From 2c76125128f326d5decfaeaac8e1c8cea34853b5 Mon Sep 17 00:00:00 2001
From: Teknium <127238744+teknium1@users.noreply.github.com>
Date: Tue, 6 Oct 2026 11:20:52 -0700
Subject: [PATCH] fix(dashboard): a cancelled 1:1 PTY handler still drops its
 channel marker

The marker discard ran after `await _legacy_pump(...)`. When the client goes away,
the handler task is cancelled while the pump is still awaiting bridge.close in a
worker thread, so the discard never ran and the channel kept a stale marker. It
now runs in a finally. test_legacy_channel_connect_pops_marker_on_disconnect hit
this on CI: TestClient cancels the app task on context exit.
---
 hermes_cli/web_routers/chat_ws.py | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/hermes_cli/web_routers/chat_ws.py b/hermes_cli/web_routers/chat_ws.py
index f463430ec3..94bf471939 100644
--- a/hermes_cli/web_routers/chat_ws.py
+++ b/hermes_cli/web_routers/chat_ws.py
@@ -626,13 +626,17 @@ async def pty_ws(ws: WebSocket) -> None:
                 _discard_active_session_file(ws.app, channel, active_session_file)
             await _pty_fail(ws, exc)
             return
-        await _legacy_pump(ws, bridge)
-        # The 1:1 PTY died with this socket; nothing survives to keep the
-        # breadcrumb, so drop the marker instead of leaking the entry (#63553).
-        # A preexisting marker belongs to a live keep-alive PTY on this channel
-        # — only the marker this handler allocated is ours to drop.
-        if not marker_preexisting:
-            _discard_active_session_file(ws.app, channel, active_session_file)
+        try:
+            await _legacy_pump(ws, bridge)
+        finally:
+            # The 1:1 PTY died with this socket; nothing survives to keep the
+            # breadcrumb, so drop the marker instead of leaking the entry (#63553).
+            # A preexisting marker belongs to a live keep-alive PTY on this channel
+            # — only the marker this handler allocated is ours to drop. In a
+            # finally: a handler cancelled mid-teardown (client gone, server
+            # shutdown) must still drop it.
+            if not marker_preexisting:
+                _discard_active_session_file(ws.app, channel, active_session_file)
         return
 
     # Keep-alive path: the PTY outlives this socket; reattach by token.