You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Effective gate treats missing coverage totals as 0% and can pass without measurements #15
Found while reviewing the architecture documentation. Verified against upstream HEAD d99515b87bbe1dde89984403e9c330d640fc5d33 using synthetic report archives and the real generate_coverage_html CLI (no gate/justification mocks).
The effective coverage path converts missing/unparseable/zero HTML totals to a numeric 0% report. The gate treats this as an evaluated percentage. At the default threshold it exits 1; at threshold 0 it exits 0, although there are no usable measurements. The raw LCOV path rejects zero instrumented lines with exit 2.
Reproduction
Create a report zip containing html_report/index.html with no parseable LLVM Totals row, an empty html_report/style.css, and lcov_report/lcov.dat with LF:0 and LH:0. Supply a valid empty YAML:
version: 1justifications: []
Run the real generate_coverage_html CLI with --yaml:
COVERAGE_THRESHOLD=100: warning about unparseable totals, effective 0%, exit 1.
COVERAGE_THRESHOLD=0: same warning, effective 0%, exit 0.
Without YAML the zero-line LCOV yields exit 2.
Expected
Unavailable measurements must produce exit 2 for every threshold, including 0. Validate usable totals before constructing/accepting an effective gate value. Cover missing/malformed HTML totals and zero instrumented lines in both HTML backends.
This violates tool_req__coverage_gate_exit_codes and tool_req__coverage_gate_no_verdict.
Activity