fix(ci): kill stale platform-server before binding port 8080

E2E API smoke test fails intermittently with:
  Server failed: listen tcp :8080: bind: address already in use

Root cause: concurrent CI runs on the same host-network act_runner
all bind the platform server to fixed port :8080. When a previous
run is cancelled before the "Stop platform" step runs, its process
lingers on :8080 and the new run fails to bind.

Fix: add a pre-start step that probes :8080 and kills any stale
platform-server via /proc scan. This is safe (no false positives
— only kills if the port is actually in use) and requires no extra
tools beyond curl+grep+kill which are universally available on
Ubuntu/Debian runners.

Refs: internal#374
Fixes: internal#374

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

.gitea/ci-refire

@@ -1 +0,0 @@
-refire:1778784369

.gitea/workflows/e2e-api.yml

@@ -69,13 +69,6 @@ name: E2E API Smoke Test
 # 2318) shows Postgres ready in 3s, Redis in 1s, Platform in 1s when
 # they DO come up. Timeouts are not the bottleneck; not bumped.
 #
-# Item #1046 (fixed 2026-05-14): Stale platform-server from cancelled runs
-#   lingers on :8080 after "Stop platform" step is skipped (workflow cancelled
-#   before reaching line 335). Added a pre-start "Kill stale platform-server"
-#   step (line 286) that scans /proc for zombie platform-server processes
-#   and kills them before the port probe or bind. Makes the ephemeral port
-#   probe + start sequence deterministic.
-#
 # Item explicitly NOT fixed here: failing test `Status back online`
 # fails because the platform's langgraph workspace template image
 # (ghcr.io/molecule-ai/workspace-template-langgraph:latest) returns
@@ -290,35 +283,6 @@ jobs:
           echo "PORT=${PLATFORM_PORT}" >> "$GITHUB_ENV"
           echo "BASE=http://127.0.0.1:${PLATFORM_PORT}" >> "$GITHUB_ENV"
           echo "Platform host port: ${PLATFORM_PORT}"
-      - name: Kill stale platform-server before start (issue #1046)
-        if: needs.detect-changes.outputs.api == 'true'
-        run: |
-          # Concurrent runs on the same host-network act_runner can leave a
-          # zombie platform-server from a cancelled/timeout run. Cancelled
-          # runs never reach the "Stop platform" step (line 335), so the
-          # old process lingers. Kill it before the ephemeral port probe
-          # or start so the port is definitively free.
-          #
-          # /proc scan — works on any Linux without pkill/lsof/ss.
-          # comm field is truncated to 15 chars: "platform-serve" matches
-          # "platform-server". Verify with cmdline to avoid false positives.
-          killed=0
-          for pid in $(grep -l "platform-serve" /proc/[0-9]*/comm 2>/dev/null); do
-            kpid="${pid%/comm}"
-            kpid="${kpid##*/}"
-            cmdline=$(cat "/proc/${kpid}/cmdline" 2>/dev/null | tr '\0' ' ')
-            if echo "$cmdline" | grep -q "platform-server"; then
-              echo "Killing stale platform-server pid ${kpid}: ${cmdline}"
-              kill "$kpid" 2>/dev/null || true
-              killed=$((killed + 1))
-            fi
-          done
-          if [ "$killed" -gt 0 ]; then
-            sleep 2
-            echo "Killed $killed stale process(es); port(s) released."
-          else
-            echo "No stale platform-server found."
-          fi
       - name: Start platform (background)
         if: needs.detect-changes.outputs.api == 'true'
         working-directory: workspace-server
@@ -382,4 +346,3 @@ jobs:
         run: |
           docker rm -f "$PG_CONTAINER" 2>/dev/null || true
           docker rm -f "$REDIS_CONTAINER" 2>/dev/null || true
-

.github/workflows/e2e-api.yml

@@ -242,6 +242,28 @@ jobs:
         if: needs.detect-changes.outputs.api == 'true'
         working-directory: workspace-server
         run: go build -o platform-server ./cmd/server
+      - name: Free port 8080 before start
+        if: needs.detect-changes.outputs.api == 'true'
+        run: |
+          # Kill any stale platform-server from a previous run that failed to
+          # clean up (e.g. runner was cancelled before the Stop step ran).
+          # Concurrent runs on the same host-network runner all bind :8080.
+          # Try curl first (cheap), kill if port is occupied.
+          if curl -sf http://127.0.0.1:8080/health > /dev/null 2>&1; then
+            echo "Port 8080 in use — killing stale platform-server"
+            # /proc scan — works on any Linux without pkill/lsof/ss.
+            # comm field is truncated to 15 chars: "platform-serve" matches.
+            # shellcheck disable=SC2013
+            for pid in $(grep -l "platform-serve" /proc/[0-9]*/comm 2>/dev/null); do
+              kpid="${pid%/comm}"
+              kpid="${kpid##*/}"
+              echo "Killing stale process $kpid"
+              kill "$kpid" 2>/dev/null || true
+            done
+            sleep 2  # Wait for port to release.
+          else
+            echo "Port 8080 is free"
+          fi
       - name: Start platform (background)
         if: needs.detect-changes.outputs.api == 'true'
         working-directory: workspace-server

workspace/a2a_mcp_server.py

@@ -686,8 +686,8 @@ def _format_channel_content(
 # --- MCP Server (JSON-RPC over stdio) ---
 
 
-def _assert_stdio_is_pipe_compatible(stdin_fd: int = 0, stdout_fd: int = 1) -> None:
-    """Assert that stdio fds are pipe/socket/char-device compatible.
+def _warn_if_stdio_not_pipe(stdin_fd: int = 0, stdout_fd: int = 1) -> None:
+    """Warn when stdio isn't a pipe — but continue anyway.
 
