| Commit message (Collapse) | Author |
|
Break up monolithic views, task helpers, networking, CLI, and
application extension setup so each module owns a focused concern.
Reorganize the test suite around the new boundaries while preserving
existing Webmention behavior and coverage.
|
|
Cap stale verification requeues so repeatedly lost or crashing tasks
eventually become terminal failures instead of cycling indefinitely.
Reject URLs that httpx cannot safely fetch before enqueueing, and
persist unexpected verification errors as terminal failures.
|
|
Bound incoming request and URL sizes so untrusted Webmention submissions
cannot consume unbounded application or database resources.
Purge old failed and deleted received Webmentions on a schedule, and use
portable SHA-256 URL fingerprints to preserve uniqueness for long URLs.
|
|
Add a periodic recovery sweep for Webmentions left in received or
verifying states after worker crashes, lost queue entries, or
interrupted processing.
Use atomic lease-style updates so stale workers cannot overwrite newer
results, and index verification state for efficient recovery scans.
|
|
Persist retry state per source revision so temporary delivery failures
survive worker restarts and duplicate queued tasks cannot bypass the
retry schedule.
Use capped exponential backoff, stop after a configurable attempt limit,
and reset the retry budget when a source changes or an admin resends
manually.
|
|
|
|
Make database initialization and upgrades explicit so new installations
cannot silently modify an existing or unversioned schema.
Keep Flask-Migrate commands available separately and package the
migration environment with the application for reliable deployment.
|