symtest-cli 1.5.5__tar.gz → 1.6.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (419) hide show
  1. {symtest_cli-1.5.5/src/symtest_cli.egg-info → symtest_cli-1.6.1}/PKG-INFO +18 -5
  2. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/README.md +17 -4
  3. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/docs/design.md +55 -6
  4. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/docs/design_en.md +66 -8
  5. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/docs/user_manual.md +179 -8
  6. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/docs/user_manual_en.md +207 -8
  7. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/pyproject.toml +1 -1
  8. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/SKILL.md +38 -0
  9. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/references/user_manual.md +207 -8
  10. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/run.py +15 -0
  11. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/inheritance_expander.py +72 -0
  12. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/prepared.py +6 -1
  13. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/base_runner.py +37 -4
  14. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/execution/executor.py +81 -17
  15. symtest_cli-1.6.1/src/symtest/core/execution/linux_cgroup.py +317 -0
  16. symtest_cli-1.6.1/src/symtest/core/execution/process_group.py +73 -0
  17. symtest_cli-1.6.1/src/symtest/core/execution/result.py +31 -0
  18. symtest_cli-1.6.1/src/symtest/core/execution/windows_job.py +367 -0
  19. symtest_cli-1.6.1/src/symtest/core/history_store.py +181 -0
  20. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/orchestration/sequence.py +23 -0
  21. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/orchestration/single.py +32 -0
  22. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/parallel_runner.py +58 -75
  23. symtest_cli-1.6.1/src/symtest/core/resource_manager.py +370 -0
  24. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/result.py +14 -1
  25. symtest_cli-1.6.1/src/symtest/core/system_memory.py +211 -0
  26. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/types.py +9 -0
  27. symtest_cli-1.6.1/src/symtest/runners/parallel_config_runner.py +631 -0
  28. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/project_config.py +22 -0
  29. {symtest_cli-1.5.5 → symtest_cli-1.6.1/src/symtest_cli.egg-info}/PKG-INFO +18 -5
  30. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest_cli.egg-info/SOURCES.txt +19 -0
  31. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/README.md +13 -0
  32. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_csv_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  33. symtest_cli-1.6.1/tests/integration/parallel/__pycache__/test_parallel_runner.cpython-312-pytest-9.0.3.pyc +0 -0
  34. symtest_cli-1.6.1/tests/integration/parallel/test_parallel_runner.py +750 -0
  35. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_config_preparation.cpython-312-pytest-9.0.3.pyc +0 -0
  36. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_h5_comparator.cpython-312-pytest-9.0.3.pyc +0 -0
  37. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_inheritance_expander.cpython-312-pytest-9.0.3.pyc +0 -0
  38. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/__pycache__/test_find.cpython-312-pytest-9.0.3.pyc +0 -0
  39. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_history_store.cpython-312-pytest-9.0.3.pyc +0 -0
  40. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_linux_cgroup.cpython-312-pytest-9.0.3.pyc +0 -0
  41. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_linux_memory.cpython-312-pytest-9.0.3.pyc +0 -0
  42. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_memory_flow.cpython-312-pytest-9.0.3.pyc +0 -0
  43. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_outputs_cleanup.cpython-312-pytest-9.0.3.pyc +0 -0
  44. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_resource_manager.cpython-312-pytest-9.0.3.pyc +0 -0
  45. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_timing_history.cpython-312-pytest-9.0.3.pyc +0 -0
  46. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_windows_job.cpython-312-pytest-9.0.3.pyc +0 -0
  47. symtest_cli-1.6.1/tests/unit/core/__pycache__/test_windows_memory.cpython-312-pytest-9.0.3.pyc +0 -0
  48. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_history_store.py +70 -0
  49. symtest_cli-1.6.1/tests/unit/core/test_linux_cgroup.py +365 -0
  50. symtest_cli-1.6.1/tests/unit/core/test_linux_memory.py +207 -0
  51. symtest_cli-1.6.1/tests/unit/core/test_memory_flow.py +358 -0
  52. symtest_cli-1.6.1/tests/unit/core/test_resource_manager.py +465 -0
  53. symtest_cli-1.6.1/tests/unit/core/test_timing_history.py +147 -0
  54. symtest_cli-1.6.1/tests/unit/core/test_windows_job.py +290 -0
  55. symtest_cli-1.6.1/tests/unit/core/test_windows_memory.py +55 -0
  56. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_extractor_channels.cpython-312-pytest-9.0.3.pyc +0 -0
  57. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_numeric_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  58. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_numeric_parity.cpython-312-pytest-9.0.3.pyc +0 -0
  59. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_script_extract.cpython-312-pytest-9.0.3.pyc +0 -0
  60. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_inheritance_expander.py +109 -0
  61. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_project_config.cpython-312-pytest-9.0.3.pyc +0 -0
  62. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/test_project_config.py +80 -0
  63. symtest_cli-1.5.5/src/symtest/core/execution/result.py +0 -24
  64. symtest_cli-1.5.5/src/symtest/core/history_store.py +0 -96
  65. symtest_cli-1.5.5/src/symtest/runners/parallel_config_runner.py +0 -355
  66. symtest_cli-1.5.5/tests/integration/parallel/__pycache__/test_parallel_runner.cpython-312-pytest-9.0.3.pyc +0 -0
  67. symtest_cli-1.5.5/tests/integration/parallel/test_parallel_runner.py +0 -208
  68. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/LICENSE +0 -0
  69. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/MANIFEST.in +0 -0
  70. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/setup.cfg +0 -0
  71. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/full_runner_example.py +0 -0
  72. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/extract_channels_script.py +0 -0
  73. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/my_analysis_comparator.py +0 -0
  74. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/my_channel_extractor.py +0 -0
  75. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/test_cases_golden_file.json +0 -0
  76. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/test_cases_simple.json +0 -0
  77. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/test_cases_simple.yaml +0 -0
  78. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/assets/templates/test_cases_steps.json +0 -0
  79. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/skill/references/field_reference.md +0 -0
  80. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/__init__.py +0 -0
  81. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/cli.py +0 -0
  82. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/__init__.py +0 -0
  83. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/compare.py +0 -0
  84. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/find.py +0 -0
  85. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/migrate.py +0 -0
  86. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/schema.py +0 -0
  87. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/commands/validate.py +0 -0
  88. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/__init__.py +0 -0
  89. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/config_io.py +0 -0
  90. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/config_schema.py +0 -0
  91. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/import_expander.py +0 -0
  92. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/jsonc.py +0 -0
  93. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/migrate.py +0 -0
  94. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/config/normalize.py +0 -0
  95. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/__init__.py +0 -0
  96. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/config_loader.py +0 -0
  97. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/execution/__init__.py +0 -0
  98. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/last_run_store.py +0 -0
  99. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/orchestration/__init__.py +0 -0
  100. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/orchestration/accept.py +0 -0
  101. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/process_worker.py +0 -0
  102. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/sequence_state.py +0 -0
  103. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/setup.py +0 -0
  104. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/test_case.py +0 -0
  105. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/validation/__init__.py +0 -0
  106. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/validation/assertions.py +0 -0
  107. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/validation/result.py +0 -0
  108. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/core/validation/validator.py +0 -0
  109. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/__init__.py +0 -0
  110. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/base_comparator.py +0 -0
  111. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/binary_comparator.py +0 -0
  112. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/csv_comparator.py +0 -0
  113. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/extractor_comparator.py +0 -0
  114. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/factory.py +0 -0
  115. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/file_comparator_base.py +0 -0
  116. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/h5_comparator.py +0 -0
  117. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/json_comparator.py +0 -0
  118. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/numeric_compare.py +0 -0
  119. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/result.py +0 -0
  120. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/script_comparator.py +0 -0
  121. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/script_extract_comparator.py +0 -0
  122. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/text_comparator.py +0 -0
  123. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/file_comparator/xml_comparator.py +0 -0
  124. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/logging_config.py +0 -0
  125. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/reporting/__init__.py +0 -0
  126. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/reporting/diagnosis.py +0 -0
  127. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/__init__.py +0 -0
  128. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/config_runner.py +0 -0
  129. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/json_runner.py +0 -0
  130. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/parallel_json_runner.py +0 -0
  131. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/parallel_yaml_runner.py +0 -0
  132. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/runners/yaml_runner.py +0 -0
  133. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/__init__.py +0 -0
  134. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/html_report.py +0 -0
  135. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/junit_xml_writer.py +0 -0
  136. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/path_resolver.py +0 -0
  137. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest/utils/report_generator.py +0 -0
  138. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest_cli.egg-info/dependency_links.txt +0 -0
  139. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest_cli.egg-info/entry_points.txt +0 -0
  140. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest_cli.egg-info/requires.txt +0 -0
  141. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/src/symtest_cli.egg-info/top_level.txt +0 -0
  142. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__init__.py +0 -0
  143. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/__init__.cpython-312.pyc +0 -0
  144. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/__init__.cpython-313.pyc +0 -0
  145. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/__init__.cpython-39.pyc +0 -0
  146. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/conftest.cpython-312-pytest-7.4.4.pyc +0 -0
  147. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/conftest.cpython-312-pytest-9.0.3.pyc +0 -0
  148. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/conftest.cpython-313-pytest-9.0.2.pyc +0 -0
  149. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/conftest.cpython-39-pytest-8.3.4.pyc +0 -0
  150. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/run_all.cpython-312.pyc +0 -0
  151. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/run_all.cpython-313.pyc +0 -0
  152. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/run_all.cpython-39.pyc +0 -0
  153. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/__pycache__/test_setup_module.cpython-312-pytest-7.4.4.pyc +0 -0
  154. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/conftest.py +0 -0
  155. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/demos/h5_filter_demo.py +0 -0
  156. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/demos/manual_report_example.py +0 -0
  157. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/demos/perf_parallel.py +0 -0
  158. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__init__.py +0 -0
  159. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/__init__.cpython-312.pyc +0 -0
  160. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/__init__.cpython-313.pyc +0 -0
  161. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/__init__.cpython-39.pyc +0 -0
  162. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/test_user_flows.cpython-312-pytest-7.4.4.pyc +0 -0
  163. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/test_user_flows.cpython-312-pytest-9.0.3.pyc +0 -0
  164. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/test_user_flows.cpython-313-pytest-9.0.2.pyc +0 -0
  165. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/__pycache__/test_user_flows.cpython-39-pytest-8.3.4.pyc +0 -0
  166. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/e2e/test_user_flows.py +0 -0
  167. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/circular_config_a.json +0 -0
  168. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/circular_config_b.json +0 -0
  169. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/hourglass_stdout_mismatch.txt +0 -0
  170. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/hourglass_stdout_sample.txt +0 -0
  171. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/inheritance_base.json +0 -0
  172. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/inheritance_chain.json +0 -0
  173. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/inheritance_circular.json +0 -0
  174. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/main_config.json +0 -0
  175. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/circular_config_a.json +0 -0
  176. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/circular_config_b.json +0 -0
  177. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/inheritance_base.json +0 -0
  178. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/inheritance_chain.json +0 -0
  179. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/inheritance_circular.json +0 -0
  180. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/main_config.json +0 -0
  181. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/missing_fields.json +0 -0
  182. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/setup_main.json +0 -0
  183. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/setup_sub.json +0 -0
  184. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/skill_test_cases_golden_file.json +0 -0
  185. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/skill_test_cases_simple.json +0 -0
  186. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/skill_test_cases_simple.yaml +0 -0
  187. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/skill_test_cases_steps.json +0 -0
  188. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/sub_json_tests.yaml +0 -0
  189. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/sub_text_tests.json +0 -0
  190. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/test_cases.json +0 -0
  191. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/test_cases.yaml +0 -0
  192. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/test_cases1.json +0 -0
  193. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/test_with_setup.json +0 -0
  194. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/migration/v1/test_with_setup.yaml +0 -0
  195. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/missing_fields.json +0 -0
  196. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/mock_fail_script.py +0 -0
  197. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/mock_pass_script.py +0 -0
  198. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/setup_main.json +0 -0
  199. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/setup_sub.json +0 -0
  200. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/sub_json_tests.yaml +0 -0
  201. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/sub_text_tests.json +0 -0
  202. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/test_cases.json +0 -0
  203. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/test_cases.yaml +0 -0
  204. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/test_cases1.json +0 -0
  205. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/test_with_setup.json +0 -0
  206. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/fixtures/test_with_setup.yaml +0 -0
  207. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_history_feature.cpython-312-pytest-7.4.4.pyc +0 -0
  208. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_history_feature.cpython-312-pytest-9.0.3.pyc +0 -0
  209. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_history_feature.cpython-313-pytest-9.0.2.pyc +0 -0
  210. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_new_cli_features.cpython-312-pytest-7.4.4.pyc +0 -0
  211. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_new_cli_features.cpython-312-pytest-9.0.3.pyc +0 -0
  212. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_new_cli_features.cpython-313-pytest-9.0.2.pyc +0 -0
  213. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_real_command_execution.cpython-312-pytest-7.4.4.pyc +0 -0
  214. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_real_command_execution.cpython-312-pytest-9.0.3.pyc +0 -0
  215. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_real_command_execution.cpython-313-pytest-9.0.2.pyc +0 -0
  216. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_sequence.cpython-312-pytest-7.4.4.pyc +0 -0
  217. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_sequence.cpython-312-pytest-9.0.3.pyc +0 -0
  218. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/__pycache__/test_sequence.cpython-313-pytest-9.0.2.pyc +0 -0
  219. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_binary_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  220. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_binary_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  221. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_binary_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  222. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_binary_compare.cpython-39-pytest-8.3.4.pyc +0 -0
  223. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_csv_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  224. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_csv_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  225. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_h5_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  226. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_h5_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  227. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_h5_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  228. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_h5_compare.cpython-39-pytest-8.3.4.pyc +0 -0
  229. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_json_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  230. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_json_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  231. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_json_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  232. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_json_compare.cpython-39-pytest-8.3.4.pyc +0 -0
  233. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_text_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  234. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_text_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  235. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_text_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  236. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_text_compare.cpython-39-pytest-8.3.4.pyc +0 -0
  237. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_xml_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  238. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_xml_compare.cpython-312-pytest-9.0.3.pyc +0 -0
  239. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/__pycache__/test_xml_compare.cpython-313-pytest-9.0.2.pyc +0 -0
  240. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_binary_compare.py +0 -0
  241. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_csv_compare.py +0 -0
  242. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_h5_compare.py +0 -0
  243. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_json_compare.py +0 -0
  244. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_text_compare.py +0 -0
  245. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/file_compare/test_xml_compare.py +0 -0
  246. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/__pycache__/test_depends_on.cpython-312-pytest-7.4.4.pyc +0 -0
  247. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/__pycache__/test_depends_on.cpython-312-pytest-9.0.3.pyc +0 -0
  248. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/__pycache__/test_parallel_runner.cpython-312-pytest-7.4.4.pyc +0 -0
  249. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/__pycache__/test_parallel_runner.cpython-313-pytest-9.0.2.pyc +0 -0
  250. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/__pycache__/test_parallel_runner.cpython-39-pytest-8.3.4.pyc +0 -0
  251. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/parallel/test_depends_on.py +0 -0
  252. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/path_handling/__pycache__/test_spaces_in_paths.cpython-312-pytest-7.4.4.pyc +0 -0
  253. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/path_handling/__pycache__/test_spaces_in_paths.cpython-312-pytest-9.0.3.pyc +0 -0
  254. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/path_handling/__pycache__/test_spaces_in_paths.cpython-313-pytest-9.0.2.pyc +0 -0
  255. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/path_handling/__pycache__/test_spaces_in_paths.cpython-39-pytest-8.3.4.pyc +0 -0
  256. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/path_handling/test_spaces_in_paths.py +0 -0
  257. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/test_history_feature.py +0 -0
  258. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/test_new_cli_features.py +0 -0
  259. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/test_real_command_execution.py +0 -0
  260. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/integration/test_sequence.py +0 -0
  261. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/run_all.py +0 -0
  262. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_binary_comparator.cpython-312-pytest-7.4.4.pyc +0 -0
  263. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_binary_comparator.cpython-312-pytest-9.0.3.pyc +0 -0
  264. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_binary_comparator.cpython-313-pytest-9.0.2.pyc +0 -0
  265. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_cli.cpython-312-pytest-7.4.4.pyc +0 -0
  266. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_cli.cpython-312-pytest-9.0.3.pyc +0 -0
  267. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_cli.cpython-313-pytest-9.0.2.pyc +0 -0
  268. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_config_io.cpython-312-pytest-7.4.4.pyc +0 -0
  269. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_config_io.cpython-312-pytest-9.0.3.pyc +0 -0
  270. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_config_io.cpython-313-pytest-9.0.2.pyc +0 -0
  271. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_h5_comparator.cpython-312-pytest-7.4.4.pyc +0 -0
  272. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_h5_comparator.cpython-313-pytest-9.0.2.pyc +0 -0
  273. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_import_expander.cpython-312-pytest-7.4.4.pyc +0 -0
  274. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_import_expander.cpython-312-pytest-9.0.3.pyc +0 -0
  275. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_import_expander.cpython-313-pytest-9.0.2.pyc +0 -0
  276. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_inheritance_expander.cpython-312-pytest-7.4.4.pyc +0 -0
  277. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_inheritance_expander.cpython-313-pytest-9.0.2.pyc +0 -0
  278. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_jsonc.cpython-312-pytest-9.0.3.pyc +0 -0
  279. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_migrate.cpython-312-pytest-7.4.4.pyc +0 -0
  280. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_migrate.cpython-312-pytest-9.0.3.pyc +0 -0
  281. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_result.cpython-312-pytest-7.4.4.pyc +0 -0
  282. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_result.cpython-312-pytest-9.0.3.pyc +0 -0
  283. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_result.cpython-313-pytest-9.0.2.pyc +0 -0
  284. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_run_all.cpython-312-pytest-7.4.4.pyc +0 -0
  285. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_run_all.cpython-312-pytest-9.0.3.pyc +0 -0
  286. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/__pycache__/test_run_all.cpython-313-pytest-9.0.2.pyc +0 -0
  287. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/__pycache__/test_compare_command.cpython-312-pytest-7.4.4.pyc +0 -0
  288. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/__pycache__/test_compare_command.cpython-312-pytest-9.0.3.pyc +0 -0
  289. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/__pycache__/test_compare_command.cpython-313-pytest-9.0.2.pyc +0 -0
  290. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/__pycache__/test_find.cpython-312-pytest-7.4.4.pyc +0 -0
  291. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/test_compare_command.py +0 -0
  292. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/commands/test_find.py +0 -0
  293. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_architecture_guard.cpython-312-pytest-7.4.4.pyc +0 -0
  294. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_architecture_guard.cpython-312-pytest-9.0.3.pyc +0 -0
  295. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_assertions.cpython-312-pytest-7.4.4.pyc +0 -0
  296. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_assertions.cpython-312-pytest-9.0.3.pyc +0 -0
  297. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_assertions.cpython-313-pytest-9.0.2.pyc +0 -0
  298. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_case_env.cpython-312-pytest-7.4.4.pyc +0 -0
  299. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_case_env.cpython-312-pytest-9.0.3.pyc +0 -0
  300. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader.cpython-312-pytest-7.4.4.pyc +0 -0
  301. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader.cpython-312-pytest-9.0.3.pyc +0 -0
  302. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader.cpython-313-pytest-9.0.2.pyc +0 -0
  303. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader_new_features.cpython-312-pytest-7.4.4.pyc +0 -0
  304. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader_new_features.cpython-312-pytest-9.0.3.pyc +0 -0
  305. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_config_loader_new_features.cpython-313-pytest-9.0.2.pyc +0 -0
  306. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution.cpython-312-pytest-7.4.4.pyc +0 -0
  307. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution.cpython-312-pytest-9.0.3.pyc +0 -0
  308. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution.cpython-312.pyc +0 -0
  309. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution.cpython-313-pytest-9.0.2.pyc +0 -0
  310. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution_new_features.cpython-312-pytest-7.4.4.pyc +0 -0
  311. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution_new_features.cpython-312-pytest-9.0.3.pyc +0 -0
  312. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_execution_new_features.cpython-313-pytest-9.0.2.pyc +0 -0
  313. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_history_store.cpython-312-pytest-7.4.4.pyc +0 -0
  314. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_history_store.cpython-313-pytest-9.0.2.pyc +0 -0
  315. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_last_run_store.cpython-312-pytest-7.4.4.pyc +0 -0
  316. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_last_run_store.cpython-312-pytest-9.0.3.pyc +0 -0
  317. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_last_run_store.cpython-313-pytest-9.0.2.pyc +0 -0
  318. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_model_v2.cpython-312-pytest-7.4.4.pyc +0 -0
  319. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_model_v2.cpython-312-pytest-9.0.3.pyc +0 -0
  320. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_orchestration_accept.cpython-312-pytest-7.4.4.pyc +0 -0
  321. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_orchestration_accept.cpython-312-pytest-9.0.3.pyc +0 -0
  322. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_parallel_results_echo.cpython-312-pytest-9.0.3.pyc +0 -0
  323. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_process_worker.cpython-312-pytest-7.4.4.pyc +0 -0
  324. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_process_worker.cpython-312-pytest-9.0.3.pyc +0 -0
  325. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_process_worker.cpython-313-pytest-9.0.2.pyc +0 -0
  326. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_sequence_state.cpython-312-pytest-7.4.4.pyc +0 -0
  327. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_sequence_state.cpython-312-pytest-9.0.3.pyc +0 -0
  328. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_setup.cpython-312-pytest-7.4.4.pyc +0 -0
  329. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_setup.cpython-312-pytest-9.0.3.pyc +0 -0
  330. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_setup.cpython-313-pytest-9.0.2.pyc +0 -0
  331. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_setup.cpython-39-pytest-8.3.4.pyc +0 -0
  332. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_validation_error.cpython-312-pytest-7.4.4.pyc +0 -0
  333. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_validation_error.cpython-312-pytest-9.0.3.pyc +0 -0
  334. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_validation_error.cpython-313-pytest-9.0.2.pyc +0 -0
  335. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_xfail.cpython-312-pytest-7.4.4.pyc +0 -0
  336. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_xfail.cpython-312-pytest-9.0.3.pyc +0 -0
  337. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/__pycache__/test_xfail.cpython-313-pytest-9.0.2.pyc +0 -0
  338. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_architecture_guard.py +0 -0
  339. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_assertions.py +0 -0
  340. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_case_env.py +0 -0
  341. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_config_loader.py +0 -0
  342. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_config_loader_new_features.py +0 -0
  343. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_execution.py +0 -0
  344. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_execution_new_features.py +0 -0
  345. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_last_run_store.py +0 -0
  346. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_model_v2.py +0 -0
  347. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_orchestration_accept.py +0 -0
  348. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_outputs_cleanup.py +0 -0
  349. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_parallel_results_echo.py +0 -0
  350. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_process_worker.py +0 -0
  351. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_sequence_state.py +0 -0
  352. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_setup.py +0 -0
  353. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_validation_error.py +0 -0
  354. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/core/test_xfail.py +0 -0
  355. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/__init__.cpython-312.pyc +0 -0
  356. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_error_analysis.cpython-312-pytest-7.4.4.pyc +0 -0
  357. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_error_analysis.cpython-312-pytest-9.0.3.pyc +0 -0
  358. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_error_analysis.cpython-313-pytest-9.0.2.pyc +0 -0
  359. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_extractor_channels.cpython-312-pytest-7.4.4.pyc +0 -0
  360. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_hourglass_tangent_comparator.cpython-312-pytest-7.4.4.pyc +0 -0
  361. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_hourglass_tangent_comparator.cpython-312-pytest-9.0.3.pyc +0 -0
  362. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_hourglass_tangent_comparator.cpython-313-pytest-9.0.2.pyc +0 -0
  363. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_numeric_compare.cpython-312-pytest-7.4.4.pyc +0 -0
  364. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_numeric_parity.cpython-312-pytest-7.4.4.pyc +0 -0
  365. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_plugin_lifecycle.cpython-312-pytest-7.4.4.pyc +0 -0
  366. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_plugin_lifecycle.cpython-312-pytest-9.0.3.pyc +0 -0
  367. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_script_comparator.cpython-312-pytest-7.4.4.pyc +0 -0
  368. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_script_comparator.cpython-312-pytest-9.0.3.pyc +0 -0
  369. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_script_comparator.cpython-313-pytest-9.0.2.pyc +0 -0
  370. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_script_extract.cpython-312-pytest-7.4.4.pyc +0 -0
  371. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_workspace_plugins.cpython-312-pytest-7.4.4.pyc +0 -0
  372. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_workspace_plugins.cpython-312-pytest-9.0.3.pyc +0 -0
  373. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/__pycache__/test_workspace_plugins.cpython-313-pytest-9.0.2.pyc +0 -0
  374. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_error_analysis.py +0 -0
  375. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_extractor_channels.py +0 -0
  376. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_hourglass_tangent_comparator.py +0 -0
  377. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_numeric_compare.py +0 -0
  378. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_numeric_parity.py +0 -0
  379. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_plugin_lifecycle.py +0 -0
  380. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_script_comparator.py +0 -0
  381. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_script_extract.py +0 -0
  382. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/file_comparator/test_workspace_plugins.py +0 -0
  383. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_json_yaml_runner.cpython-312-pytest-7.4.4.pyc +0 -0
  384. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_json_yaml_runner.cpython-312-pytest-9.0.3.pyc +0 -0
  385. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_json_yaml_runner.cpython-312.pyc +0 -0
  386. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_json_yaml_runner.cpython-313-pytest-9.0.2.pyc +0 -0
  387. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_json_yaml_runner.cpython-39-pytest-8.3.4.pyc +0 -0
  388. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_filter.cpython-312-pytest-7.4.4.pyc +0 -0
  389. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_filter.cpython-312-pytest-9.0.3.pyc +0 -0
  390. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_filter.cpython-313-pytest-9.0.2.pyc +0 -0
  391. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_tag_filter.cpython-312-pytest-7.4.4.pyc +0 -0
  392. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_tag_filter.cpython-312-pytest-9.0.3.pyc +0 -0
  393. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_test_case_tag_filter.cpython-313-pytest-9.0.2.pyc +0 -0
  394. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/__pycache__/test_with_deps.cpython-312-pytest-9.0.3.pyc +0 -0
  395. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/test_json_yaml_runner.py +0 -0
  396. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/test_test_case_filter.py +0 -0
  397. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/test_test_case_tag_filter.py +0 -0
  398. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/runners/test_with_deps.py +0 -0
  399. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_binary_comparator.py +0 -0
  400. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_cli.py +0 -0
  401. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_config_io.py +0 -0
  402. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_config_preparation.py +0 -0
  403. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_h5_comparator.py +0 -0
  404. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_import_expander.py +0 -0
  405. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_jsonc.py +0 -0
  406. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_migrate.py +0 -0
  407. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_result.py +0 -0
  408. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/test_run_all.py +0 -0
  409. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_html_report.cpython-312-pytest-7.4.4.pyc +0 -0
  410. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_html_report.cpython-312-pytest-9.0.3.pyc +0 -0
  411. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_junit_xml_writer.cpython-312-pytest-7.4.4.pyc +0 -0
  412. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_junit_xml_writer.cpython-312-pytest-9.0.3.pyc +0 -0
  413. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_junit_xml_writer.cpython-313-pytest-9.0.2.pyc +0 -0
  414. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_report_generator.cpython-312-pytest-7.4.4.pyc +0 -0
  415. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_report_generator.cpython-312-pytest-9.0.3.pyc +0 -0
  416. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/__pycache__/test_report_generator.cpython-313-pytest-9.0.2.pyc +0 -0
  417. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/test_html_report.py +0 -0
  418. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/test_junit_xml_writer.py +0 -0
  419. {symtest_cli-1.5.5 → symtest_cli-1.6.1}/tests/unit/utils/test_report_generator.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: symtest-cli
3
- Version: 1.5.5
3
+ Version: 1.6.1
4
4
  Summary: Regression testing for CLI applications, with multi-step workflows and numerical file comparison.
5
5
  Author-email: Xiaotong Wang <xiaotongwang98@gmail.com>
6
6
  License: MIT
@@ -214,11 +214,24 @@ symtest run test_cases.json --parallel --workers 4
214
214
  symtest run test_cases.json --parallel --execution-mode process
215
215
  ```
216
216
 
217
- Thread mode currently supports CPU-token allocation, solver thread environment
217
+ Thread mode supports CPU-token allocation, solver thread environment
218
218
  variables, and longest-processing-time-first scheduling using estimates or
219
- runtime history. Process mode provides execution isolation but does not yet use
220
- the resource scheduler. Memory enforcement, priority semantics, and broader
221
- resource scheduling remain active areas of development.
219
+ runtime history. On Windows and Linux, enabling `--history-dir` or declaring
220
+ `min_memory_mb` additionally turns on opportunistic memory admission: each
221
+ case's process-group memory peak is measured via a platform backend (a
222
+ Windows Job Object for the commit-memory peak; a per-command cgroup v2 for
223
+ the whole-group `memory.peak` on Linux, which requires delegation rights and
224
+ degrades to measurement-off without them) and stored in history per source,
225
+ and a case starts only when CPU tokens, the effective memory margin
226
+ (the case's own estimated demand plus a reserve), a physical-memory
227
+ watermark, and (only under potential peak crowding) a minimum startup interval
228
+ are all satisfied — queued
229
+ instead of launching into a memory shortage. Active tasks hold no full
230
+ peak reservations, so occasional overcommit is accepted (zero-OOM not
231
+ guaranteed). Process mode
232
+ provides execution isolation but does not yet use the resource scheduler.
233
+ Linux memory scheduling, priority semantics, and broader resource scheduling
234
+ remain active areas of development.
222
235
 
223
236
  ## CI and reports
224
237
 
@@ -186,11 +186,24 @@ symtest run test_cases.json --parallel --workers 4
186
186
  symtest run test_cases.json --parallel --execution-mode process
187
187
  ```
188
188
 