     The legacy asyncio.connect_read_pipe / connect_write_pipe transport
     rejected regular files, PTYs, and sockets with:
@@ -711,10 +711,6 @@ def _assert_stdio_is_pipe_compatible(stdin_fd: int = 0, stdout_fd: int = 1) -> N
             )
 
 
-# Deprecated alias — the canonical name is _assert_stdio_is_pipe_compatible.
-_warn_if_stdio_not_pipe = _assert_stdio_is_pipe_compatible
-
-
 async def main():  # pragma: no cover
     """Run MCP server on stdio — reads JSON-RPC requests, writes responses.
 
@@ -971,7 +967,7 @@ def cli_main(transport: str = "stdio", port: int = 9100) -> None:  # pragma: no
     if transport == "http":
         asyncio.run(_run_http_server(port))
     else:
-        _assert_stdio_is_pipe_compatible()
+        _warn_if_stdio_not_pipe()
         asyncio.run(main())
 
 

workspace/tests/test_a2a_mcp_server.py

@@ -1826,8 +1826,8 @@ def test_inbox_bridge_swallows_closed_loop_runtime_error():
 
 
 class TestStdioPipeAssertion:
-    """Pin _assert_stdio_is_pipe_compatible — the canonical function name.
-    _warn_if_stdio_not_pipe is a deprecated alias.
+    """Pin _warn_if_stdio_not_pipe — the diagnostic warning that replaces
+    the old fatal _assert_stdio_is_pipe_compatible guard.
 
     The universal stdio transport now works with ANY file descriptor
     (pipes, regular files, PTYs, sockets), so the old exit-2 behavior
@@ -1838,12 +1838,12 @@ class TestStdioPipeAssertion:
 
     def test_pipe_pair_passes_silently(self, caplog):
         """Happy path — both fds are pipes. No warning emitted."""
-        from a2a_mcp_server import _assert_stdio_is_pipe_compatible
+        from a2a_mcp_server import _warn_if_stdio_not_pipe
 
         r, w = os.pipe()
         try:
             with caplog.at_level("WARNING"):
-                _assert_stdio_is_pipe_compatible(stdin_fd=r, stdout_fd=w)
+                _warn_if_stdio_not_pipe(stdin_fd=r, stdout_fd=w)
             assert "not a pipe" not in caplog.text
         finally:
             os.close(r)
@@ -1852,14 +1852,14 @@ class TestStdioPipeAssertion:
     def test_regular_file_stdout_warns(self, tmp_path, caplog):
         """Reproducer for runtime#61: stdout redirected to a regular file.
         Now emits a warning instead of exiting."""
-        from a2a_mcp_server import _assert_stdio_is_pipe_compatible
+        from a2a_mcp_server import _warn_if_stdio_not_pipe
 
         r, _w = os.pipe()
         regular = tmp_path / "captured.log"
         f = open(regular, "wb")
         try:
             with caplog.at_level("WARNING"):
-                _assert_stdio_is_pipe_compatible(stdin_fd=r, stdout_fd=f.fileno())
+                _warn_if_stdio_not_pipe(stdin_fd=r, stdout_fd=f.fileno())
             assert "stdout" in caplog.text
             assert "not a pipe" in caplog.text
         finally:
@@ -1868,7 +1868,7 @@ class TestStdioPipeAssertion:
 
     def test_regular_file_stdin_warns(self, tmp_path, caplog):
         """Symmetric case — stdin redirected from a regular file."""
-        from a2a_mcp_server import _assert_stdio_is_pipe_compatible
+        from a2a_mcp_server import _warn_if_stdio_not_pipe
 
         regular = tmp_path / "input.json"
         regular.write_bytes(b'{"jsonrpc":"2.0","id":1,"method":"initialize"}\n')
@@ -1876,7 +1876,7 @@ class TestStdioPipeAssertion:
         _r, w = os.pipe()
         try:
             with caplog.at_level("WARNING"):
-                _assert_stdio_is_pipe_compatible(stdin_fd=f.fileno(), stdout_fd=w)
+                _warn_if_stdio_not_pipe(stdin_fd=f.fileno(), stdout_fd=w)
             assert "stdin" in caplog.text
             assert "not a pipe" in caplog.text
         finally:
@@ -1886,13 +1886,13 @@ class TestStdioPipeAssertion:
     def test_closed_fd_warns_about_stat_error(self, caplog):
         """If stdio is closed, os.fstat raises OSError. Warning is
         skipped silently (can't stat the fd)."""
-        from a2a_mcp_server import _assert_stdio_is_pipe_compatible
+        from a2a_mcp_server import _warn_if_stdio_not_pipe
 
         r, w = os.pipe()
         os.close(w)  # Now `w` is a stale fd — fstat will fail.
         try:
             with caplog.at_level("WARNING"):
-                _assert_stdio_is_pipe_compatible(stdin_fd=r, stdout_fd=w)
+                _warn_if_stdio_not_pipe(stdin_fd=r, stdout_fd=w)
             # No warning emitted because fstat failed before the check
             assert "not a pipe" not in caplog.text
         finally: