max(sizes) - min(sizes) <= 1
+    assert sorted(bucket_files) == fake_test_files
     assert output["clang_tidy"] is True
     assert output["clang_tidy_mode"] in ["nosplit", "split"]
     assert output["clang_format"] is True
@@ -509,14 +502,24 @@ def test_compute_integration_test_buckets_at_threshold_stays_single() -> None:
 
 
 def test_compute_integration_test_buckets_just_over_threshold_splits() -> None:
-    """One file over the threshold triggers the 3-bucket fan-out, balanced."""
+    """One file over the threshold fans out fully when the weights demand it."""
     n = determine_jobs.INTEGRATION_TESTS_SPLIT_THRESHOLD + 1
     files = [f"tests/integration/test_{i:02d}.py" for i in range(n)]
-    run, buckets = determine_jobs._compute_integration_test_buckets(False, files)
+    with patch.object(
+        determine_jobs,
+        "load_integration_durations",
+        return_value=dict.fromkeys(files, 200.0),
+    ):
+        run, buckets = determine_jobs._compute_integration_test_buckets(False, files)
     assert run is True
-    assert [b["name"] for b in buckets] == ["1/3", "2/3", "3/3"]
-    union = [path for b in buckets for path in b["tests"]]
+    # threshold+1 files x 200s caps at the maximum bucket count.
+    n_buckets = determine_jobs.INTEGRATION_TESTS_SPLIT_BUCKETS
+    assert [b["name"] for b in buckets] == [
+        f"{i + 1}/{n_buckets}" for i in range(n_buckets)
+    ]
+    union = sorted(path for b in buckets for path in b["tests"])
     assert union == sorted(files)
+    # Equal weights => bucket sizes are balanced (difference at most 1).
     sizes = [len(b["tests"]) for b in buckets]
     assert max(sizes) - min(sizes) <= 1
 
@@ -526,7 +529,7 @@ def test_compute_integration_test_buckets_run_all_with_empty_glob_disables_run()
 ):
     """run_all=True but glob returns no files => run suppressed (otherwise
     pytest would collect tests outside tests/integration/)."""
-    with patch.object(determine_jobs, "_all_integration_test_files", return_value=[]):
+    with patch.object(determine_jobs, "all_integration_test_files", return_value=[]):
         run, buckets = determine_jobs._compute_integration_test_buckets(True, [])
     assert run is False
     assert buckets == []
@@ -552,6 +555,13 @@ def test_determine_integration_tests(
         assert run_all is True
         assert test_files == []
 
+    # Dependency pins and the session init fixture trigger run_all
+    for trigger in sorted(determine_jobs.INTEGRATION_TESTS_TRIGGER_FILES):
+        with patch.object(determine_jobs, "changed_files", return_value=[trigger]):
+            run_all, test_files = determine_jobs.determine_integration_tests()
+            assert run_all is True
+            assert test_files == []
+
     # Python files directly in esphome/ do NOT trigger tests
     with patch.object(
         determine_jobs, "changed_files", return_value=["esphome/config.py"]
@@ -3139,3 +3149,86 @@ def test_memory_impact_elf_layouts_are_found(tmp_path: Path) -> None:
         elf.write_text("")
 