189
- Thread mode currently supports CPU-token allocation, solver thread environment
189
+ Thread mode supports CPU-token allocation, solver thread environment
190
190
  variables, and longest-processing-time-first scheduling using estimates or
191
- runtime history. Process mode provides execution isolation but does not yet use
192
- the resource scheduler. Memory enforcement, priority semantics, and broader
193
- resource scheduling remain active areas of development.
191
+ runtime history. On Windows and Linux, enabling `--history-dir` or declaring
192
+ `min_memory_mb` additionally turns on opportunistic memory admission: each
193
+ case's process-group memory peak is measured via a platform backend (a
194
+ Windows Job Object for the commit-memory peak; a per-command cgroup v2 for
195
+ the whole-group `memory.peak` on Linux, which requires delegation rights and
196
+ degrades to measurement-off without them) and stored in history per source,
197
+ and a case starts only when CPU tokens, the effective memory margin
198
+ (the case's own estimated demand plus a reserve), a physical-memory
199
+ watermark, and (only under potential peak crowding) a minimum startup interval
200
+ are all satisfied — queued
201
+ instead of launching into a memory shortage. Active tasks hold no full
202
+ peak reservations, so occasional overcommit is accepted (zero-OOM not
203
+ guaranteed). Process mode
204
+ provides execution isolation but does not yet use the resource scheduler.
205
+ Linux memory scheduling, priority semantics, and broader resource scheduling
206
+ remain active areas of development.
194
207
 
