Repository navigation
Fix Makefile and documentation paths left behind by moved files - #4166
DanielMBouyou wants to merge 1 commit into
Conversation
Five references still point at files that were moved or renamed by merged pull requests, and nothing in CI notices, because no test asserts that a path named in the build or the docs resolves. Makefile, test-fuzz: invoked tests/cross_fw/sdl/fuzz/quantize_api.py, removed in openvinotoolkit#3695, which added fuzz_target.py in the same commit. The filename is the only change; both scripts call atheris.Setup(sys.argv, ...) and then atheris.Fuzz(), so argument handling and behaviour are unchanged. This target is the only invocation of the fuzz target in the repository, so no CI job caught the breakage. Makefile, install-openvino-dev: installed from examples/post_training_quantization/openvino/yolov8/, renamed to yolo26 in openvinotoolkit#3844. docs/usage/IgnoredScope.md, tests/post_training/README.md and the --memory-monitor help text in tests/post_training/conftest.py: tools/ paths moved in openvinotoolkit#3939. The documented render_dot_to_svg command also passed -m, which the script does not register; its parser accepts -i and --input_file. Adds tests/docs/test_makefile_paths.py, asserting that Python scripts written out literally in Makefile recipes exist. It is deliberately narrow: it does not cover requirements files, paths built from Make variables, or python -m invocations, so a green run is not a claim that every target works. install-models-hub-torch is left alone. It installs from tests/torch/models_hub_test/, deleted in openvinotoolkit#3553, and the only repair is removing the target, which is a maintainer's decision.
github-actions
Bot
added
documentation
|
Bonjour gentle bump on this one. The CI never actually ran: as this is my first PR to nncf, the workflows needed a maintainer to approve them, and the pending runs expired after 30 days. I have just reopened the PR to re-queue them. Could a maintainer approve the run and take a look when you have a moment? It is a small cleanup that fixes stale script paths in the Makefile and a few docs left behind by #3695 and #3939, with a narrow test that guards against the same drift. Happy to rebase onto develop if that helps. Thanks! |
|
@DanielMBouyou The main problem is that the Makefile was not being used as intended and has become outdated. (it was fixed in #4235) |
Changes
Repoint five references left pointing at files that were renamed or removed by merged refactors:
Makefile,test-fuzz— invokedquantize_api.py; Rework fuzzing tests #3695 removed it and addedfuzz_target.pyin the same commit. Filename only: both callatheris.Setup(sys.argv, ...), behaviour unchanged.Makefile,install-openvino-dev— installs fromopenvino/yolov8/, renamedyolo26in [OpenVINO] YOLO26 sample #3844.docs/usage/IgnoredScope.md,tests/post_training/README.md, a--memory-monitorhelp string —tools/paths moved in Reorganization tools directory #3939.render_dot_to_svg.pytakes-i, not-m.Adds
tests/docs/test_makefile_paths.py: scripts named in Makefile recipes must exist.install-models-hub-torchinstalls fromtests/torch/models_hub_test/, deleted in #3553; removing a target seemed like your call.Reason for changes
make test-fuzzfails on its only line;make install-openvino-devon its third, a requirements file that no longer exists. No CI job runs the fuzz target, so nothing caught it.Related tickets
None; no issue or PR covers these paths.
Tests
Test fails on
developnamingMakefile:166, passes after.pytest tests/docsgreen (14),tests/commonunchanged, Ruff v0.15.18 check and format clean. Thememory_monitoredits are text, checked by inspection.