#Running BDD

Features run through one of two paths, both driven by the same TestRunner.

#ferridriver bdd (the CLI)

Primary path. Runs Gherkin features through the core test runner with Rust and / or JS / TS step bodies — no Node, no Bun:

# Rust steps only
ferridriver bdd tests/features/

# With JavaScript / TypeScript step files
ferridriver bdd \
  --steps 'steps/**/*.{js,ts}' \
  --tags '@smoke and not @wip' \
  --workers 4 \
  --reporter junit \
  --reporter terminal \
  tests/features/

--steps is repeatable and overrides [test].steps from the config file. Defaults to steps/**/*.{js,ts} and step_definitions/**/*.{js,ts} when omitted.

#Flags

--steps GLOB              JS / TS step files (repeatable)
--tags EXPR               tag filter: @smoke and not @wip, etc.
--workers N               parallel workers (default: CPU count)
--reporter NAME           reporter name (repeatable)
--strict                  treat undefined / pending steps as failures (default)
--no-strict               report undefined / pending steps without failing
--dry-run                 parse + report scenarios without executing
--fail-fast               stop after first failure
--step-timeout MS         override per-step timeout
--order MODE              defined | random | random:SEED
--language LANG           default Gherkin keyword language (en, de, fr, ...)
--shard X/N               run shard X of N across CI machines
--world-parameters JSON   pass JSON object to this.parameters

Plus the shared browser flags (--backend, --headless, --executable-path, --connect, --auto-connect, --user-data-dir).

#Rust / cargo test

Use bdd_main!() as the binary entry point and the shared runner flags:

// tests/bdd.rs
ferridriver_bdd::bdd_main!();

# Cargo.toml
[[test]]
name = "bdd"
path = "tests/bdd.rs"
harness = false

[dev-dependencies]
ferridriver-bdd  = "0.4"
ferridriver-test = "0.4"

cargo test --test bdd -- -j 4 --tags '@smoke and not @wip'

Tests run headed by default; pass --headless to opt into headless mode.

#Reporters

Specify multiple reporters by repeating --reporter, e.g. --reporter terminal --reporter junit. Each name is matched exactly (no name:path syntax). File-writing reporters emit into the run's output directory — junit writes junit.xml, json writes results.json. Set the path or per-reporter options with a [[test.reporter]] table in the config file. A name that is neither built in nor a path to a reporter module is refused with the list above; see custom reporters.

#The cucumber-json document

cucumber-json writes the document cucumber's own ecosystem reads, field for field: a feature's id, line, description and tags[] (each with the line it was written on), an element's id (feature;rule;scenario), keyword, line and description, and a step's line, arguments (a data table as {rows:[{cells}]}, a doc string as {content, line}) and result.duration in nanoseconds. Hooks appear as hidden steps keyworded Before or After.

Feature entries are keyed on the file AND the project, so a run over several projects produces one entry per project rather than merging them; the entry's id carries the project name.

There is no cucumber-html reporter. The HTML report is produced from this document by the tool that owns the format (cucumber-html-reporter, multiple-cucumber-html-reporter), which reads exactly the file cucumber-json writes.

#Environment variables

The shared runner env vars apply (see CLI reference — FERRIDRIVER_DEBUG, FERRIDRIVER_PROFILE, RUST_LOG, etc.).