         assert find_elf_path(build_path) == elf, f"{platform} ELF not found"
+
+
+def test_compute_integration_test_buckets_no_durations_full_fanout() -> None:
+    """Without recorded durations the fan-out stays at the maximum."""
+    files = [f"tests/integration/test_{i:03d}.py" for i in range(15)]
+    with patch.object(determine_jobs, "load_integration_durations", return_value={}):
+        run, buckets = determine_jobs._compute_integration_test_buckets(False, files)
+    assert run is True
+    assert len(buckets) == determine_jobs.INTEGRATION_TESTS_SPLIT_BUCKETS
+    assert sorted(f for b in buckets for f in b["tests"]) == files
+
+
+def test_compute_integration_test_buckets_adaptive_count() -> None:
+    """A small recorded total weight collapses to one bucket above the threshold."""
+    files = [f"tests/integration/test_{i:03d}.py" for i in range(15)]
+    with patch.object(
+        determine_jobs,
+        "load_integration_durations",
+        return_value=dict.fromkeys(files, 10.0),
+    ):
+        run, buckets = determine_jobs._compute_integration_test_buckets(False, files)
+    assert run is True
+    # 15 files x 10s recorded = 150s, under the per-bucket weight target.
+    assert [b["name"] for b in buckets] == ["1/1"]
+    assert buckets[0]["tests"] == files
+
+
+def test_compute_integration_test_buckets_duration_weighted() -> None:
+    """Heavy files spread across buckets instead of clustering by sorted name."""
+    files = [f"tests/integration/test_{i:03d}.py" for i in range(12)]
+    durations = dict.fromkeys(files, 10.0)
+    durations[files[0]] = 600.0
+    durations[files[1]] = 600.0
+    with patch.object(
+        determine_jobs, "load_integration_durations", return_value=durations
+    ):
+        run, buckets = determine_jobs._compute_integration_test_buckets(False, files)
+    assert run is True
+    assert len(buckets) >= 2
+    heavy_buckets = [b for b in buckets if set(files[:2]) & set(b["tests"])]
+    assert len(heavy_buckets) == 2, "heavy files should land in different buckets"
+    assert sorted(f for b in buckets for f in b["tests"]) == files
+
+
+def test_load_integration_durations_missing_or_corrupt(tmp_path: Path) -> None:
+    """Missing or unparsable durations data degrades to an empty mapping."""
+    with patch.object(helpers, "root_path", str(tmp_path)):
+        assert determine_jobs.load_integration_durations() == {}
+        durations_file = tmp_path / helpers.INTEGRATION_TEST_DURATIONS_FILE
+        durations_file.parent.mkdir(parents=True)
+        durations_file.write_text("not json")
+        assert determine_jobs.load_integration_durations() == {}
+        durations_file.write_text('{"tests/integration/test_a.py": 12.5}')
+        assert determine_jobs.load_integration_durations() == {
+            "tests/integration/test_a.py": 12.5
+        }
+        # Non-positive entries are dropped, valid ones survive
+        durations_file.write_text(
+            '{"tests/integration/test_a.py": 12.5, "tests/integration/test_b.py": -1}'
+        )
+        assert determine_jobs.load_integration_durations() == {
+            "tests/integration/test_a.py": 12.5
+        }
+        # One non-numeric entry cannot discard the whole recording
+        durations_file.write_text(
+            '{"tests/integration/test_a.py": 12.5, "tests/integration/test_b.py": null}'
+        )
+        assert determine_jobs.load_integration_durations() == {
+            "tests/integration/test_a.py": 12.5
+        }
+        # A non-dict top level degrades to empty
+        durations_file.write_text("[12.5]")
+        assert determine_jobs.load_integration_durations() == {}
+
+
+def test_committed_integration_durations_are_sane() -> None:
+    """The committed recording itself holds positive bounded floats."""
+    raw = json.loads(
+        (Path(helpers.root_path) / helpers.INTEGRATION_TEST_DURATIONS_FILE).read_text()
+    )
+    assert raw, "committed durations file missing or empty"
+    assert all(isinstance(v, (int, float)) and 0 < v < 86400 for v in raw.values())
+    assert all(k.startswith("tests/integration/test_") for k in raw)

tests/script/test_helpers.py

@@ -2120,3 +2120,30 @@ def test_get_cpp_changed_components_independent_of_cwd(
     assert helpers.get_cpp_changed_components(
         ["tests/components/time/__init__.py"]
     ) == ["time"]
+
+
+def test_lpt_partition_balances_skewed_weights() -> None:
+    """Heavy items spread across groups instead of clustering."""
+    items = [f"i{n}" for n in range(6)]
+    weights = {"i0": 100.0, "i1": 90.0, "i2": 10.0, "i3": 10.0, "i4": 5.0, "i5": 5.0}
+    groups = helpers.lpt_partition(items, weights, 2)
+    group_weights = sorted(sum(weights[i] for i in g) for g in groups)
+    # Contiguous split would give 200 vs 20; LPT lands at 110 vs 110
+    assert group_weights == [110.0, 110.0]
+    assert sorted(i for g in groups for i in g) == items
+
+
+def test_lpt_partition_more_groups_than_items() -> None:
+    """Surplus groups come back empty; every item still lands somewhere."""
+    items = ["a", "b"]
+    groups = helpers.lpt_partition(items, {"a": 1.0, "b": 1.0}, 4)
+    assert len(groups) == 4
+    assert sorted(i for g in groups for i in g) == items
+    assert sum(not g for g in groups) == 2
+
+
+def test_lpt_partition_tie_determinism() -> None:
+    """Equal weights assign in input order, so output is reproducible."""
+    items = [f"i{n}" for n in range(4)]
+    weights = dict.fromkeys(items, 1.0)
+    assert helpers.lpt_partition(items, weights, 2) == [["i0", "i2"], ["i1", "i3"]]