195
208
  ## CI and reports
196
209
 
@@ -87,8 +87,10 @@ expand_imports() # 递归展开 import 引用(合并 setup、注入
87
87
  resolve_inheritance() # 解析 extends 继承链(deep-merge、去重 abstract)
88
88
  │
89
89
  ▼
90
- apply_variables() # 注入 --var 全局变量 + case-level variables,
91
- # 递归替换 {placeholder} 占位符
90
+ apply_variables() # 注入 --var 全局变量 + case-level variables(全局同名覆盖),
91
+ # 先递归展开变量值之间的 {引用}(与声明顺序无关,
92
+ # 循环引用报 ValueError 并给出引用链),
93
+ # 再单遍替换 {placeholder} 占位符
92
94
  │
93
95
  ▼
94
96
  展开形态 schema 校验 # 展开后每个可执行用例必须完整
@@ -155,7 +157,7 @@ raw 形态的模板和继承用例允许局部 `execution` 字典,复用完整
155
157
  - JSON/YAML Runner 仅注入不同的 config_loader,通用逻辑统一在顺序 / 并行通用运行器中
156
158
  - xfail:`expected_failure=True` 时 `passed → xpassed`(意外通过,计入失败);任意非 passed → `xfailed`(预期失败,不计入失败)
157
159
  - JUnit XML 状态映射:`xfailed → <skipped>`、`xpassed → <failure>`、`timeout → <error>`、`failed` 按消息类型分 `<failure>` / `<error>`
158
- - 历史:累计平均更新;写入前先按 `regression_threshold` 与旧均值做回归检测;`--update-history` 清除后重记
160
+ - 历史:耗时累计平均更新,写入前先按 `regression_threshold` 与旧均值做回归检测;内存峰值独立记录且**按来源隔离**——Windows 提交峰值写 `last_peak_memory_bytes` / `max_peak_memory_bytes` / `peak_memory_count` / `peak_memory_source`,Linux cgroup 峰值写 `*_linux` 同构字段(`max_peak_memory_bytes_linux` 等,来源由字段名表达),两套口径数值不可比、互不覆盖;有效测量即可记录(含 failed / timeout),调度估算按平台读对应字段的历史最大值;`--update-history` 清除整条记录(耗时 + 内存)
159
161
  - `--last-failed`:每次运行只覆写本次参与执行的用例状态,子集运行不丢失其余用例状态
160
162
 
