Skip to content

CHANGELOG.md

Latest commit

 

History

History
77 lines (54 loc) · 8.74 KB

CHANGELOG.md

File metadata and controls

77 lines (54 loc) · 8.74 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

0.4.0 - 2026-07-20

Changed

  • deploytasks:rollup now words an empty --group selection the same way as deploytasks:run ("No tasks matched the requested group(s)." instead of "No task slots matched the requested group(s)."), and deploytasks:skip deduplicates a repeated --group value in its error messages like the other commands already did.
  • deploytasks:host:install now scaffolds the configured host.directory instead of always creating deploy/host-tasks/ — a host with a custom directory no longer ends up with a task directory the other commands never look at. The default is unchanged.

Fixed

  • A run that lost its lock lease mid-run no longer reports "Run skipped: another process is already running." — which hid the tasks that did execute. It now reports the early stop and how many tasks ran before it (RunResult gained a leaseLost flag alongside locked).

Security

  • Filesystem storage now refuses to operate on a directory containing a composer.json — a storage path misconfigured to the project root could otherwise let deploytasks:reset/rollup delete project files (composer.json, package.json, … match the record-name pattern). A dedicated state directory never contains one.
  • deploytasks:reset now strips terminal control bytes from a rejected --group value before echoing it back, so a malicious group name can no longer inject escape sequences into the operator's console.
  • Filesystem storage no longer recognizes record files whose name ends in a newline — the record-name pattern is anchored to the absolute end of the filename (\z), matching the trailing-newline hardening already applied to task ids and group names.

0.3.0 - 2026-07-19

Changed

  • Renamed the container-scope generator command from deploytasks:generate:container to deploytasks:generate. The bare deploytasks:<verb> form now consistently means the container scope (matching run, reset, rollup, skip, …), while the host scope keeps its deploytasks:host:<verb> prefix. The old name no longer exists — update any deploy scripts, Makefiles, or CI that call it. See UPGRADE.md.

0.2.0 - 2026-07-17

Added

  • Symfony 8 support — the bundle now installs on Symfony 8.x (requires PHP 8.4+ there); Symfony 6.4 LTS and 7.x remain supported unchanged.

0.1.0 - 2026-07-17

First release of DeployTasksBundle — a Symfony bundle for running one-time deploy tasks (data migrations, cache warmups, seed scripts) from the console, tracking each so it executes exactly once per environment.

Added

  • Task contract — implement DeployTaskInterface (getDescription(), run(OutputInterface): TaskResult) and declare the task with #[AsDeployTask] (id, priority, env, timeout, slowTaskThreshold, transactional, description, groups); tasks are discovered and tagged automatically. run() returns SUCCESS, FAILURE, or SKIPPED.
  • Task runner — deploytasks:run executes pending tasks in a deterministic order (priority, then the date embedded in the task id, then registration order), with dry-run, single-task (--id), re-run-all, and group targeting, plus a live [i/N] progress line and per-task durations. A task returning SKIPPED (preconditions not met) is deferred: its slot stays pending and it is retried on the next deploy.
  • Storage backends — filesystem (default; one JSON record per slot), Doctrine DBAL (SQLite/MySQL/MariaDB/PostgreSQL, composite (id, group) key, optional table auto-creation), and in-memory (testing). Plug in any backend through TaskStorageInterface / TransactionalStorageInterface; opt into DDL provisioning with SchemaManageableInterface.
  • Transactions — a per-backend transaction_mode (none, per_task, all_or_nothing) wraps task execution in database transactions, up to an all-or-nothing whole-run transaction that rolls back every side effect on failure; individual tasks opt out of per_task wrapping via #[AsDeployTask(transactional: false)].
  • Execution durations — every run records how long the task took (milliseconds), shown by deploytasks:status and deploytasks:show.
  • Concurrency protection — optional run lock via symfony/lock, refreshed between tasks so long deploys keep the lease; degrades to a no-op when the component is absent.
  • Task groups — tasks declare one or more groups via #[AsDeployTask(groups: …)] to stage a deploy (e.g. predeploy / postdeploy); a multi-group task records one slot per group, and every command accepts a repeatable --group filter.
  • Lifecycle events — BeforeTaskEvent, AfterTaskEvent, TaskFailedEvent (optional; requires symfony/event-dispatcher).
  • PSR-3 logging — run- and task-level lifecycle logged through the application logger, auto-routed to a dedicated soviann_deploy_tasks Monolog channel when available and falling back to a NullLogger otherwise.
  • Console commands — fourteen deploytasks:* commands covering execution (run), inspection (status, show), state operations (skip, reset, rollup), scaffolding (generate:container), schema management (create-schema), and the host-scope suite (host:install, host:generate, host:skip, host:reset, host:rollup, host:config).
  • Host-scope tasks — a self-contained bin/deploy-tasks-host.sh runner executes tracked *.sh tasks on the host, outside the container, with a Symfony-compatible .env cascade and flock-based concurrency; deploytasks:host:install scaffolds the runner, the task directory, and the .gitignore block in one idempotent step.
  • Process helper — ProcessRunnerTrait wraps symfony/process for tasks that shell out, streaming sanitized child output and enforcing a per-call timeout.
  • Extension points — per-instance task ids (TaskIdProviderInterface) and custom ordering (TaskSorterInterface).
  • Configuration — the soviann_deploy_tasks config tree covers storage, events, lock, generation, and host settings, with scalar shorthands (storage: database, events: false, lock: false) and a tunable slow-task warning threshold (global slow_task_threshold plus a per-task attribute override).
  • Requirements — PHP 8.2+, Symfony 6.4 LTS or 7.x.

Security

  • Storage path guards — filesystem storage refuses to store execution records under a web-served document root, and generated-file paths (deploytasks:generate:container, deploytasks:host:generate, --dir) are canonicalized with symlinks resolved and validated against path traversal before anything is written.
  • Private, atomic records — filesystem records are written atomically with 0700 directories and 0600 files, so a crashed save cannot leave a partial record and other local users cannot read deploy history.
  • Console output sanitization — untrusted text rendered by the commands (task and process error output) is scrubbed of ANSI/C0/C1 control sequences, Unicode bidi and format characters (Trojan-Source-style reordering), and console formatter tags through a single ConsoleSanitizer pipeline; ProcessRunnerTrait applies the same scrub to streamed child output.
  • Destructive-command gates — deploytasks:reset, deploytasks:rollup, and their host-scope equivalents prompt for confirmation and require an explicit --force when run non-interactively; deploytasks:skip warns before overwriting an existing execution record.
  • Strict identifier validation — task ids and group names are validated against a strict allowlist (exact-line anchoring, no trailing-newline smuggling), rejected when they exceed the configured database column length (preventing silent truncation and task re-runs), and checked for case-insensitive collisions at container build or boot.
  • Credential-safe logging — DBAL failure context is scrubbed of full exception objects before it reaches any log handler, so database credentials cannot leak into shared log sinks.
  • Lock hardening — the run lock's lease is refreshed between tasks, a mid-run lock failure aborts the run cleanly instead of crashing it, and the TTL semantics are documented so operators size it against the longest single task.