From 73066749e3526b27ae14bb15396dcd6fd969ddda Mon Sep 17 00:00:00 2001
From: EdisonJwa <me@edison.network>
Date: Sat, 16 May 2026 12:14:28 +0800
Subject: [PATCH] fix(audio,linux): isolate zbus blocking probe on a fresh OS
 thread

LinuxGnomeWaylandBackend::probe() called zbus::blocking::Connection::session()
directly. The blocking facade internally constructs a current-thread tokio
runtime and block_on()s its async D-Bus client. probe() runs from
PttController::new (sync) which is called from start_audio (async on the
bridge tokio runtime). Nested runtimes panic with 'Cannot start a runtime
from within a runtime'.

Symptom on Linux: the first voice-channel join surfaced a SnackBar
'Could not join channel: join: task N panicked ...' while the channel-move
command had already succeeded server-side. User saw 'channel joined but voice
not enabled'.

Fix: run the cheap blocking probe on a dedicated std::thread (no ambient
runtime), join it synchronously, propagate the version / error. Probe is
microseconds; the join cost is negligible.
---
 .../chanora_audio/src/ptt_backends/linux.rs   | 35 ++++++++++++++-----
 1 file changed, 26 insertions(+), 9 deletions(-)

diff --git a/crates/chanora_audio/src/ptt_backends/linux.rs b/crates/chanora_audio/src/ptt_backends/linux.rs
index 68146e5..a0ac894 100644
--- a/crates/chanora_audio/src/ptt_backends/linux.rs
+++ b/crates/chanora_audio/src/ptt_backends/linux.rs
@@ -231,16 +231,33 @@ enum WorkerCmd {
 impl LinuxGnomeWaylandBackend {
     /// Synchronous probe via the blocking proxy. Cheap; runs once
     /// at `try_select` time before we commit to the live flow.
+    ///
+    /// `zbus::blocking` internally spins up its own current-thread
+    /// tokio runtime to drive the async D-Bus client. If we called
+    /// it directly here we would panic with "Cannot start a runtime
+    /// from within a runtime" because `try_select()` is invoked
+    /// from `PttController::new`, which is called from the audio
+    /// engine start path, which runs on the bridge's tokio
+    /// multi-thread runtime. We therefore push the blocking probe
+    /// onto a fresh OS thread (no ambient runtime), join it
+    /// synchronously, and propagate the result. The probe is in
+    /// the millisecond range so the join cost is negligible.
     fn probe() -> Result<Self, PttBackendError> {
-        let conn = BlockingConnection::session()
-            .map_err(|e| PttBackendError::Init(format!("session bus: {e}")))?;
-        let version = {
-            let proxy = BlockingGlobalShortcutsProxy::new(&conn)
-                .map_err(|e| PttBackendError::Init(format!("proxy: {e}")))?;
-            proxy
-                .version()
-                .map_err(|e| PttBackendError::Init(format!("portal version: {e}")))?
-        };
+        let join_result = std::thread::Builder::new()
+            .name("chanora-ptt-portal-probe".to_string())
+            .spawn(|| -> Result<u32, String> {
+                let conn = BlockingConnection::session()
+                    .map_err(|e| format!("session bus: {e}"))?;
+                let proxy = BlockingGlobalShortcutsProxy::new(&conn)
+                    .map_err(|e| format!("proxy: {e}"))?;
+                proxy
+                    .version()
+                    .map_err(|e| format!("portal version: {e}"))
+            })
+            .map_err(|e| PttBackendError::Init(format!("probe thread spawn: {e}")))?
+            .join()
+            .map_err(|_| PttBackendError::Init("probe thread panicked".to_string()))?;
+        let version = join_result.map_err(PttBackendError::Init)?;
         info!(
             target: "chanora_audio",
             portal_version = version,