161
163
  ### 4.2 测试双模式
@@ -166,12 +168,43 @@ raw 形态的模板和继承用例允许局部 `execution` 字典,复用完整
166
168
 
167
169
  ### 4.3 并行与资源调度
168
170
 
169
- - LPT 策略:长任务先行;启用 history 时优先用 `.symtest` 历史 `avg_duration` 排序,其次才看声明的 `estimated_time`
171
+ - LPT 策略:长任务先行;存在历史耗时样本时优先用 `avg_duration`,只有内存记录时回退声明的 `estimated_time`。
172
+
173
+ 线程内存感知模式按运行前的需求来源决定是否预热:没有显式需求和本平台
174
+ 有效历史峰值的用例只学习内存,不计入耗时历史。典型流程为首轮内存、次轮
175
+ 耗时、第三轮历史耗时调度。例外:Linux 无 cgroup v2(平台峰值学习能力
176
+ 缺失)时预热等待永远等不到峰值,直接首轮记录耗时,调度 / 回归检测保持
177
+ 可用。告警要求至少 3 个先前的有效通过耗时样本,先比较再更新。Windows Job basic accounting 的用户态 + 内核态 CPU 时间
178
+ 从 100 ns tick 换算成秒,经 ExecutionResult → TestResult → 重试求和 →
179
+ 序列求和 → 独立 CPU 均值 / 样本数历史。任一部分缺测或包含缓存恢复步骤
180
+ 则省略 CPU 总值;None 表示未知,零值有效。CPU 查询失败保留执行结果。
181
+ 墙钟耗时继续用于调度、报告和超时;**性能回归告警按度量来源选口径**:本次
182
+ 带 CPU 测量(Windows)时只用 CPU 历史比较(同样要求 ≥3 个先前 CPU 样本,
183
+ 样本不足不回退墙钟,避免并行负载下的墙钟误报),无 CPU 测量(Linux /
184
+ 测量失败)退化为墙钟口径。旧历史保留,`--update-history` 按运行前准入信息
185
+ 重建耗时计数。本次不做并发归一化,也不引入 Linux CPU 统计。
170
186
  - `AtomicSemaphore` 资源池:`safe_capacity = max(1, cpu_count − 2)`;每个 case 按 `cpu_cores` acquire/release
171
187
  - 按权重比例分配相对 CPU 核数,并自动注入 `OMP_NUM_THREADS` / `MKL_NUM_THREADS` / `NPROC`
172
188
  - 环境变量优先级:`os.environ` < 调度器注入 < case 级 env(经 subprocess 注入,不改进程全局环境)
173
189
  - 线程模式共享内存 + 结果/打印锁;进程模式经 process worker 隔离;可回退顺序执行
174
190
 
191
+ **内存感知调度——机会式准入**(线程模式 + 启用 `--history-dir` 或声明 `min_memory_mb` 时启用,Windows/Linux 均可;其余情况走原 CPU-only 路径,行为不变):
192
+
193
+ - 内存需求估算优先级:显式 `min_memory_mb` > history 历史最大峰值 × 安全系数(默认 1.2,`memory_margin` 可配) > 可配置默认需求(`memory_default_demand_mb`,默认 4096 MB);不再整池独占。**峰值按来源隔离读取**:Windows 读提交峰值(`max_peak_memory_bytes`),Linux 读 cgroup 峰值(`max_peak_memory_bytes_linux`),两套口径数值不可比,不跨平台混用
194
+ - `JointResourceManager` 联合资源池:CPU 令牌 + 系统快照在同一把锁下原子评估——消除"持 CPU 等内存、持内存等 CPU"的部分占有死锁。快照来源按平台:Windows `GlobalMemoryStatusEx` 可用物理内存 + `GetPerformanceInfo` 提交余量(两种口径分开建模);Linux `/proc/meminfo` 的 `MemAvailable`,且若处于有限额的 cgroup v2 中,对当前组及祖先组取 `memory.max − memory.current` 最小余量与 `MemAvailable` 取更小(有效余量),总量钳制到 `min(MemTotal, 最小组限额)`——Linux 无系统级提交上限概念,单一有效预算同时充当提交余量角色
195
+ - **机会式准入判据**(全部满足才启动):CPU 令牌充足;峰值可能拥挤时距上次授予 ≥ 最小启动间隔(`memory_min_start_interval_s`,默认 1s,错开启动防扎堆);有效余量 ≥ 新任务估算需求 + 保留量(`memory_commit_reserve_mb`,默认 2048 MB;Windows 为提交余量,Linux 为有效余量,同一配置键);可用物理内存 ≥ 物理低水位(`memory_physical_watermark_mb`,默认 2048 MB,独立检查不与需求量挂钩)。提交与物理两种口径分开建模、互不混用(Linux 单一预算下两项检查同源,保留各自语义)
196
+ - **不重复计账**:不再为运行中任务保留完整未来峰值——已占用内存自然体现在系统余量中;未来再次增长导致超额的风险明确接受(允许偶发 OOM,不保证零失败)。实测证明旧的"完整预留 + 快照兜底"策略在 31.8 GiB 机器上把 2/4 worker 并发压到 1
197
+ - 平铺与依赖调度均在提交到线程池前检查准入。暂未获准的就绪用例留在调度队列,后续符合条件的用例可以先运行。调度器在用例结束时或按轮询周期重试,启动间隔即将结束时提前重试
198
+ - 拒绝原因(含关键数值:需求/保留量/余量/水位)来自实际判断路径,记入 `waiter.deny_reason` 与调度日志,供验收归因
199
+ - 物理低水位校验:已知总物理内存时,非零水位必须严格小于总量,否则初始化抛出 `ValueError`,避免永久排队。请降低 `memory_physical_watermark_mb`,或设为 0 禁用该检查。
200
+ - 内存不足排队等待而非降级(CPU 的 10s 降级逻辑不适用内存);需求未知(`None`)或需求加保留量超过系统提交上限(结构性不可满足,如 pagefile 过小 / cgroup 限额过小)时跳过余量检查放行(物理水位与启动间隔仍生效),交由用例自身失败暴露问题,不永久排队
201
+ - Linux 测量与降级:每条命令一个独立 cgroup v2 子组。shell 把自身 PID 写入 `cgroup.procs`,发送确认字节后 `exec`,后代继承组成员身份。测量后端消费确认字节后把输出交给执行器;管道 EOF 后轮询 `cgroup.procs` 排空才读 `memory.peak`。若无法创建测量子组或迁移失败,执行器回退到普通子进程,命令只执行一次,峰值未知;准入仍按手动/默认需求执行。不设 `memory.max` 硬上限、不主动杀任务(超时整组清理与 Windows 语义对齐)
202
+ - 组内全部后代与启动器共享命令超时期限;管道 EOF 后继续等待组内进程结束,期限耗尽则终止整组并报告超时
203
+ - Windows Job 排空查询的 PID 数组使用指针宽度;缓冲区不足或列表不完整时扩容重试(每轮最多 8 次)。查询失败抛出 `JobError`,不作为超时;执行器记录警告,保留 `communicate` 已取得的输出和返回码,不重新执行命令,峰值保持未知。此时无法确认后代排空,关闭 Job 仍清理残留进程。
204
+ - 上述参数为 `symtest.json` run 段配置键(CLI 第一版不提供旗标);进程模式不接入资源调度(保持既有行为)
205
+
206
+ 仅在存在活动任务且“活动任务估算峰值之和 + 新任务需求 + 提交保留量”大于当前提交余量时,才要求距上次授予达到启动间隔;余量充足或活动任务全部结束时立即放行。该估算可能重复计入已占用内存,只用于保守地触发间隔,不作为硬预留。快照或需求未知且存在活动任务时保守保留间隔。
207
+
175
208
  ### 4.4 DAG 依赖调度语义
176
209
 
177
210
  - `depends_on` 非空自动切换拓扑模式;无依赖时走零开销 fast path
@@ -185,12 +218,19 @@ raw 形态的模板和继承用例允许局部 `execution` 字典,复用完整
185
218
 
186
219
  PathResolver 解析(系统命令直通、shell builtin 平台包装、复合命令拆分解析)
187
220
  → 声明产物清理(`execution.outputs` 声明的文件在 subprocess 启动前删除;只允许声明文件,声明目录或 workspace 根一律拒绝;路径 resolve 后必须在 workspace 内,删除失败大声失败记为执行错误;每次 attempt 都执行,`--resume` 跳过的 step 天然不清理)
188
- → subprocess 隔离执行(捕获 stdout / stderr / returncode / duration)
221
+ → subprocess 隔离执行(捕获 stdout / stderr / returncode / duration;测量启用时经平台进程组后端整组执行——Windows Job Object 以挂起创建→绑定→恢复捕获 `PeakJobMemoryUsed`,Linux cgroup v2 以 shell 先迁移进组再 exec 捕获 `memory.peak`,超时均终止整组,测量失败回退普通 subprocess 路径、峰值保持未知)
189
222
  → 断言校验(return_code / contains / matches / compare_files)
190
223
  → 失败按 `retry_count` 重试
191
224
  → 超时 kill 整个进程组
192
225
  → 返回附带 `next_action_hint` 的结构化结果。
193
226
 
227
+ 进程组内存峰值是**执行事实**(原则 2),随结果可选携带 `peak_memory_bytes` /
228
+ `peak_memory_source`(Windows 为提交内存口径 `windows_job_commit`,Linux 为
229
+ cgroup v2 记账口径 `linux_cgroup_v2`——含文件缓存,非物理内存峰值;两口径
230
+ 数值不可比;无可测量时字段缺席,不用 0 表示未知)。聚合规则:retry 取各
231
+ attempt 最大值;sequence 取实际执行步骤的最大值(步骤串行,不求和;
232
+ `--resume` 跳过的步骤不补造数据)。
233
+
194
234
  ### 4.6 断言与文件比较集成
195
235
 