tests/script/test_platformio_install_deps.py

@@ -35,8 +35,8 @@ def _load_script():
 def test_spec_key_collapses_destinations() -> None:
     """Two specs delivering one package share a directory and one key."""
     mod = _load_script()
-    assert mod.spec_key("esphome/noise-c @ 0.1.21") == "noise-c"
-    assert mod.spec_key("esphome/noise-c@0.1.21") == "noise-c"
+    assert mod.spec_key("esphome/noise-c @ 0.1.24") == "noise-c"
+    assert mod.spec_key("esphome/noise-c@0.1.24") == "noise-c"
     assert mod.spec_key("ESP32Async/AsyncTCP @ ^3.4.10") == mod.spec_key(
         "esp32async/asynctcp @ 3.5.0"
     )
@@ -54,23 +54,23 @@ def test_parse_specs_and_cli_args(tmp_path: Path) -> None:
         "[env:a]\n"
         "platform = fake/platform@1\n"
         "lib_deps =\n"
-        "    esphome/noise-c @ 0.1.21\n"
+        "    esphome/noise-c @ 0.1.24\n"
         "    ${common.lib_deps}\n"
         "    internal_lib\n"
         "[env:b]\n"
         "lib_deps =\n"
-        "    esphome/noise-c @ 0.1.21\n"
+        "    esphome/noise-c @ 0.1.24\n"
     )
     mod = _load_script()
     args = Namespace(libraries=True, platforms=True, tools=False)
     libs, platforms, tools = mod.parse_specs(str(ini), args)
     # exact-string duplicates collapse; distinct version pins survive
-    assert libs == ["esphome/noise-c @ 0.1.21"]
+    assert libs == ["esphome/noise-c @ 0.1.24"]
     assert platforms == ["fake/platform@1"]
     assert tools == []
     assert mod.build_cli_args(libs, platforms, tools) == [
         "-l",
-        "esphome/noise-c @ 0.1.21",
+        "esphome/noise-c @ 0.1.24",
         "-p",
         "fake/platform@1",
     ]
@@ -162,13 +162,13 @@ def test_parallel_install_behavior(tmp_path: Path) -> None:
     mod.parallel_install(
         cls,
         [
-            "esphome/noise-c @ 0.1.21",
-            "esphome/noise-c @ 0.1.21",
+            "esphome/noise-c @ 0.1.24",
+            "esphome/noise-c @ 0.1.24",
             "esphome/already @ 1.0",
             "https://x/framework.tar.xz",
         ],
     )
-    assert cls.calls == ["esphome/noise-c @ 0.1.21"]
+    assert cls.calls == ["esphome/noise-c @ 0.1.24"]
     assert cls.lock_events == ["lock", "unlock"]
 
 
@@ -205,7 +205,7 @@ def test_parallel_install_runs_dependency_waves(tmp_path: Path) -> None:
     mod = _load_script()
     cls = _reset_fake(str(tmp_path))
     cls.deps = {
-        "esphome/noise-c @ 0.1.21": [
+        "esphome/noise-c @ 0.1.24": [
             {"owner": "esphome", "name": "libsodium", "version": "^1.0"},
             {"name": "SPI"},
         ],
@@ -213,12 +213,12 @@ def test_parallel_install_runs_dependency_waves(tmp_path: Path) -> None:
             {"owner": "esphome", "name": "libsodium", "version": "^1.0"},
         ],
     }
-    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.21", "esphome/wg @ 1.0"])
+    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.24", "esphome/wg @ 1.0"])
     assert len(cls.calls) == 3  # the shared dep installs exactly once
     assert {mod.spec_key(c) for c in cls.calls} == {"noise-c", "wg", "libsodium"}
     # Wave-1 strings carry no compatibility; the dependency wave does
     compats = dict(cls.compat_calls)
-    assert compats["esphome/noise-c @ 0.1.21"] is None
+    assert compats["esphome/noise-c @ 0.1.24"] is None
     dep_compat = next(v for k, v in cls.compat_calls if "libsodium" in k)
     assert dep_compat is not None  # mirrors pio's install_dependency
 
