Skip to content

Execution and failures ​

Discovery and order ​

flowstride run reads testDir (default flows). Supply a file or directory to narrow the target. Discovery includes nested folders. Files ending in .first.flow are prioritized for setup. An explicit handoff determines when a dependent file becomes eligible to run.

Reports preserve paths relative to the flow root: admin/login.flow and customer/login.flow remain separate tests and have distinct evidence names.

A sequential run uses one worker. Cloud can distribute files across workers, subject to workspace and machine limits. Use a named handoff when a later file needs data from a successful earlier file. FS releases the consumer after the producer passes and marks it blocked if the producer exhausts its retries. Process-global values belong to one worker.

Local extracted values, generated-value memory and unsaved local sessions reset between files, not between every scenario. Named browser profiles are also file-local. Use independent data and distinct names when scenarios or users share a file. Variables and profiles explain these boundaries.

What happens on failure? ​

A required step failure marks its attempt failed and leaves that attempt's remaining steps untested. A soft assertion records the failure and allows subsequent steps in the same attempt to run. If retries is configured, Flowstride reruns the failed scenario before proceeding. After the final failed attempt, the runner continues to later scenarios by default. --bail stops later work after that final failure. Cloud uses sequential execution for bail runs.

sh
npx flowstride run flows --ci --bail

flow.try makes an action optional. Reserve it for an optional banner or similar condition; using it around a required assertion can hide a defect.

Registered cleanup runs after each attempt. A conditional skip records that a scenario was deliberately skipped before actions started.

Focus and debug stops ​

only.Scenario: focuses one scenario within that file. Only one focus marker is allowed per file. Other files remain eligible. stop.When and other stop. blocks halt after the first successfully executed command in that block, even when the block contains further commands. They are debugging controls. Remove focus and stop markers before a complete acceptance run; CI records incomplete execution as a nonzero result.

Use scenario tags to select part of a suite and flow.pause to resume manually in the dashboard during a local run.

Interactive and CI lifecycles ​

Interactive execution keeps the local dashboard server open for inspection after the suite ends. Stop it with Ctrl+C when finished.

Native CI forces headless mode, writes summary.json and junit.xml, and exits. Failed, incomplete, or crashed execution returns nonzero. SIGINT/SIGTERM are recorded as interrupted runs. A configured local webServer is owned and stopped by the run.

Test outcome and Cloud reporting outcome are separate. A passing test with a failed upload remains a passing test with a reporting problem. Inspect both fields in the CI artifacts.

Explicit retries and distributed runs ​

Retries are opt in and preserve earlier attempt diagnostics. A later success is reported as flaky; external application changes are not rolled back. Native browser workflows explains retry configuration, independent users, matrices, file sharding and complete result merging.

Read a command failure ​

The terminal and UI retain the command, relevant target/resource, source file/line and the original error/call log where available. Native assertions also retain target and condition checks with expected/actual values. Later unexecuted work is not labelled successful. See assertions and diagnostics for examples across browser, API, sessions, plugins, games, audio and OTP.

Flowstride is MIT licensed. Flowstride Cloud offers free and paid workspace plans.