196
236
  - `compare_files` 是一等断言,经 ComparatorFactory 按类型分发(详见 §6)
@@ -202,7 +242,7 @@ PathResolver 解析(系统命令直通、shell builtin 平台包装、复合
202
242
  | 状态 | 内容 | 更新时机 / 语义 |
203
243
  |---|---|---|
204
244
  | 项目默认配置 `symtest.json` | `run` 组运行选项 + `report` 组报告路径 | 运行启动时读取(CLI > 选中的参数文件 > 内置默认);默认入口 `.symtest/symtest.json`(宽容校验,问题只警告),`--defaults FILE` 显式替代入口(严格校验,问题报错退出,不与默认文件叠加);只读,运行中不更新 |
205
- | 历史记录 | per-case `avg_duration` / `last_duration` / `run_count` | 运行后累计平均更新;先回归检测后写入 |
245
+ | 历史记录 | per-case `avg_duration` / `last_duration` / `run_count` + 可选内存字段(Windows `last_peak_memory_bytes` / `max_peak_memory_bytes` / `peak_memory_count` / `peak_memory_source`;Linux 为同构 `*_linux` 字段) | 运行后累计平均更新;先回归检测后写入;内存只要有有效测量即记录(含失败用例),估算读平台自身字段的历史最大值 |
206
246
  | `last_run.json` | case → 上次状态 | 覆写参与执行的用例;未执行的保留旧状态 |
207
247
  | `sequence_state/<case>.json` | steps + case expected 的 SHA-256 配置哈希、已通过的 step 集 | 每步 pass 后保存;全部通过后清理 |
208
248
  | `sequence_state/cache/*.log` | 步骤输出缓存 | 与上同步;供 resume 拼接复现 |
@@ -325,6 +365,15 @@ JSON/YAML 配置,学习成本更低且无交互界面维护负担。
325
365
  | 项目级默认配置(`.symtest/symtest.json`) | 常用运行参数(并行、误差分析、依赖展开、报告路径)项目内固化,`symtest run` 免重复输入。三级优先级(CLI > 项目配置 > 内置默认)靠 tri-state 布尔项实现(argparse 默认 `None`,`--no-xxx` 显式关闭);未知键/坏 JSON 宽容处理只警告——配置是纯加分项,不阻断运行。替代各仓库自建入口脚本保存默认参数的用法 |
326
366
  | `--defaults FILE` 显式默认参数文件 | 同一 workspace 需要多套默认参数(如回归 vs 验证)时免建项目 runner。显式文件**完全替代** `.symtest/symtest.json`(不叠加合并,防止验证配置漏写键时静默继承回归设置),并切换为**严格校验**(缺失/坏 JSON/未知键/类型错报错退出)——显式指定即用户意图,静默回退到另一套设置比报错更危险。默认入口保持宽容语义不变,与 `--no-config` 互斥 |
327
367
  | Logging 统一化 | 通过 `logging` 模块集中管理,CLI 入口激活控制台输出,库用户按需启用 |
368
+ | Windows Job Object 测量进程组峰值 | 挂起创建→绑定 Job→恢复执行消除后代逃逸竞态;峰值由操作系统持续维护,无需采样即可捕捉短暂尖峰;`PeakJobMemoryUsed` 为提交内存口径,如实命名为 `peak_memory_source=windows_job_commit`,不冒充物理内存峰值 |
369
+ | Linux cgroup v2 测量进程组峰值 | 命令必须**先进组再执行**——启动后迁移 PID 会漏掉已创建的子进程;用 shell 把自身 pid 写入 `cgroup.procs` 后 `exec` 实现(不使用 `preexec_fn`,规避多线程 fork 死锁风险)。每命令一个独立子组,`memory.peak` 高水位由内核持续维护;该口径含文件缓存,如实命名 `linux_cgroup_v2`,不冒充物理内存或提交内存峰值。不设 `memory.max` 硬上限(只测不限),无委派权限/老内核时降级为不测量,准入不受影响 |
370
+ | 内存峰值历史按来源隔离 | Windows commit 峰值与 Linux cgroup 峰值口径不同且数值不可比:分别写 `max_peak_memory_bytes` 与 `max_peak_memory_bytes_linux` 字段,调度估算按平台读取对应字段——跨平台混用会把 5 GiB 提交峰值当作 cgroup 峰值(或反之)参与准入,得出错误余量判断 |
371
+ | Linux 快照钳制 cgroup 限额余量 | 只看 `/proc/meminfo` 会把宿主机内存当作可用额度,容器/限额环境下准入形同虚设;对当前组及祖先组取 `memory.max − memory.current` 最小余量与 `MemAvailable` 取更小,总量同步钳制使物理水位校验在限额环境中正确拦截 |
372
+ | CPU+内存联合原子准入 | 两个独立信号量会产生"持 CPU 等内存、持内存等 CPU"的部分占有死锁;联合池在同一把锁下原子评估池令牌 + 系统快照(物理可用与提交余量分别检查,二者防护目标不同) |
373
+ | 机会式内存准入(不做完整峰值预留) | 旧的"系统余量须覆盖运行中任务的完整预留 + 新请求"会重复计入已分配内存(实测把 2/4 worker 并发压到 1),且把提交峰值误用于物理内存检查。机会式准入只检查新任务自身需求 + 保留量;已占用内存由系统余量自然体现;峰值可能拥挤时授予间保持最小启动间隔错开扎堆;允许偶发 OOM,不保证零失败 |
374
+ | 内存排队而非降级 | CPU 令牌超时降级到 1 核是吞吐启发式;内存降级启动正是并行 OOM 失败的来源,故内存不足持续等待(周期唤醒重评外部内存变化),结构性不可满足(需求加提交保留量超提交上限)才跳过检查放行 |
375
+ | 内存历史用最大值而非平均 | 峰值口径天然只增,偶发高峰被平均稀释后会在下次调度再次 OOM;估算加安全系数(默认 1.2,可配) |
376
+ | 内存测量失败回退普通路径 | 测量是加分项不是验收条件:后端失败/平台不支持/进程模式时峰值保持未知(None,不用 0 伪装),执行与判定语义完全不变;Linux 无 cgroup 委派权限时准入仍可用手动/默认需求执行,启动日志明确提示无法学习整组峰值 |
328
377
  | 控制台 log/report 通道分离 | 控制台(stderr logging)只承载进度与摘要日志;报告仅在显式 `--output-format text/json/html` 时走 stdout(可重定向落盘),细节经 `--text-report`/`--html-report`/`--junit-xml` 写文件。修复原 `print_report` 把整份文本报告经 `logger.info` 重放到终端的通道混流——`--error-analysis-all` 下逐用例 error_stats 会淹没 pass/failed 摘要 |
329
378
 
330
379
  ---
@@ -112,8 +112,10 @@ duplicate-name check # extends references must be unambiguous
112
112
  resolve_inheritance() # Resolve extends chains (deep-merge, remove abstract)
113
113
  │
114
114
  ▼
115
- apply_variables() # Inject --var global variables + case-level variables,
116
- # recursively replace {placeholder} tokens
115
+ apply_variables() # Inject --var global variables + case-level variables (globals
116
+ # win on key conflicts); first expand {references} among variable
117
+ # values recursively (order-independent, circular references raise
118
+ # ValueError with the chain), then replace {placeholder} tokens
117
119
  │
118
120
  ▼
119
121
  expanded-form validation # every executable case must be complete after expansion
@@ -181,7 +183,7 @@ the exact file and field.
181
183
  - JSON/YAML runners differ only by injected config_loader; common logic lives in the generic sequential / parallel runners
182
184
  - xfail: when `expected_failure=True`, `passed → xpassed` (unexpected pass, counted as a suite failure); any non-passed status → `xfailed` (expected failure, not counted)
183
185
  - JUnit XML status mapping: `xfailed → <skipped>`, `xpassed → <failure>`, `timeout → <error>`, `failed` split into `<failure>` / `<error>` by message type
184
- - History: cumulative average update; regression is checked against the prior average before writing; `--update-history` clears then records
186
+ - History: duration uses cumulative average update; regression is checked against the prior average before writing; memory peaks are recorded independently and **isolated by source** — Windows commit peaks write `last_peak_memory_bytes` / `max_peak_memory_bytes` / `peak_memory_count` / `peak_memory_source`, Linux cgroup peaks write the parallel `*_linux` fields (e.g. `max_peak_memory_bytes_linux`; the field name carries the source), the two bases are incomparable and never overwrite each other; any run with a valid measurement records (including failed / timeout), scheduling estimates use the platform's own historical maximum; `--update-history` clears the whole record (duration + memory)
185
187
  - `--last-failed`: each run overwrites only the statuses of executed cases; subset runs never lose states of unexecuted cases
186
188
 
187
189
  ### 4.2 Two Case Modes
@@ -193,12 +195,50 @@ JSON Schema; it is not restated here.
193
195
 
194
196
  ### 4.3 Parallelism and Resource Scheduling
195
197
 
196
- - LPT strategy: long tasks first; with history enabled, historical `avg_duration` takes precedence over declared `estimated_time`
198
+ - LPT strategy: long tasks first; a historical timing sample makes `avg_duration` take precedence over declared `estimated_time`; memory-only entries fall back to the declared estimate.
199
+
200
+ Timing history excludes memory-aware thread cases whose pre-run memory estimate
201
+ has neither a declared demand nor a valid platform peak. Memory is still learned
202
+ on the first run; the second can collect timing, the third can schedule from it.
203
+ Exception: on Linux without cgroup v2 (platform peak learning unavailable) the
204
+ warm-up wait can never succeed, so timing history starts on the first run and
205
+ scheduling/regression detection stay functional.
206
+ Regression requires 3 prior successful timing samples, checked before updating.
207
+ Windows Job basic accounting supplies user + kernel CPU seconds (100 ns ticks
208
+ converted to seconds), propagated through ExecutionResult → TestResult → retry
209
+ sum → sequence sum → optional CPU history averages/counts. Any missing part or
210
+ cached resumed step invalidates the aggregate; None is unknown, zero is valid.
211
+ CPU query failure preserves execution/verdicts. Wall durations retain their
212
+ scheduling/report/timeout meaning; **regression warnings pick the metric by
213
+ source**: when this run carries a CPU measurement (Windows) only CPU history is
214
+ compared (also requiring ≥3 prior CPU samples; immature samples do not fall
215
+ back to wall time, avoiding wall-clock false positives under parallel load),
216
+ and runs without a CPU measurement (Linux / query failure) fall back to the
217
+ wall-clock metric. Existing history
218
+ is preserved; `--update-history` rebuilds timing counts using pre-run admission
219
+ knowledge. There is no concurrency normalization or Linux CPU accounting here.
197
220
  - `AtomicSemaphore` resource pool: `safe_capacity = max(1, cpu_count − 2)`; per-case acquire/release by `cpu_cores`
198
221
  - Relative CPU cores assigned proportionally by weight; auto-injects `OMP_NUM_THREADS` / `MKL_NUM_THREADS` / `NPROC`
199
222
  - Environment variable precedence: `os.environ` < scheduler injection < case-level env (via subprocess, never mutating global `os.environ`)
200
223
  - Thread mode shares memory with result/print locks; process mode isolates via process workers; sequential fallback available
201
224
 