@@ -229,11 +229,11 @@ def test_dependency_wave_excludes_url_specs(tmp_path: Path) -> None:
     mod = _load_script()
     cls = _reset_fake(str(tmp_path))
     cls.deps = {
-        "esphome/noise-c @ 0.1.21": [
+        "esphome/noise-c @ 0.1.24": [
             {"name": "vendored", "version": "https://github.com/x/y.git"},
         ],
     }
-    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.21"])
+    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.24"])
     assert {mod.spec_key(c) for c in cls.calls} == {"noise-c"}
 
 
@@ -348,13 +348,13 @@ def test_warm_store_still_walks_dependencies(tmp_path: Path) -> None:
     """Already-installed top-level packages still feed the dependency
     wave; a warm store can be missing a transitive dep."""
     mod = _load_script()
-    cls = _reset_fake(str(tmp_path), installed={"esphome/noise-c @ 0.1.21"})
+    cls = _reset_fake(str(tmp_path), installed={"esphome/noise-c @ 0.1.24"})
     cls.deps = {
-        "esphome/noise-c @ 0.1.21": [
+        "esphome/noise-c @ 0.1.24": [
             {"owner": "esphome", "name": "libsodium", "version": "^1.0"},
         ],
     }
-    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.21"])
+    mod.parallel_install(cls, ["esphome/noise-c @ 0.1.24"])
     assert [mod.spec_key(c) for c in cls.calls] == ["libsodium"]
 
 

