Outputs

Every campaign writes into a per-design workspace, output/<top>/. The user-facing deliverables live at the top of that directory; transient internal state lives in output/<top>/.faultflow/ and is what --clean removes.

Directory layout

output/<top>/
├── coverage.rpt           # human-readable coverage report
├── rule_check.rpt         # DFT rule-check report (rule_check command)
├── report.rpt             # unified report (shell `report` command)
├── <top>_scan.v           # scan deliverable (after scan + techmap)
├── <top>_scan.json        # scanned generic JSON
├── scan.rpt               # scan-insertion report
└── .faultflow/            # internal workspace (removed by --clean)
    ├── faultflow.sqlite    # campaign database (combinational + scan)
    ├── logs/               # Yosys / nl2bench / quaigh logs
    ├── manifests/
    ├── intermediate/       # <top>.json, <top>_gate.v, coverage_report.json
    ├── verification/       # iverilog verification reports
    └── generated_scripts/  # rendered yosys_synth.tcl

In OpenTestability oracle mode (run), an oracle_response.json is written to the output root.

coverage.rpt

A plain-text summary of the campaign: the coverage headline, the fault classification counts, the policy in effect, and per-node detail. The file name is set by [report] output. (The interactive shell additionally writes report.rpt, a unified report with a scan-focused headline; the batch CLI flow produces coverage.rpt.)

coverage_report.json

The machine-readable report, written to output/<top>/.faultflow/intermediate/coverage_report.json and validated against schemas/coverage.schema.json. Its top-level structure is:

Section	Contents
metadata	Tool/run identity and input fingerprints
policy	The active policies (collapsing, clock/reset inclusion, unsupported-cell mode)
summary	The coverage and fault-count totals (see below)
run	Per-run statistics, including ATPG round counts and timing
per_node	Per-net fault counts and detection status
undetected_faults	The remaining undetected/redundant faults

The summary block

Field	Meaning
total_raw_faults	Every enumerated fault, including all tagged exclusions
structural_eligible	Faults eligible by structure: supported cell, not collapsed, not clock/reset (the structural denominator)
denominator	The effective (testable) denominator: structural_eligible minus proven-redundant faults
detected	Faults detected by the test set
undetected	Faults neither detected nor proven redundant
redundant	Faults proven untestable (SAT UNSAT) under the current model
collapsed	Faults removed by collapsing
excluded_blackbox	Faults on blackboxed-cell outputs
excluded_clock	Clock-net faults (unless include_clock_faults)
excluded_reset	Reset-net faults (unless include_reset_faults)
excluded_scan, excluded_scan_internal, excluded_scan_chain	Scan-cell-internal and scan-path-only faults (scan campaigns)
test_coverage_percent	detected / denominator x 100 — credits proven-redundant faults by removing them from the denominator; the headline figure
fault_coverage_percent	detected / structural_eligible x 100 — counts proven-redundant faults against you; the conservative figure
coverage_percent	The headline coverage figure (equal to test_coverage_percent)

Note

total_raw_faults = denominator + redundant + collapsed + excluded_blackbox + excluded_clock + excluded_reset + excluded_scan + excluded_cross_domain. Exclusions are always tagged and counted, never silently dropped, and a zero denominator is raised as an error rather than producing a NaN.

patterns.test

Important

Despite being a documented config key ([atpg] output, default patterns.test), this file is not written by a normal sim / run_atpg run — combinational or scan. It is only ever written as an internal fallback inside scan_check’s vector-source resolution, when the optional Quaigh comparison path (quaigh atpg ... -o patterns.test) runs because no vectors were found any other way. In practice it is rarely present — don’t rely on it as a routine deliverable.

The reliable ways to get generated vectors out are the campaign database (faultflow.sqlite) and coverage_report.json below, or sim --scan --export-patterns PATH — a distinct, JSON-format mechanism (see Running without the shell). In that file a transition pattern carries launch: loc, a functional clock pulse between the load and the capture, or los, one more shift with scan enable on and each chain’s launch_scan_in bit at its scan input. External vectors supplied with sim --ext use the same plain-text format as patterns.test and require a same-stem .bench sidecar that fixes the primary-input order.

faultflow.sqlite

The campaign database under .faultflow/. It holds the fault list, per-fault detection records, generated vectors, run metadata, and timing — and is shared by the combinational and scan campaigns. status reads its current state; --clean deletes it.

Warning

The database uses the default (rollback) SQLite journal mode, not WAL. WAL is not reliable on a Windows drive mounted into WSL (/mnt/c/...), so do not switch it on when the workspace lives on such a path.

oracle_response.json

Produced by run (OpenTestability oracle mode). A flat JSON document — validated against schemas/oracle_response.schema.json — carrying backend, coverage_percent, fault_coverage_percent, vector_count, a terminal_reason (target_reached / exhausted / timeout), and the undetected_faults and per_node arrays. See External tools.