llama.cpp runtime arm: serve-script runtime switch, harness containerless support, master build

autoresearch.sh

@@ -24,7 +24,7 @@ restore() {
     echo "[harness] tearing down candidate, restoring deployed baseline..." >&2
     ssh "$HOST" 'sudo -n systemctl stop frank-large.service 2>/dev/null;
                  podman rm -f frank-large-mimo 2>/dev/null;
-                 true' >/dev/null 2>&1 || true
+                 pkill -x llama-server 2>/dev/null; true' >/dev/null 2>&1 || true
     ssh "$HOST" 'sudo -n systemctl start frank-large.service' >/dev/null 2>&1 || true
     exit $rc
 }
@@ -40,14 +40,13 @@ rsync -a "$REPO/scripts/frank-large.env" "$REPO/scripts/serve-frank-large.sh" \
 ssh "$HOST" 'sudo -n systemctl stop frank-large.service;
              podman rm -f frank-large-mimo 2>/dev/null; true' >/dev/null
 
-# 3. Boot the candidate detached. The serve script's own gates run first;
-#    a gate failure exits here with the reason.
+# 3. Boot the candidate detached; the launcher's own output IS the serve
+#    stage output (gate failures print there before any container exists).
 echo "[harness] starting candidate..." >&2
-if ! ssh "$HOST" "setsid nohup env FRANK_LARGE_ENV_FILE=$RDIR/frank-large.env \
+ssh "$HOST" "setsid nohup env FRANK_LARGE_ENV_FILE=$RDIR/frank-large.env \
         bash $RDIR/serve-frank-large.sh >$RDIR/serve.log 2>&1 </dev/null & echo ok" \
-        | grep -q ok; then
-    echo "ERROR: could not launch candidate" >&2; exit 1
-fi
+    | tee /tmp/autoresearch-launch.log >&2
+grep -q ok /tmp/autoresearch-launch.log || { echo "ERROR: launch failed" >&2; exit 1; }
 
 # 4. Wait for health (503 while loading; anything else = boot failure).
 echo "[harness] waiting for health..." >&2
@@ -55,10 +54,15 @@ if ! ssh "$HOST" "for i in \$(seq 1 $HEALTH_TRIES); do
         code=\$(curl -s -o /dev/null -w '%{http_code}' --max-time 3 \
               http://127.0.0.1:7770/health 2>/dev/null || echo 000)
         [ \"\$code\" = 200 ] && echo UP && exit 0
+        st=\$(podman inspect -f '{{.State.Status}}' frank-large-mimo 2>/dev/null || echo gone)
+        if [ \"\$st\" = exited ] || { [ \"\$st\" = gone ] && ! pgrep -f serve-frank-large >/dev/null; }; then echo DOWN; exit 1; fi
         sleep 5
     done; exit 1" | grep -q UP; then
     echo "ERROR: candidate never became healthy; serve.log tail:" >&2
-    ssh "$HOST" "tail -30 $RDIR/serve.log" >&2 || true
+    ssh "$HOST" "tail -30 $RDIR/serve.log; echo '--- engine log ---';
+                 podman logs --tail 120 frank-large-mimo 2>&1 |
+                 grep -aiE 'error|raise|failed|traceback|assert|kernel' |
+                 tail -30" >&2 || true
     exit 1
 fi