From 69f8f4cfd7f7a30fd4b9308e86f422c594981875 Mon Sep 17 00:00:00 2001
From: KoalaDev <6156589+Shik3i@users.noreply.github.com>
Date: Fri, 26 Jun 2026 10:01:04 +0200
Subject: [PATCH] fix(extension): trust HOST_BLOCKED as authoritative (join
 race, EC-5)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

hcmHandleBlocked no longer re-checks local control mode before acting. Background
only sends HOST_BLOCKED to a gated guest, so the message itself is authoritative —
adopt host-only/guest role from it. Fixes the join race where a HOST_BLOCKED
arriving before the CONTROL_MODE broadcast was silently ignored.

Also corrects the EC-4 note: "let catch-up re-sync" is invalid (sync is
event-driven, no continuous catch-up); the right fix is a buffer-aware deferred
snap-back.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 docs/host-control-mode-EDGECASES.md | 17 ++++++++++-------
 extension/content.js                |  7 ++++++-
 2 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/docs/host-control-mode-EDGECASES.md b/docs/host-control-mode-EDGECASES.md
index 5a2f1c5..501bd42 100644
--- a/docs/host-control-mode-EDGECASES.md
+++ b/docs/host-control-mode-EDGECASES.md
@@ -39,13 +39,16 @@ Deferred by decision (see §8): host grace period on disconnect (EC-10).
 
 ### Pre-test self-audit (open, watch during device testing)
 - **EC-4/EC-1 snap-back thrash:** for *involuntary* events we still actively seek+play,
-  which can fight a buffering player for the duration of the stall. Likely fix: on
-  involuntary, don't re-play — let catch-up re-sync — or use an exponential cooldown.
-  Decide after watching it on Netflix/YouTube.
-- **Control-mode race at join:** a `HOST_BLOCKED` arriving before content.js learns the
-  mode is ignored (`hcmIsGuestGated()` false). Fix: trust `HOST_BLOCKED` as
-  authoritative (background only sends it to gated guests) instead of re-checking local
-  mode. Small change, deferred pending test.
+  which can fight a buffering player for the duration of the stall. NOTE: "just let
+  catch-up re-sync" is NOT a valid fix — sync is event-driven, there is no continuous
+  catch-up loop, so skipping the snap-back risks leaving the guest stuck paused/behind
+  until the host next acts. Correct fix is a **buffer-aware deferred snap-back**: when
+  involuntary + buffering, wait for readyState>=3 (à la pollSeekReady) then seek+play
+  once — avoids thrash AND guarantees re-sync. Build after device testing reveals which
+  players fire pause() vs only 'waiting'.
+- ~~**Control-mode race at join:**~~ FIXED — `hcmHandleBlocked` now treats `HOST_BLOCKED`
+  as authoritative (adopts host-only/guest role) instead of re-checking local mode,
+  since background only sends it to gated guests.
 - **Dialog/badge text is English-only** — content.js has no i18n loader; the in-page
   strings aren't localized yet. Follow-up.
 
diff --git a/extension/content.js b/extension/content.js
index 6514556..387a907 100644
--- a/extension/content.js
+++ b/extension/content.js
@@ -175,7 +175,12 @@
 
     // Entry point: background told us our local action was blocked in host-only.
     function hcmHandleBlocked(action, target) {
-        if (!hcmIsGuestGated()) return;
+        // HOST_BLOCKED is only ever sent to a gated guest (background verifies
+        // host-only + !host before sending), so it's authoritative. Adopt the
+        // role/mode from it in case our CONTROL_MODE broadcast hasn't landed yet
+        // (join race, EC-5) — otherwise we'd miss the dialog/snap-back.
+        hcmControlMode = 'host-only';
+        hcmAmHost = false;
         if (Date.now() < hcmSnapBackCooldownUntil) return; // EC-4 loop guard
         if (hcmDesynced) return; // already solo, nothing to do