225
+ **Memory-aware scheduling — opportunistic admission** (enabled for thread mode when `--history-dir` is set or `min_memory_mb` is declared, on both Windows and Linux; otherwise the original CPU-only path is kept unchanged):
226
+
227
+ - Demand estimate priority: explicit `min_memory_mb` > historical max peak × safety factor (default 1.2, configurable via `memory_margin`) > configurable default demand (`memory_default_demand_mb`, default 4096 MB); no more whole-pool exclusive runs. **Peaks are read per source**: Windows reads the commit peak (`max_peak_memory_bytes`), Linux reads the cgroup peak (`max_peak_memory_bytes_linux`); the two scales are not comparable and never mixed across platforms
228
+ - `JointResourceManager` joint pool: CPU tokens + a system snapshot are evaluated atomically under a single lock — eliminating the partial-holding deadlock of "holding CPU while waiting for memory and vice versa". Snapshot sources are platform-specific: Windows uses `GlobalMemoryStatusEx` available physical memory plus `GetPerformanceInfo` commit margin (two separately modelled budgets); Linux uses `MemAvailable` from `/proc/meminfo`, additionally clamped by the smallest `memory.max − memory.current` margin over the current cgroup and all ancestors when running under a limited cgroup v2 (the effective margin), with the total clamped to `min(MemTotal, smallest group limit)` — Linux has no system-level commit-limit concept, so the single effective budget also plays the commit-margin role
229
+ - **Admission criteria** (all must hold to start): CPU tokens suffice; under potential peak crowding, at least the minimum startup interval (`memory_min_start_interval_s`, default 1s) has elapsed since the previous grant (staggered starts, avoiding thundering herds that all see the same headroom); effective margin ≥ the new task's estimated demand + reserve (`memory_commit_reserve_mb`, default 2048 MB; the commit margin on Windows, the effective margin on Linux — same config key); available physical memory ≥ the physical watermark (`memory_physical_watermark_mb`, default 2048 MB, an independent pressure check not tied to the demand). The commit and physical budgets are modelled separately and never mixed (on Linux both checks draw from the same single budget, each keeping its own semantics)
230
+ - **No double counting**: full future peaks are no longer reserved for running tasks — memory already in use shows up naturally in the system headroom; the residual risk of a running task growing again is explicitly accepted (occasional OOM allowed, zero-failure not guaranteed). Measurements proved the old "full reservations + snapshot backstop" policy squeezed 2/4-worker concurrency down to 1 on a 31.8 GiB machine
231
+ - Both flat and dependency scheduling check admission before submitting cases to the thread pool. Denied ready cases stay in the scheduler queue, and later eligible cases can run first. The scheduler retries on case completion or at the polling interval, with earlier retries when a startup interval expires
232
+ - Denial reasons (with key numbers: demand / reserve / headroom / watermark) come from the actual decision path and are stored on `waiter.deny_reason` plus the scheduler log for acceptance attribution
233
+ - Physical watermark validation: when total physical memory is known, a nonzero watermark must be strictly below it; otherwise initialization raises `ValueError` instead of waiting forever. Lower `memory_physical_watermark_mb`, or set it to 0 to disable this check.
234
+ - Memory shortage queues instead of downgrading (the CPU 10s-downgrade heuristic does not apply to memory); unknown demand (`None`) or demand plus reserve beyond the system commit limit (structurally unsatisfiable, e.g. a small pagefile / a small cgroup limit) skip the margin check and run (physical watermark and startup interval still apply), leaving the problem to the case's own failure rather than queueing forever
235
+ - Linux measurement and fallback: each command gets a fresh cgroup v2 subgroup. The wrapping shell writes its PID to `cgroup.procs`, sends an acknowledgment byte, then calls `exec`. Descendants inherit group membership. The backend consumes the acknowledgment before handing the output to the executor. After pipe EOF the executor polls `cgroup.procs` until the group drains, then reads `memory.peak`. If the measurement subgroup cannot be created or migration fails, the executor runs the command once as a plain subprocess with an unknown peak. Admission still uses declared or default demand. There is no `memory.max` hard limit or proactive termination. Timeout cleanup covers the whole group, as on Windows.
236
+ - All descendants in the group share the launcher's command deadline. After pipe EOF, execution still waits for the group to finish; reaching the deadline terminates the group and reports a timeout.
237
+ - Windows Job drain queries use pointer-sized PID entries and resize insufficient buffers or incomplete lists (at most 8 retries per poll). Query failures raise `JobError`, not a timeout; the executor logs a warning, preserves the output and return code already obtained from `communicate`, does not rerun the command, and leaves the peak unknown. Descendant completion cannot be confirmed in this case; closing the Job still cleans up remaining processes.
238
+ - These parameters are `symtest.json` run-section config keys (no CLI flags in the first version); process mode does not participate in resource scheduling (unchanged behavior)
239
+
240
+ The startup interval applies only while tasks are active and their summed estimated peaks plus the new demand and commit reserve exceed current commit headroom. Ample headroom or an empty active set permits immediate admission. This conservative signal may double-count allocated memory; it triggers spacing, never a hard reservation. With active tasks and unknown demand or headroom, spacing is retained conservatively.
241
+
202
242
  ### 4.4 DAG Dependency Scheduling Semantics
203
243
 
204
244
  - Non-empty `depends_on` automatically switches to topological mode; no deps walks a zero-overhead fast path
@@ -217,10 +257,20 @@ workspace root are always rejected; resolved
217
257
  paths must stay inside the workspace, deletion failure fails loudly as an execution
218
258
  error; runs on every attempt, and steps skipped by `--resume` are naturally not
219
259
  cleaned) → subprocess-isolated execution (captures
220
- stdout / stderr / returncode / duration) → assertions (return_code / contains /
260
+ stdout / stderr / returncode / duration; when measurement is enabled the process group runs via the platform backend — on Windows a Job Object created suspended → assigned → resumed captures the group's commit-memory peak `PeakJobMemoryUsed`, on Linux the wrapping shell migrates into a per-command cgroup v2 before `exec` and the executor later reads `memory.peak`; on timeout both terminate the whole group; measurement failure falls back to the plain subprocess path with the peak left unknown) → assertions (return_code / contains /
221
261
  matches / compare_files) → retries on failure up to `retry_count` → kills the
222
262
  whole process group on timeout → structured result with `next_action_hint`.
223
263
 
264
+ The process-group memory peak is an **execution fact** (Principle 2), optionally
265
+ carried on results as `peak_memory_bytes` / `peak_memory_source` (`windows_job_commit`
266
+ for the Windows commit-memory basis, `linux_cgroup_v2` for the Linux cgroup v2
267
+ accounting basis — which includes file cache; neither is a physical-memory peak and
268
+ the two scales are not comparable; fields are absent when no valid
269
+ measurement exists — never 0-for-unknown). Aggregation: retry takes the maximum
270
+ across attempts; a sequence takes the maximum across actually-executed steps
271
+ (steps run serially, no summation; steps skipped by `--resume` contribute no
272
+ fabricated data).
273
+
224
274
  ### 4.6 Assertions and File Comparison Integration
225
275
 
226
276
  - `compare_files` is a first-class assertion dispatched via ComparatorFactory (see §6)
@@ -232,7 +282,7 @@ whole process group on timeout → structured result with `next_action_hint`.
232
282
  | State | Content | Update timing / semantics |
233
283
  |---|---|---|
234
284
  | Project defaults `symtest.json` | `run` group run options + `report` group report paths | Read once at run start (CLI > project config > built-in defaults); read-only, never updated during a run |
235
- | History | per-case `avg_duration` / `last_duration` / `run_count` | Cumulative average after runs; regression-checked before writing |
285
+ | History | per-case `avg_duration` / `last_duration` / `run_count` + optional memory fields (`last_peak_memory_bytes` / `max_peak_memory_bytes` / `peak_memory_count` / `peak_memory_source` on Windows; parallel `*_linux` fields on Linux) | Duration: cumulative average after runs, regression-checked before writing (passed cases only); memory: recorded for any run with a valid measurement, scheduling estimates use the platform's own historical max |
236
286
  | `last_run.json` | case → last status | Overwrites executed cases; unexecuted retain old status |
237
287
  | `sequence_state/<case>.json` | SHA-256 config hash (steps + case expected), set of passed steps | Saved after each passing step; cleaned up after full pass |
238
288
  | `sequence_state/cache/*.log` | Step output cache | Kept in sync; spliced back on resume |
@@ -309,8 +359,7 @@ no interactive UI maintenance burden.
309
359
  | Semaphore-based CPU core management | More fine-grained than thread pool worker count; allows different cases to declare different core requirements |
310
360
  | LPT scheduling strategy | Long tasks start first, reducing tail latency; prioritizes `.symtest` historical data |
311
361
  | Cumulative average history update | Simple and intuitive; single anomalies are naturally diluted over many runs |
312
- | Regression check before update | Compare against old average first, then update; ensures comparison is against the "historical baseline" |
313
- | `.symtest/` directory | Non-intrusive to the user's directory view; JSON format for easy debugging; state files centrally managed |
362
+ | Regression check before update | Compare against old average first, then update; ensures comparison is against the "historical baseline" || `.symtest/` directory | Non-intrusive to the user's directory view; JSON format for easy debugging; state files centrally managed |
314
363
  | Environment variable injection | Scientific computing solvers often ignore Python-level thread control |
315
364
  | Comparator factory + plugin discovery | Creates comparators by file type; workspace `comparators/` directory for auto-discovered plugins |
316
365
  | subprocess isolated execution | Each test case runs in an independent subprocess, ensuring tests don't affect each other |
@@ -326,6 +375,15 @@ no interactive UI maintenance burden.
326
375
  | JUnit XML output | Compatible with GitLab CI / Jenkins / CircleCI and other major CI system test report formats |
327
376
  | Project-level default configuration (`.symtest/symtest.json`) | Frequently used run options (parallel, error analysis, dependency expansion, report paths) are stored per project so `symtest run` needs no repeated input. Three-level precedence (CLI > project config > built-in defaults) is implemented with tri-state booleans (argparse defaults to `None`; `--no-xxx` turns a flag off explicitly); unknown keys and bad JSON are handled tolerantly with a warning only — the config is a pure bonus and never blocks a run. Replaces per-repo entry scripts whose only job was storing default arguments |
328
377
  | Console log/report channel separation | The console (stderr logging) carries progress and summary logs only. Reports reach stdout solely via an explicit `--output-format text/json/html` (redirectable to a file), with full details written to disk via `--text-report`/`--html-report`/`--junit-xml`. Fixes the channel confusion where the former `print_report` replayed the whole text report through `logger.info` — with `--error-analysis-all` the per-case error_stats flooded the terminal and buried the pass/failed summary |
378
+ | Windows Job Object measures process-group peaks | Create-suspended → assign → resume eliminates the descendant-escape race; the OS maintains the peak continuously, catching short spikes without sampling; `PeakJobMemoryUsed` is commit memory, honestly named `peak_memory_source=windows_job_commit`, never presented as a physical-memory peak |
379
+ | Linux cgroup v2 measures process-group peaks | The command must **enter the group before executing** — migrating a pid after start would miss children spawned in between; implemented by a wrapping shell that writes its own pid into `cgroup.procs` and then `exec`s (no `preexec_fn`, avoiding fork deadlocks in multithreaded parents). One fresh sub-group per command; the kernel maintains the `memory.peak` high-water mark. That basis includes file cache, honestly named `linux_cgroup_v2`, never presented as physical or commit memory. No `memory.max` hard limit (measure, don't restrict); without delegation rights or on old kernels, measurement degrades to off without affecting admission |
380
+ | Peak history isolated by source | Windows commit peaks and Linux cgroup peaks use different bases with incomparable numbers: they are stored as `max_peak_memory_bytes` vs `max_peak_memory_bytes_linux`, and admission reads the platform's own field — mixing them across platforms would feed a 5 GiB commit peak into cgroup admission (or vice versa) and produce wrong headroom decisions |
381
+ | Linux snapshot clamps cgroup limit margins | Reading only `/proc/meminfo` treats host memory as the available budget, making admission meaningless inside containers/limited groups; the smallest `memory.max − memory.current` margin over the current group and ancestors is combined with `MemAvailable` via min, and the total is clamped so the physical watermark check still traps correctly under limits |
382
+ | Joint atomic CPU+memory admission | Two independent semaphores allow a partial-holding deadlock ("holding CPU while waiting for memory and vice versa"); the joint pool evaluates pool tokens + system snapshot atomically under one lock (physical availability and commit margin are checked separately — they guard different failure modes) |
383
+ | Opportunistic memory admission (no full peak reservations) | The old policy requiring system headroom to cover all active reservations plus the new request double-counted allocated memory (measured to squeeze 2/4-worker concurrency to 1) and misused commit peaks for the physical-memory check. Opportunistic admission checks only the new task's own demand + reserve; memory in use shows up naturally in the system headroom; a minimum startup interval staggers grants only under potential peak crowding; occasional OOM is accepted, zero-failure is not guaranteed |
384
+ | Memory queues instead of downgrading | The CPU token timeout-downgrade is a throughput heuristic; downgraded memory launch is precisely the source of parallel OOM failures, so memory shortage waits indefinitely (periodic wake-ups re-evaluate external memory changes); only structurally unsatisfiable requests (demand plus commit reserve beyond the commit limit) skip the check and run |
385
+ | Memory history uses the maximum, not the average | Peaks are naturally monotone; an occasional spike diluted by averaging would OOM again at the next scheduling; estimates add a safety factor (default 1.2, configurable) |
386
+ | Memory measurement failure falls back to the plain path | Measurement is a bonus, not an acceptance criterion: on backend failure / unsupported platform / process mode the peak stays unknown (None, never 0-for-unknown) and execution and verdict semantics are unchanged; on Linux without cgroup delegation rights admission still runs on manual/default demands, with the startup log clearly stating that whole-group peaks cannot be learned |
329
387
 
