COMP: Fix the Codecov wiring that was dropping three modules from the percentage

[RafaelPalomar]

What this fixes

The Codecov percentage on preview was being computed over a subset of the extension, and had become non-comparable across commits. Three independent faults, all silent — no warning, no failed job, just a wrong number.

1. Flag paths were dropping three whole modules

flags.<flag>.paths in .codecov.yml removes files from the report, it does not merely un-flag them. The list had fallen behind the module layout:

Entry	Reality
LiverMarkups/	module deleted in #585
Modeling/	never existed — ADR-0023 dropped it
LiverResectionsLib/	one level too high; it is LiverResections/LiverResectionsLib/
py flag -> **/Testing/Python/	discarded anyway by the ignore list below it
(absent)	Liver/, LiverSegmentation/, SlicerLiverInteractionLib/

So the shell (11 .py), Stage-2 segmentation (29 .py) and the shared interaction lib (13 .py) contributed zero to the reported figure, even though the coverage job was measuring the first two. Confirmed against the live report: only LiverResections, LiverVolumetry and VascularTerritories appear.

Flag membership now comes from the upload alone, which is the thing that already knows the answer.

2. Two flags, one session

A single codecov-action step with files: a,b + flags: cxx,py is one upload carrying both flags, so neither flag isolates its language. The tell is sessions: 1 on a repo with two flags — visible on every commit in the API. Now one upload step per flag.

3. Unmeasured and unmerged code

• SlicerLiverInteractionLib was missing from the --cov targets, so the base four modules share was never measured at all.
• LiverVolumetryLib, VascularTerritoriesLib and SlicerLiverInteractionLib are staged into qt-scripted-modules/ like LiverResectionsLib, but had no .coveragerc [paths] alias — their launched-leg records stayed on build-tree paths instead of merging onto the source files.

Why the trend line was lying

Faults 1 and 2 change the denominator, so the percentage moves on its own. The 56.29% -> 51.61% step on 2026-08-20 lands on an ADR-plus-workflow commit that adds no source, while the measured base jumped from 118 files / 12,478 lines to 122 / 14,524. Read as a test-quality regression, that would have been wrong.

For the record, the current 52.0% is still soft — it will move once the three missing modules enter the denominator, probably downward, and that will be the first honest number rather than a regression.

Scope

Configuration and documentation only. No test, source or build-logic changes. Coverage remains informational per ADR-0021 — this fixes what the number is computed over, not whether it gates.

ADR-0021 gains a "Configuration traps" section recording both faults, the staged-package [paths] obligation, and the reading rule that would have caught this sooner: when a coverage swing has no plausible cause in the diff, check the file and line totals before believing it.

Verification

• .codecov.yml, .coveragerc and ci.yml parse (yaml.safe_load, configparser).
• Workflow structure asserted: the coverage job now carries exactly two codecov-action steps, coverage-cxx.xml/cxx and coverage-py.xml/py.
• Every staged qt-scripted-modules/ Lib package now has a [paths] alias (checked against the CMakeLists.txt DESTINATION_DIR entries).
• Full pre-commit suite green, including Validate GitHub Workflows.

The real proof is the next post-merge run: expect sessions: 2 and the three modules present in the file list.

Reviewer's map

• .codecov.yml — load-bearing: the flags block. The comment explains why paths must not come back.
• .github/workflows/ci.yml — load-bearing: the upload step split. Mechanical: one added --cov line.
• .coveragerc — mechanical: three aliases matching the existing pattern.
• Docs/adr/0021-coverage-measurement.md — documentation of the above.

---

Authored with assistance from Claude (Anthropic Claude Opus 5).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 54.75%. Comparing base (4421370) to head (0d1a247).

Additional details and impacted files

@@             Coverage Diff             @@
##           preview     #629      +/-   ##
===========================================
+ Coverage    52.00%   54.75%   +2.75%     
===========================================
  Files          122      139      +17     
  Lines        14863    17449    +2586     
  Branches      2504     2519      +15     
===========================================
+ Hits          7729     9555    +1826     
- Misses        6746     7503     +757     
- Partials       388      391       +3     

Flag	Coverage Δ
cxx	54.75% <ø> (+2.75%)	⬆️
py	54.75% <ø> (+2.75%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 17 files with indirect coverage changes

🚀 New features to boost your workflow:

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