What's New

For the complete history, see the full CHANGELOG.md on GitHub.

Next

0.200.1-next

• Restore remote cache publication through Insights while retaining atomic cache generations.
• Expand the Terrabuild agent skill with current target-policy, cache, phase, batching, locking, recovery, reporting, and Insights guidance.

0.200.0-next

• Document that Insights reporting is mandatory once a run connects while remote artifact transfer remains a cache concern.
• Replace debug diagnostics atomically without allowing diagnostic I/O to mask build failures.
• Fail file-lock acquisition on filesystem errors while continuing to wait for genuine contention.
• Emit terminal build progress only after scheduler outcomes are finalized.
• Retry abandoned container cleanup after daemon errors and timeouts.
• Keep builds usable through remote cache read and publication failures.
• Recover interrupted workspace restores conservatively when their transaction index is unreadable.
• Exclude Git repository metadata from workspace project discovery by default.
• Validate smoke-test reports against the current diagnostic schema.
• Recover interrupted restores through an indexed startup path instead of repeated workspace scans.
• Reap daemon-owned containers after cancellation or abrupt Terrabuild termination.
• Refuse cache clearing while another Terrabuild process is using the profile.
• Publish remote cache artifacts as verified atomic generations.
• Recover the previous local cache entry after an interrupted publication.
• Preserve completed execution results when publication or report rendering fails.
• Version diagnostic reports when their result vocabulary changes.
• Fail graph preparation when target actions cannot be completely evaluated.
• Distinguish logical cache reuse from physical execution for restored batch members.
• Clean failed and process-abandoned cache staging directories safely.
• Report tasks prevented by failed dependencies as blocked instead of failed executions.
• Report cached failure summaries as summaries instead of restores.
• Keep expanded forwarded environment secrets out of container arguments and cached summaries.
• Recover interrupted output restores before configuration and project hashing begin.
• Recover interrupted cache output restores from a durable transaction journal.
• Report realized restore fallbacks and include batch staging and publication in finalization timing.
• Dispose synthetic batch cache entries after their logs are copied to member results.
• Restore explicitly empty cached output sets by removing stale declared files.
• Invalidate cached targets when declared forwarded environment values change without retaining their secret values.
• Rebuild targets when remotely cached output files disappear before restoration.
• Keep execution batches separate when container CPU limits or extension scripts differ.
• Prevent targets with different names from being combined into one execution batch.
• Define the last command as the source of a target's default artifact mode.
• Select the latest website-capable release tag when preparing local documentation.
• Render caching and batching criteria as proper documentation lists.
• Keep links to the Next-only target policy guide within the active documentation version.
• Correct target reference examples to keep managed files separate from external images.
• Explain deployment target policies through plan ownership and side-effect use cases.
• Connect introductory and reference documentation to the target policy decision guide.
• Add symptom-based troubleshooting for cache, propagation, batching, locks, and environment inputs.
• Turn the playground quick start into an annotated target-policy tutorial.
• Explain batch-mode use cases, tradeoffs, locking differences, and troubleshooting.
• Add ownership-based caching guidance and concrete artifact-mode recipes.
• Add a decision guide for target scheduling, reuse, batching, environment, and locking policies.
• Expose the explain command in the documentation navigation and command index.
• Clarify how terrabuild clear --all treats idle and active target locks.
• Document inherited named locks in the workspace and project target references.
• Document the diagnostic reason used when cached summaries have no restorable outputs.
• Explain that successful external cache hits reuse summaries without restoring files.
• Drain captured process output streams concurrently to prevent full-pipe deadlocks.
• Finalize started Insights builds as failed when runner operations raise an exception.
• Apply named target locks while cached outputs are restored into the workspace.
• Rebuild offline targets when a cached summary exists without its managed outputs.
• Report named locks and lock-wait timing separately from target execution.
• Keep batch output and log staging inside named target lock leases.
• Remove idle target lock files with terrabuild clear --all.
• Keep named-lock waiters stable during sustained contention.
• Serialize targets that declare the same named lock across concurrent Terrabuild processes.
• Keep execution diagnostics stable when concurrent containers use unique names.
• Restore cached outputs transactionally without exposing cache directories to the runner.
• Pass container arguments without shell re-parsing and isolate concurrent container names.
• Publish cache entries atomically and persist remotely downloaded summaries locally.
• Encapsulate cache output and log storage behind the cache entry API.
• Clarify that external cache hits reuse execution summaries without restoring artifact files.
• Document that explicitly selected lazy targets execute on cache misses.
• Keep phase barriers from rebuilding downstream targets that already have valid cached artifacts.
• Recompute target cache keys when evaluated output patterns change.

Full Changelog: https://github.com/magnusopera/terrabuild/compare/0.199.0-next...509079dbdddaf12e5d2bf4f29c559116b47c3292