330
388
  ---
331
389
 
@@ -1277,6 +1277,34 @@ test_cases:
1277
1277
 
1278
1278
  > **安全设计**:只有 `variables` 字典中存在的 key 才会被替换。`{xxx}` 若无匹配不会报错,而是原样保留。因此 `expected.output_matches` 中的正则模式(如 `{2,}`、`\d{4}`)不受影响。
1279
1279
 
1280
+ ### 嵌套引用(变量引用变量)
1281
+
1282
+ 变量值可以引用其他变量,展开结果与声明顺序无关:
1283
+
1284
+ ```bash
1285
+ symtest run test_cases.json --var BASE_PATH=/tmp
1286
+ ```
1287
+
1288
+ ```json
1289
+ {
1290
+ "test_cases": [
1291
+ {
1292
+ "name": "求解器测试",
1293
+ "variables": {"work": "{BASE_PATH}/work"},
1294
+ "execution": { "command": "python", "args": ["{work}/run.py"], "expected": { "return_code": 0 } }
1295
+ }
1296
+ ]
1297
+ }
1298
+ ```
1299
+
1300
+ `{work}` 的值为 `{BASE_PATH}/work`,最终展开为 `/tmp/work`。用例变量与全局变量合并后(全局同名覆盖)统一展开;`setup` 仅使用全局变量,全局变量之间也可相互引用。
1301
+
1302
+ 规则与限制:
1303
+
1304
+ - 变量值中未匹配的占位符原样保留(与上方安全设计一致)。
1305
+ - 变量直接或间接引用自身(如 `a → b → a`)是配置错误,加载时报告完整引用链,如 `Circular variable reference detected: a -> b -> a`。
1306
+ - **已知边界**:变量值中的字面 `{name}` 若恰好与某个已定义变量同名,会被递归展开。当前不支持转义字面花括号,请通过调整变量命名规避。
1307
+
1280
1308
  ### 用法
1281
1309
 
1282
1310
  #### CLI
@@ -1617,7 +1645,7 @@ test_cases:
1617
1645
  |---|---|
1618
1646
  | `cpu_cores` | 所需 CPU 核心数,默认 1。框架用信号量控制分配,超限任务排队等待 |
1619
1647
  | `estimated_time` | 预估耗时(秒),用于 LPT 调度(长任务优先启动) |
1620
- | `min_memory_mb` | 预估内存(MB),目前仅用于日志警告 |
1648
+ | `min_memory_mb` | 内存需求估算量(MB)。内存管理启用时直接作为准入的需求估算;未启用时仅参与日志提示 |
1621
1649
  | `priority` | 优先级 0-10,目前仅用于信息标注 |
1622
1650
 
1623
1651
  框架行为:
@@ -1625,15 +1653,95 @@ test_cases:
1625
1653
  - 任务启动时自动注入 `OMP_NUM_THREADS`、`MKL_NUM_THREADS`、`NPROC` 环境变量,防止求解器线程失控
1626
1654
  - 按 `estimated_time` 降序调度(LPT 策略);若启用 `--history-dir`,优先使用历史 `avg_duration` 排序
1627
1655
 
1656
+ ### 内存感知调度(机会式准入,Windows / Linux)
1657
+
1658
+ 大规模并行求解器(如 Abaqus)偶发因内存不足失败。在**线程模式**下(Windows 或
1659
+ Linux),启用 `--history-dir` 或声明 `min_memory_mb` 后,框架进入机会式内存准入:
1660
+ 根据历史需求和当前内存压力决定是否启动新任务,减少并行引起的内存失败;
1661
+ 允许偶发超出内存,不保证零 OOM。
1662
+
1663
+ **峰值测量**(按平台选择后端):
1664
+
1665
+ - **Windows**:每个命令经 Job Object 以"挂起创建 → 绑定 → 恢复"方式执行,
1666
+ 启动器及其后代进程(求解器)整组被统计**提交内存峰值**(commit memory
1667
+ 口径,非物理内存峰值)。若管道读取及主进程等待已完成,但 Job 排空查询
1668
+ 失败,则记录警告,保留命令的输出和返回码,不报告超时、不重新执行,
1669
+ 峰值标记为未知;此时无法确认后代全部完成,关闭 Job 仍会清理残留进程;
1670
+ - **Linux**:每个命令在独立 cgroup v2 子组中执行(要求 cgroup v2 且有委派
1671
+ 权限、内核 ≥ 6.8 提供 `memory.peak`)——包装 shell 先把自身 pid 写入
1672
+ `cgroup.procs` 再 `exec`,命令及其全部后代从一开始就在组内;读取的
1673
+ `memory.peak` 为 **cgroup 记账口径**(含匿名内存、文件缓存与部分内核
1674
+ 内存),非物理内存峰值,与 Windows 提交口径数值不可比。
1675
+
1676
+ 两平台一致的语义:retry 取各次尝试的最大值;步骤序列取实际执行步骤的最大值。
1677
+ 启动器退出后仍等待组内全部后代结束;整组进程共享命令的 `timeout`,超过期限
1678
+ 会终止整组并报告超时,不使用固定的后代清理宽限时间。Linux 无 cgroup 权限时
1679
+ 自动降级为不测量(见下"限制与说明")。
1680
+
1681
+ **系统余量快照**(按平台):
1682
+
1683
+ - **Windows**:可用物理内存(`GlobalMemoryStatusEx`)与提交余量
1684
+ (`GetPerformanceInfo`)分别建模;
1685
+ - **Linux**:`/proc/meminfo` 的 `MemAvailable` 为基准;若处于有限额的
1686
+ cgroup v2 中,另对当前组及祖先组取 `memory.max − memory.current` 最小
1687
+ 余量,与 `MemAvailable` 取更小(**有效余量**)——避免在容器/限额环境
1688
+ 只看到宿主机内存;Linux 无系统级提交上限概念,有效余量同时充当提交
1689
+ 余量角色,总量钳制到 `min(MemTotal, 最小组限额)`。
1690
+
1691
+ **准入判据**(全部满足才启动,否则排队等待并周期重评):
1692
+
1693
+ 1. CPU 令牌充足;
1694
+ 2. 仅在峰值可能拥挤时要求距上次授予 ≥ 最小启动间隔(默认 1s,规则见下);
1695
+ 3. 有效余量 ≥ 新任务估算需求 + 保留量(默认 2048 MB;Windows 为提交余量,
1696
+ Linux 为有效余量,同一配置键);
1697
+ 4. 可用物理内存 ≥ 物理低水位(默认 2048 MB,独立压力检查,与任务需求无关)。
1698
+
1699
+ 提交与物理两种口径分开检查、互不混用(Linux 单一预算下两项检查同源,保留
1700
+ 各自语义)。**不做完整峰值预留**:已占用内存自然体现在系统余量中,系统余量
1701
+ 只需覆盖新任务自身需求——避免旧策略把已分配内存重复计入而把并发压到 1。
1702
+ 内存压力大时只暂停启动新任务,已运行任务继续执行。
1703
+ 每次拒绝的具体原因(需求/保留量/余量/水位数值)记入调度日志,便于归因。
1704
+ 等待准入的用例留在调度队列,后续满足准入条件且依赖已完成的用例可以先运行。
1705
+
1706
+ **需求估算优先级**:
1707
+
1708
+ 1. 显式声明的 `min_memory_mb`(用户判断优先);
1709
+ 2. history 中的历史最大峰值 × 安全系数(默认 1.2)。**按来源隔离读取**:
1710
+ Windows 读提交峰值(`max_peak_memory_bytes`),Linux 读 cgroup 峰值
1711
+ (`max_peak_memory_bytes_linux`),不跨平台混用;
1712
+ 3. 两者皆无(首次运行):使用可配置的默认需求(默认 4096 MB)——不再整池
1713
+ 独占;首次估算不准的风险接受,运行后更新历史。
1714
+
1715
+ **准入参数配置**(`symtest.json` 的 `run` 段;CLI 第一版不提供旗标):
1716
+
1717
+ | 配置键 | 类型 | 默认值 | 说明 |
1718
+ |---|---|---|---|
1719
+ | `memory_margin` | float | 1.2 | 历史峰值估算需求的安全系数 |
1720
+ | `memory_commit_reserve_mb` | int | 2048 | 余量保留(MB),准入时需额外覆盖(Windows 为提交余量保留,Linux 为有效余量保留) |
1721
+ | `memory_physical_watermark_mb` | int | 2048 | 物理内存低水位(MB),低于该值暂停新任务 |
1722
+ | `memory_default_demand_mb` | int | 4096 | 需求未知 case 的默认估算需求(MB) |
1723
+ | `memory_min_start_interval_s` | float | 1.0 | 峰值可能拥挤时的最小启动间隔(秒),0 关闭 |
1724
+
1725
+ 仅在存在活动任务且“活动任务估算峰值之和 + 新任务需求 + 提交保留量”大于当前提交余量时,才要求距上次授予达到启动间隔;余量充足或活动任务全部结束时立即放行。该估算可能重复计入已占用内存,只用于保守地触发间隔,不作为硬预留。快照或需求未知且存在活动任务时保守保留间隔。
1726
+
1727
+ **限制与说明**:
1728
+ - 内存不足持续等待(这正是防止 OOM 失败的机制),等待期间每分钟输出一次心跳日志(含拒绝原因);
1729
+ - 物理低水位校验:已知总物理内存时,非零水位必须严格小于总量,否则初始化抛出 `ValueError`,避免永久排队。请降低 `memory_physical_watermark_mb`,或设为 0 禁用该检查;
1730
+ - 估算需求加保留量超过系统提交上限(如 pagefile 过小 / cgroup 限额过小,结构性不可满足)时,警告并跳过该项检查放行(物理水位与启动间隔仍生效),交由用例自身失败暴露问题;
1731
+ - 进程模式(`--execution-mode process`)暂不接入内存调度与峰值测量;
1732
+ - Linux 无法创建测量子组、内核无 `memory.peak` 或命令进程迁移失败时,命令以普通子进程执行一次,整组峰值未知。准入仍按手动/默认需求执行。不设 `memory.max` 硬上限、不主动杀任务。
1733
+ - 测量失败自动回退普通执行路径,峰值标记为未知,不影响测试判定。
1734
+
1628
1735
  ## 历史记录与回归检测
1629
1736
 