tests/script/test_sync_dependency_versions.py

@@ -0,0 +1,219 @@
+"""Unit tests for script/sync_dependency_versions.py."""
+
+from pathlib import Path
+import subprocess
+import sys
+
+import pytest
+import yamlrocks
+
+sys.path.insert(0, str((Path(__file__).parent / ".." / ".." / "script").resolve()))
+
+import sync_dependency_versions as sync_mod  # noqa: E402
+
+PRECOMMIT = """\
+# See https://pre-commit.com for more information
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    # Ruff version.
+    rev: v0.1.0
+    hooks:
+      - id: ruff
+  - repo: https://github.com/PyCQA/flake8
+    rev: 7.0.0
+    hooks:
+      - id: flake8
+  - repo: https://github.com/asottile/pyupgrade
+    rev: v3.0.0
+    hooks:
+      - id: pyupgrade
+  - repo: https://github.com/pre-commit/mirrors-clang-format
+    rev: v13.0.1
+    hooks:
+      - id: clang-format
+  - repo: https://github.com/adrienverge/yamllint.git
+    rev: v1.0.0
+    hooks:
+      - id: yamllint
+  - repo: local
+    hooks:
+      - id: pylint
+"""
+
+REQ_TEST = """\
+pylint==4.0.8
+flake8==7.1.0
+ruff==0.2.0  # comment
+pyupgrade==3.0.0
+"""
+
+REQ_DEV = """\
+clang-format==13.0.1
+yamllint==1.0.0
+"""
+
+RUFF_REPO = "https://github.com/astral-sh/ruff-pre-commit"
+DUPLICATE_RUFF_BLOCK = f"  - repo: {RUFF_REPO}\n    rev: v0.3.0\n    hooks: []\n"
+
+EXPECTED_DRIFT = ["ruff: 0.1.0 -> 0.2.0", "flake8: 7.0.0 -> 7.1.0"]
+EXPECTED_PRECOMMIT = PRECOMMIT.replace("rev: v0.1.0", "rev: v0.2.0").replace(
+    "rev: 7.0.0", "rev: 7.1.0"
+)
+
+
+@pytest.fixture
+def root(tmp_path: Path) -> Path:
+    """A fake checkout where ruff (v-prefixed) and flake8 (bare) have drifted."""
+    (tmp_path / ".pre-commit-config.yaml").write_text(PRECOMMIT)
+    (tmp_path / "requirements_test.txt").write_text(REQ_TEST)
+    (tmp_path / "requirements_dev.txt").write_text(REQ_DEV)
+    return tmp_path
+
+
+def _load(text: str) -> object:
+    return yamlrocks.loads(text.encode(), option=yamlrocks.OPT_ROUND_TRIP)
+
+
+@pytest.mark.parametrize(
+    ("requirements", "expected"),
+    [
+        ("prek==0.5.1  # comment\n", "0.5.1"),
+        ("Prek==0.5.1\n", "0.5.1"),
+        ("other==1.0\nprek==0.5.1\n", "0.5.1"),
+        ("prek>=0.5.1\n", None),
+        ("prek-extra==0.5.1\n", None),
+        ("", None),
+    ],
+)
+def test_read_requirement_version(requirements: str, expected: str | None) -> None:
+    assert sync_mod.read_requirement_version(requirements, "prek") == expected
+
+
+def test_find_repo_entry() -> None:
+    entry = sync_mod.find_repo_entry(_load(PRECOMMIT), RUFF_REPO)
+    assert entry["rev"] == "v0.1.0"
+
+
+@pytest.mark.parametrize(
+    ("text", "message"),
+    [
+        ("hooks: []\n", "missing key 'repos'"),
+        ("repos:\n  - rev: 1.0.0\n", "missing key 'repo'"),
+        (PRECOMMIT + DUPLICATE_RUFF_BLOCK, "found 2"),
+        ("repos:\n  - repo: other\n    rev: 1.0.0\n", "found 0"),
+    ],
+)
+def test_find_repo_entry_errors(text: str, message: str) -> None:
+    with pytest.raises(sync_mod.SyncError, match=message):
+        sync_mod.find_repo_entry(_load(text), RUFF_REPO)
+
+
+@pytest.mark.parametrize(
+    ("rev", "expected"),
+    [("v0.1.0", ("v", "0.1.0")), ("7.0.0", ("", "7.0.0")), ("'1.0'", ("", "1.0"))],
+)
+def test_current_rev(rev: str, expected: tuple[str, str]) -> None:
+    doc = _load(f"repos:\n  - repo: {RUFF_REPO}\n    rev: {rev}\n")
+    assert sync_mod.current_rev(doc["repos"][0], RUFF_REPO) == expected
+
+
+@pytest.mark.parametrize(
+    ("block", "message"),
+    [("    hooks: []\n", "has no rev"), ("    rev: 1.0\n", "not a string: 1.0")],
+)
