Changeset 1b1cc89 in Klonkt for src/services/ActivityPubService.js

Timestamp:

07/01/2026 12:45:51 AM (2 months ago)

Author:

Robin Genis <roboburr@…>

Branches:

main

Children:

e1c4f3e

Parents:

0f2d441

Message:

fix(federation): deliver Follow via the retry queue (byh)

followActor() sent the Follow with a plain deliver() (fire-and-forget), so a
first-attempt failure (peer down/timeout/transient 5xx) left the follow stuck
on 'pending' forever — the Accept can only come back once the Follow lands.
Route it through deliverWithRetry() like deliverCreate/unfollowActor already do:
immediate attempt, then backoff retries via the ap_delivery queue (the worker
re-signs from the slug key).

• src/services/ActivityPubService.js — followActor Follow delivery -> deliverWithRetry
• CHANGELOG.md / .nl / .de — Unreleased Fixed: follow-retry + this session's cover/right-click/boost fixes

Co-Authored-By: Claude <noreply@…>

File:

• src/services/ActivityPubService.js (modified) (1 diff)

src/services/ActivityPubService.js

@@ -1662 +1662 @@
   fwStmts().ins.run(site.slug, actor.id, ai.handle, ai.name, ai.icon, ai.url, actor.inbox, followId, 'pending', autoBoost ? 1 : 0);
   const follow = { '@context': AP_CONTEXT, id: followId, type: 'Follow', actor: me, object: actor.id };
-  try { await deliver(actor.inbox, follow, `${me}#main-key`, keys.private_pem); }
-  catch (e) { console.warn('[AP] follow deliver failed:', e.message); }
+  // Deliver via the retry queue: a Follow that fails the first attempt (peer down,
+  // timeout, transient 5xx) is retried with backoff instead of staying stuck on
+  // 'pending' forever — the Accept can only come back once the Follow lands.
+  await deliverWithRetry(site.slug, actor.inbox, follow, `${me}#main-key`, keys.private_pem);
   console.log('[AP] follow', site.slug, '→', actor.id);
   // Follow + feature in one step → backfill their recent posts into the Cirkel right away.