[stevevanhooser]

The general test job (run_tests.yml → tools/tasks/testToolbox.m → matbox.tasks.testToolbox) collects every test under tools/tests, so it also ran the cross-language symmetry tests. Those already run in the symmetry workflows: NDR-matlab's test-symmetry.yml and NDR-python's cross-language one. There they get the artifacts they read and run in order.

In the general job they only repeated that work. Python never runs here, so since #145 every pythonArtifacts read case shows up as skipped: a standing set of 7 skips on every run.

Changes

• All 14 symmetry test classes (7 make, 7 read) are now tagged: classdef (TestTags = {'Symmetry'}) ….
• tools/tasks/testToolbox.m passes ExcludeTags="Symmetry". MatBox's testToolbox has no folder exclusion, but it does support ExcludeTags. The action calls this repo-local wrapper when it exists, so run_tests.yml needs no change.
• New tools/tests/+ndr/+unittest/TestSymmetryTestsAreTagged.m: fails if any test under ndr.symmetry lacks the tag. The exclusion is only as good as the tags, so a new symmetry test can't quietly slip back into the general job.
• Both symmetry INSTRUCTIONS.md files: a rule to tag new classes, and the readArtifacts example now shows the tag.
• New codecov.yml: the project check allows a coverage drop of up to 3% against the base, instead of Codecov's default of failing on any drop. The patch check is unchanged (target: auto). See Coverage below.

The symmetry workflows collect tests with TestSuite.fromPackage and don't filter by tag, so they're unaffected.

Result

• General job: 395 tests ran, 0 failed (before: 415). It removed exactly the 21 symmetry test cases (7 make + 7 read × 2 sources) and added the tag guard, which passed. Skips went from 10 to 3.
• Coverage: 54.94% → 52.21% (−2.73%, 262 lines). Those 262 lines of +ndr reader/format code are reached only by the symmetry tests, and no unit test covers them. With no codecov.yml, codecov/project failed on that drop. The maintainer chose to allow drops of up to 3% (option 2 in the discussion below), so this drop passes and a larger one still fails. Unit tests for those lines would be the lasting fix.

Verification

I wrote this without access to MATLAB, so CI is its first run. The test syntax (TestTags class attribute, Test.Tags) and the ExcludeTags option are taken from MatBox's own testToolbox.m source. I couldn't reach Codecov's validator from my environment, so this PR's codecov/project check is what confirms codecov.yml.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QjkvVXsdCVw6NNe5SnGNyX

[github-actions]

Test Results

395 tests   - 20   392 ✅  - 13   36s ⏱️ +10s
 41 suites  - 13     3 💤  -  7 
  1 files   ± 0     0 ❌ ± 0 

Results for commit af078b9. ± Comparison against base commit e2e96be.

This pull request removes 21 and adds 1 tests. Note that renamed tests count towards both.

ndr.symmetry.makeArtifacts.format.readOmeZarr ‑ testReadOmeZarrArtifacts
ndr.symmetry.makeArtifacts.format.readSpikeGadgets ‑ testReadSpikeGadgetsArtifacts
ndr.symmetry.makeArtifacts.reader.readData ‑ testReadDataArtifacts
ndr.symmetry.makeArtifacts.reader.readPrairieview ‑ testReadPrairieviewArtifacts
ndr.symmetry.makeArtifacts.reader.readTiffstack ‑ testReadTiffstackArtifacts
ndr.symmetry.makeArtifacts.reader.readVideo ‑ testReadVideoArtifacts
ndr.symmetry.makeArtifacts.reader.readVld ‑ testReadVldArtifacts
ndr.symmetry.readArtifacts.format.readOmeZarr ‑ testReadOmeZarrArtifacts(SourceType=matlabArtifacts)
ndr.symmetry.readArtifacts.format.readOmeZarr ‑ testReadOmeZarrArtifacts(SourceType=pythonArtifacts)
ndr.symmetry.readArtifacts.format.readSpikeGadgets ‑ testReadSpikeGadgetsArtifacts(SourceType=matlabArtifacts)
…

ndr.unittest.TestSymmetryTestsAreTagged ‑ testEverySymmetryTestIsTagged

♻️ This comment has been updated with latest results.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.21%. Comparing base (e2e96be) to head (af078b9).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #146      +/-   ##
==========================================
- Coverage   54.94%   52.21%   -2.73%     
==========================================
  Files         164      164              
  Lines        9598     9598              
==========================================
- Hits         5274     5012     -262     
- Misses       4324     4586     +262     

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

🚀 New features to boost your workflow:

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

[stevevanhooser]

codecov/project is red. The tests are fine. The red reflects coverage this job no longer measures, so it needs a maintainer decision.

• Tests: 395 ran, 0 failed. Before this PR: 415. The difference is 21 symmetry test cases removed (7 make + 7 read × 2 sources) and the new tag guard added. Skips went from 10 to 3, so the 7 pythonArtifacts skips are gone, as intended.
• Coverage: 54.94% → 52.21% (−2.73%, 262 lines). Those 262 lines of +ndr reader/format code are exercised only by the symmetry tests, and no unit test covers them. Codecov's default (there's no codecov.yml) fails on any drop.

Options:

1. Accept the drop: merge with codecov/project red. Coverage then reports only what the unit tests reach.
2. Add a codecov.yml with a project threshold (for example coverage.status.project.default.threshold: 3%). Simple, but it also hides future real drops of up to that size.
3. Upload coverage from the symmetry workflow too: run its suites with CodeCoveragePlugin and upload with a Codecov flag, so the combined coverage counts again. This keeps the measurement but is more CI work, and the status can flap while uploads land.
4. Write unit tests for those 262 lines: the durable fix, but a separate piece of work.

I haven't changed the threshold, since that would quietly weaken a check. Tell me which option you want and I'll do it.

---

Generated by Claude Code

[stevevanhooser]

Went with option 2: af078b9 adds a codecov.yml allowing a project coverage drop of up to 3% (this one is 2.73%). Larger drops still fail, and the patch check is unchanged.

---

Generated by Claude Code