+def test_current_rev_errors(block: str, message: str) -> None:
+    doc = _load(f"repos:\n  - repo: {RUFF_REPO}\n{block}")
+    with pytest.raises(sync_mod.SyncError, match=message):
+        sync_mod.current_rev(doc["repos"][0], RUFF_REPO)
+
+
+def test_sync_reports_without_writing(root: Path) -> None:
+    assert sync_mod.sync(root, write=False) == EXPECTED_DRIFT
+    assert (root / ".pre-commit-config.yaml").read_text() == PRECOMMIT
+
+
+def test_sync_writes_keeps_layout_and_is_idempotent(root: Path) -> None:
+    assert sync_mod.sync(root, write=True) == EXPECTED_DRIFT
+    assert (root / ".pre-commit-config.yaml").read_text() == EXPECTED_PRECOMMIT
+    assert sync_mod.sync(root, write=True) == []
+
+
+def test_sync_does_not_touch_a_config_that_matches(root: Path) -> None:
+    (root / ".pre-commit-config.yaml").write_text(EXPECTED_PRECOMMIT)
+    before = (root / ".pre-commit-config.yaml").stat().st_mtime_ns
+    assert sync_mod.sync(root, write=True) == []
+    assert (root / ".pre-commit-config.yaml").stat().st_mtime_ns == before
+
+
+def test_sync_missing_requirement_pin(root: Path) -> None:
+    (root / "requirements_dev.txt").write_text("")
+    with pytest.raises(sync_mod.SyncError, match="no 'clang-format==' pin"):
+        sync_mod.sync(root, write=True)
+
+
+def test_sync_propagates_config_errors(root: Path) -> None:
+    (root / ".pre-commit-config.yaml").write_text(PRECOMMIT + DUPLICATE_RUFF_BLOCK)
+    with pytest.raises(sync_mod.SyncError, match="found 2"):
+        sync_mod.sync(root, write=True)
+
+
+def test_main_check_reports_drift(
+    root: Path, capsys: pytest.CaptureFixture[str]
+) -> None:
+    assert sync_mod.main(["--check", "--root", str(root)]) == 1
+    assert capsys.readouterr().out.splitlines() == EXPECTED_DRIFT
+    assert (root / ".pre-commit-config.yaml").read_text() == PRECOMMIT
+
+
+def test_main_writes_then_check_is_clean(
+    root: Path, capsys: pytest.CaptureFixture[str]
+) -> None:
+    assert sync_mod.main(["--root", str(root)]) == 0
+    assert capsys.readouterr().out.splitlines() == EXPECTED_DRIFT
+    assert sync_mod.main(["--check", "--root", str(root)]) == 0
+    assert capsys.readouterr().out == ""
+
+
+def test_main_reports_sync_error(
+    root: Path, capsys: pytest.CaptureFixture[str]
+) -> None:
+    (root / "requirements_dev.txt").write_text("")
+    assert sync_mod.main(["--root", str(root)]) == 1
+    assert (
+        "error: requirements_dev.txt: no 'clang-format==' pin"
+        in capsys.readouterr().err
+    )
+
+
+def test_main_defaults_to_repo_root(monkeypatch: pytest.MonkeyPatch) -> None:
+    seen: dict[str, object] = {}
+
+    def fake_sync(root: Path, *, write: bool) -> list[str]:
+        seen["root"] = root
+        seen["write"] = write
+        return []
+
+    monkeypatch.setattr(sync_mod, "sync", fake_sync)
+    assert sync_mod.main([]) == 0
+    assert seen == {"root": sync_mod.REPO_ROOT, "write": True}
+
+
+def test_repository_is_in_sync() -> None:
+    """The real checkout must match; a failure here means a rev has drifted.
+
+    Also proves every SYNC_TARGETS entry still resolves in the real files.
+    """
+    assert sync_mod.sync(sync_mod.REPO_ROOT, write=False) == []
+
+
+def test_cli_entry_point(root: Path) -> None:
+    """Run the script the way the workflow does, as a subprocess."""
+    script = Path(sync_mod.__file__)
+    result = subprocess.run(
+        [sys.executable, str(script), "--check", "--root", str(root)],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+    assert result.returncode == 1
+    assert result.stdout.splitlines() == EXPECTED_DRIFT

