[samtrion]

Problem

stryker-config.json has "thresholds": { "high": 80, "low": 60, "break": 0 }. The scheduled mutation run (run 40488130044) logged The final mutation score is 31.98 %, far below low 60, and still concluded success, despite 504 survivors and 8348 NoCoverage mutants. A mutation score regression can never be noticed from the workflow status.

Root cause

break = 0 disables the gate. (high and low only colour the report; only break makes Stryker exit non-zero.) The current low score is, at least partly, a separate problem (duplicate Roslyn variants and test projects that "did not report any test", 8 log lines); the share of each is inferred from the logs, not measured.

Change

break: 0 -> break: 25. low stays 60, so break <= low holds (Stryker rejects break > low with an InputException; any higher break would require raising low/high in the same change).

Provisional, blocked on the variant fix. The baseline will change once the duplicate-Roslyn-variant fix lands, so this PR is a draft. Before merging: re-measure with a dispatch run on the new baseline, then set break about 5 points below that observed score (and raise low/high if needed). 25 is only "a few points below today's measured 32 %".

Before / After

metric	before	after
thresholds.break	0	25 (provisional)
score	31.98 % (run 40488130044)	32.02 % on probe run 40805541569 (same code, only thresholds differ); score is stable across runs
conclusion at ~32 % with break=0	success	not measured with break=25 (see probe)
gate propagation	unproven	confirmed: throwaway branch ci/stryker-break-gate-probe with high=low=break=100, run https://github.com/dailydevops/frameshift/actions/runs/40805541569 logged Final mutation score is below threshold break. Crashing... and the job concluded failure

A dispatch run on this PR branch was not done: it would take about 2.5 h and, with a score of ~32 % vs break 25, would just be green; the gate itself was proven with the probe run. Mutants created / compile errors / NoCoverage are unchanged (18800 / - / 8348 in the probe run).

Risk / what this does not change

• Does not change the mutation score, mutants or test runtime.
• Once merged, any run with a score below 25 % fails the workflow (flaky coverage or timeouts could dip the score by a few points; 7 points margin).
• Should not be merged before the variant fix; the value then needs to be re-derived.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.69%. Comparing base (98b8eb0) to head (6cda4b0).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #297      +/-   ##
==========================================
+ Coverage   95.66%   95.69%   +0.02%     
==========================================
  Files         111      111              
  Lines        7031     7031              
  Branches     1427     1427              
==========================================
+ Hits         6726     6728       +2     
+ Misses        140      138       -2     
  Partials      165      165              

☔ 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.