1630
- 通过 `--history-dir` 指定一个目录,框架会在该目录下维护一个 `.symtest` 文件,记录每个 case 的历史运行时间。
1737
+ 通过 `--history-dir` 指定一个目录,框架会在该目录下维护 `history.json`,记录每个 case 的历史信息。
1631
1738
 
1632
1739
  ### 工作原理
1633
1740
 
1634
- 1. **首次运行**:目录下没有 `.symtest`,自动创建空文件,排序仍使用配置中的 `estimated_time`
1635
- 2. **后续运行**:读取 `.symtest` 中的历史数据,优先使用 `avg_duration` 做调度排序
1636
- 3. **回归检测**:每次运行结束后,如果某 case 耗时超过历史均值的阈值倍数(默认 1.5),打印警告
1741
+ 1. **首次运行**:自动创建空 `history.json`,按配置 `estimated_time` 排序。线程内存感知模式下,缺少显式需求和本平台峰值历史的用例只采集内存,不计入耗时历史。
1742
+ 2. **后续运行**:有了内存峰值后采集通过耗时,下一轮可按 `avg_duration` 调度;内存准入使用本平台历史最大峰值。只有内存的记录回退配置 `estimated_time`。
1743
+ 3. **回归检测**:至少有 3 个先前有效通过耗时样本后,耗时超过历史均值的阈值倍数(默认 1.5)才打印警告。
1744
+ 4. **内存峰值**:耗时只记录通过的 case,内存峰值只要有有效测量就记录(含失败/超时的 case——这些数据同样保护下一次调度决策)
1637
1745
 
1638
1746
  ### CLI 用法
1639
1747
 
@@ -1675,7 +1783,14 @@ success = runner.run_tests()
1675
1783
  "case_name_1": {
1676
1784
  "avg_duration": 3.5,
1677
1785
  "last_duration": 3.2,
1678
- "run_count": 5
1786
+ "run_count": 5,
1787
+ "last_peak_memory_bytes": 5500000000,
1788
+ "max_peak_memory_bytes": 6100000000,
1789
+ "peak_memory_count": 3,
1790
+ "peak_memory_source": "windows_job_commit",
1791
+ "last_peak_memory_bytes_linux": 5100000000,
1792
+ "max_peak_memory_bytes_linux": 5300000000,
1793
+ "peak_memory_count_linux": 2
1679
1794
  }
1680
1795
  }
1681
1796
  }
@@ -1685,7 +1800,56 @@ success = runner.run_tests()
1685
1800
  |---|---|
1686
1801
  | `avg_duration` | 累计平均耗时(秒),用于调度排序和回归基线 |
1687
1802
  | `last_duration` | 最近一次运行耗时 |
1688
- | `run_count` | 历史运行次数 |
1803
+ | `run_count` | 有效通过耗时样本数,不含内存预热 |
1804
+ | `avg_cpu_time` / `last_cpu_time` | Windows Job 平均 / 最近一次 CPU 秒数(可选);`avg_cpu_time` / `cpu_run_count` 在本次带 CPU 测量时用于 Windows 回归告警 |
1805
+ | `cpu_run_count` / `cpu_time_source` | 有效 CPU 样本数 / `windows_job`(可选) |
1806
+ | `last_peak_memory_bytes` | 最近一次进程组内存峰值(字节,Windows Job Object 提交内存口径;Windows 内存管理启用时记录,含失败用例) |
1807
+ | `max_peak_memory_bytes` | 历史最大提交内存峰值,Windows 内存调度估算使用该值 |
1808
+ | `peak_memory_count` | 有效峰值测量次数(Windows) |
1809
+ | `peak_memory_source` | 测量来源(`windows_job_commit` = Windows Job Object 提交内存峰值) |
1810
+ | `last_peak_memory_bytes_linux` | 最近一次进程组内存峰值(字节,Linux cgroup v2 记账口径,含文件缓存;Linux 且有 cgroup 委派权限时记录) |
1811
+ | `max_peak_memory_bytes_linux` | 历史最大 cgroup 内存峰值,Linux 内存调度估算使用该值 |
1812
+ | `peak_memory_count_linux` | 有效峰值测量次数(Linux) |
1813
+
1814
+ 内存字段为可选且**按来源隔离**:Windows 写提交峰值字段,Linux 写 `*_linux`
1815
+ 字段,两套口径数值不可比、互不覆盖——同一份 history 在两平台先后使用时,
1816
+ 各平台只读自己的字段,不会把另一平台的峰值当作本平台需求估算。未启用
1817
+ 测量或测量失败时不产生对应字段;旧格式 history 无需迁移即可继续使用。
1818
+
1819
+ ### 内存预热、样本可信度与 Windows CPU 时间
1820
+
1821
+ 线程模式启用内存感知调度时,如果用例没有显式 `min_memory_mb`,也没有
1822
+ 本平台先前记录的有效内存峰值,本次只学习内存,不计入墙钟 / CPU 耗时历史;
1823
+ 本次结果中的实际耗时仍保留。规则按用例执行,新增用例也需要预热。
1824
+ 测量持续失败时,该用例持续预热,直到获得峰值或显式提供 `min_memory_mb`。
1825
+ 其他执行模式仍正常记录通过用例的耗时。
1826
+
1827
+ 例外:Linux 上 cgroup v2 不可用(无委派权限,如 CI 托管 runner)时,平台
1828
+ 永远学不到整组峰值,此时跳过预热、首轮即记录耗时——准入仍按默认需求估算
1829
+ 工作,调度排序与回归告警不受影响。
1830
+
1831
+ 典型流程:第 1 轮学习内存,第 2 轮开始采集耗时,第 3 轮即可按历史耗时
1832
+ 调度。只有内存的历史条目按配置 `estimated_time` 排序。告警则固定要求至少
1833
+ **3 个先前的有效通过耗时样本**,所以第 2–4 轮积累样本,第 5 轮最早可以
1834
+ 报警;仍使用默认 1.5 倍阈值。调度不需要等满 3 个样本。
1835
+
1836
+ Windows Job 测量会在结果 JSON / Python 结果字典中增加可选 `cpu_time`
1837
+ (秒)和 `cpu_time_source: "windows_job"`,表示命令及其全部后代(含已退出
1838
+ 进程)的用户态 + 内核态 CPU 时间。重试和多步骤求和,不包含 SymTest 自身
1839
+ 验证 / 文件比较所消耗的 CPU;多线程命令的 CPU 时间可能大于墙钟时间。
1840
+ 真实的零值有效,测量不可用则省略字段;任一重试 / 步骤缺少测量,或序列
1841
+ 恢复了缓存步骤时,省略总 CPU 时间,避免把局部测量当成完整值。
1842
+ 失败执行也可在结果中保留已测 CPU 时间;通过且非预热的用例另外在 history
1843
+ 记录 `avg_cpu_time`、`last_cpu_time`、`cpu_run_count`、`cpu_time_source`。
1844
+ Linux 和普通 subprocess 不产生此 Windows 指标。CPU 时间不替换调度、超时或
1845
+ JUnit 耗时;**性能回归告警按度量来源选口径**——本次带 CPU 测量(Windows)
1846
+ 时只用 CPU 均值 / 样本数比较(同样要求至少 3 个先前 CPU 样本,样本不足时
1847
+ 本次不告警、不回退墙钟,避免并行负载下的墙钟误报),无 CPU 测量(Linux /
1848
+ 测量失败)退化为墙钟口径。
1849
+
1850
+ 这些规则排除冷内存样本和过早告警,不能消除任意 CPU / I/O 争用。
1851
+ 旧历史仍兼容,其耗时样本继续计数;需要替换旧的混合条件样本时,保留已学
1852
+ 内存历史运行一次 `--update-history` 即可。
1689
1853
 
1690
1854
  ### 回归警告示例
1691
1855
 
@@ -1695,6 +1859,13 @@ success = runner.run_tests()
1695
1859
  ⚠ WARNING: Case 'heavy_simulation' regressed: 18.2s vs avg 10.5s (1.73x slower)
1696
1860
  ```
1697
1861
 
1862
+ Windows 上带 CPU 测量的运行使用 CPU 口径(消息标注 `CPU`,多线程命令的
1863
+ CPU 时间可大于墙钟):
1864
+
1865
+ ```
1866
+ ⚠ WARNING: Case 'heavy_simulation' regressed: 26.4s CPU vs avg 12.1s CPU (2.18x slower)
1867
+ ```
1868
+
1698
1869
  ### 不启用历史记录
1699
1870
 
1700
1871
  不传 `--history-dir` 时行为与之前完全一致,不创建任何额外文件。
@@ -1714,7 +1885,7 @@ symtest run config.json --history-dir ./hist --update-history
1714
1885
  - 清除 `.symtest` 中**本次运行涉及的 case** 的历史条目
1715
1886
  - 未参与本次运行的 case 的历史数据保留不动
1716
1887
  - 本次运行的回归检测不会产生误报(无历史基线可比)
1717
- - 本次通过 case 的耗时会被记录为全新起点
1888
+ - 本次通过且非预热的 case 从头积累耗时;是否预热按运行前的内存信息判定,已有峰值仍允许采样。3 个先前样本的告警门槛重新计数
1718
1889
  - 报告中显示 `History Reset` 计数
1719
1890
 
1720
1891
  > **注意**:需搭配 `--history-dir` 使用。清零范围仅限本次运行的 case,不影响其他 case 的历史数据。