tests/script/test_update_integration_test_durations.py

@@ -0,0 +1,130 @@
+"""Unit tests for script/update_integration_test_durations.py."""
+
+import json
+from pathlib import Path
+import sys
+from unittest.mock import patch
+
+import pytest
+
+# Add the script directory to Python path so we can import the module
+script_dir = str((Path(__file__).parent / ".." / ".." / "script").resolve())
+sys.path.insert(0, script_dir)
+
+import helpers  # noqa: E402
+import update_integration_test_durations as uitd  # noqa: E402
+
+JUNIT_TEMPLATE = """<?xml version="1.0" encoding="utf-8"?>
+<testsuites><testsuite>{testcases}</testsuite></testsuites>
+"""
+
+KNOWN = {
+    "tests/integration/test_a.py",
+    "tests/integration/test_b.py",
+}
+
+
+def _write_junit(path: Path, testcases: str) -> None:
+    path.write_text(JUNIT_TEMPLATE.format(testcases=testcases), encoding="utf-8")
+
+
+def test_collect_durations_sums_per_file(tmp_path: Path) -> None:
+    """Testcases from the same module sum."""
+    _write_junit(
+        tmp_path / "a.xml",
+        '<testcase classname="tests.integration.test_a" name="t1" time="1.5"/>'
+        '<testcase classname="tests.integration.test_a" name="t2" time="2.0"/>'
+        '<testcase classname="tests.integration.test_b" name="t1" time="4.0"/>',
+    )
+    assert uitd.collect_durations(tmp_path, KNOWN) == {
+        "tests/integration/test_a.py": 3.5,
+        "tests/integration/test_b.py": 4.0,
+    }
+
+
+def test_collect_durations_class_based_testcase(tmp_path: Path) -> None:
+    """A class-based classname still maps to its module file."""
+    _write_junit(
+        tmp_path / "a.xml",
+        '<testcase classname="tests.integration.test_a.TestFoo" name="t" time="2.5"/>',
+    )
+    assert uitd.collect_durations(tmp_path, KNOWN) == {
+        "tests/integration/test_a.py": 2.5
+    }
+
+
+def test_collect_durations_unknown_module_skipped(
+    tmp_path: Path, capsys: pytest.CaptureFixture[str]
+) -> None:
+    """A classname that maps to no known file is skipped with a warning."""
+    _write_junit(
+        tmp_path / "a.xml",
+        '<testcase classname="tests.integration.test_gone" name="t" time="2.5"/>',
+    )
+    assert uitd.collect_durations(tmp_path, KNOWN) == {}
+    assert "test_gone" in capsys.readouterr().err
+
+
+def test_collect_durations_skips_skipped_testcases(tmp_path: Path) -> None:
+    """Skipped testcases do not record a bogus zero duration."""
+    _write_junit(
+        tmp_path / "a.xml",
+        '<testcase classname="tests.integration.test_a" name="t" time="0">'
+        "<skipped/></testcase>",
+    )
+    assert uitd.collect_durations(tmp_path, KNOWN) == {}
+
+
+def test_collect_durations_unexpected_classname_aborts(tmp_path: Path) -> None:
+    """A classname outside tests.integration means the junit layout changed."""
+    _write_junit(
+        tmp_path / "a.xml",
+        '<testcase classname="tests.unit_tests.test_x" name="t" time="9.0"/>',
+    )
+    with pytest.raises(SystemExit):
+        uitd.collect_durations(tmp_path, KNOWN)
+
+
+def test_collect_durations_empty_dir_aborts(tmp_path: Path) -> None:
+    """No junit XML at all is a hard error, not an empty recording."""
+    with pytest.raises(SystemExit):
+        uitd.collect_durations(tmp_path, KNOWN)
+
+
+def test_main_merges_partial_run(tmp_path: Path) -> None:
+    """A partial run merges over the previous data instead of truncating it."""
+    tests_dir = tmp_path / "tests" / "integration"
+    tests_dir.mkdir(parents=True)
+    for name in ("test_a", "test_b", "test_c"):
+        (tests_dir / f"{name}.py").write_text("", encoding="utf-8")
+    durations_file = tmp_path / helpers.INTEGRATION_TEST_DURATIONS_FILE
+    durations_file.write_text(
+        json.dumps(
+            {
+                "tests/integration/test_a.py": 5.0,
+                "tests/integration/test_b.py": 7.0,
+                "tests/integration/test_gone.py": 9.0,
+            }
+        ),
+        encoding="utf-8",
+    )
+    junit_dir = tmp_path / "junit"
+    junit_dir.mkdir()
+    _write_junit(
+        junit_dir / "a.xml",
+        '<testcase classname="tests.integration.test_a" name="t" time="6.0"/>',
+    )
+    with (
+        patch.object(helpers, "root_path", str(tmp_path)),
+        patch.object(uitd, "DURATIONS_FILE", durations_file),
+    ):
+        # 1 of 3 files covered: refused without --allow-partial
+        with patch.object(sys, "argv", ["uitd", str(junit_dir)]):
+            assert uitd.main() == uitd.EXIT_LOW_COVERAGE
+        with patch.object(sys, "argv", ["uitd", str(junit_dir), "--allow-partial"]):
+            assert uitd.main() == 0
+    # test_a updated, test_b kept, deleted test_gone dropped
+    assert json.loads(durations_file.read_text()) == {
+        "tests/integration/test_a.py": 6.0,
+        "tests/integration/test_b.py": 7.0,
+    }