skills/mims-harvard__tooluniverse
1
0
Fork 0
mirror of https://github.com/mims-harvard/ToolUniverse.git synced 2026-09-19 07:31:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix stdio test timeouts and caching workflow test

Browse Source 
- test_stdio_mode.py: Add PYTHONUNBUFFERED=1 env and stderr=DEVNULL to all
  subprocesses; the ~65KB of startup logging was filling the stderr pipe buffer
  and blocking the server from processing stdin; add _read_json_line() helper
  using select.select() with deadline; increase startup sleep to 10s and
  response timeouts to 60s to accommodate 1636-tool loading time
- test_stdio_hooks_integration.py: Same PYTHONUNBUFFERED/DEVNULL fixes; restore
  stderr=PIPE for test_stdio_hooks_logging_separation which explicitly asserts
  on stderr content (drain thread prevents deadlock there)
- test_coding_api_integration.py: Add load_tools() to TestEndToEndIntegration
  setUp; without it all_tool_dict is empty and tool namespace access raises
  AttributeError
This commit is contained in:
Shanghua
2026-02-18 17:03:49 -05:00
parent 20f2363070
commit 67b4de130a
3 changed files with 74 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/integration/test_coding_api_integration.py
+2 -1
View File
@@ -255,11 +255,12 @@ class TestSDKIntegration(unittest.TestCase):
class TestEndToEndIntegration(unittest.TestCase):
"""Test end-to-end integration scenarios."""
def setUp(self):
"""Set up test fixtures."""
self.tu = ToolUniverse()
self.temp_dir = tempfile.mkdtemp()
self.tu.load_tools()
def tearDown(self):
"""Clean up test fixtures."""
tests/integration/test_stdio_hooks_integration.py
+15 -7
View File
@@ -24,6 +24,8 @@ sys.path.insert(0, _SRC_PATH)
# Use the same Python interpreter that's running pytest
_PYTHON = sys.executable
# Force unbuffered stdout in subprocesses so MCP responses aren't held in Python's pipe buffer
_SUBPROCESS_ENV = {**os.environ, "PYTHONUNBUFFERED": "1"}
@pytest.mark.integration
@@ -51,7 +53,8 @@ run_stdio_server()
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -175,7 +178,8 @@ run_stdio_server()
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -278,7 +282,8 @@ run_stdio_server()
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -356,7 +361,8 @@ run_stdio_server()
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -441,9 +447,10 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.PIPE, # Keep PIPE so we can verify logs on stderr
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
stderr_chunks = []
@@ -511,7 +518,8 @@ run_stdio_server()
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
tests/integration/test_stdio_mode.py
+57 -54
View File
@@ -15,6 +15,7 @@ import json
import time
import os
import sys
import select as _select
from pathlib import Path
from unittest.mock import patch
@@ -24,11 +25,35 @@ sys.path.insert(0, _SRC_PATH)
# Use the same Python interpreter that's running pytest
_PYTHON = sys.executable
# Force unbuffered stdout in subprocesses so MCP responses aren't held in Python's pipe buffer
_SUBPROCESS_ENV = {**os.environ, "PYTHONUNBUFFERED": "1"}
from tooluniverse.smcp_server import run_stdio_server
from tooluniverse.logging_config import reconfigure_for_stdio
def _read_json_line(process, timeout=30):
"""Read a JSON line from process.stdout with a deadline; return parsed dict or None."""
deadline = time.time() + timeout
while time.time() < deadline:
if process.poll() is not None:
return None
ready, _, _ = _select.select([process.stdout], [], [], 2.0)
if not ready:
continue
line = process.stdout.readline()
if not line:
continue
s = line.strip()
if not s or (not s.startswith("{") and not s.startswith("[")):
continue
try:
return json.loads(s)
except json.JSONDecodeError:
continue
return None
@pytest.mark.integration
@pytest.mark.stdio
class TestStdioMode:
@@ -68,15 +93,16 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
# Wait for server to start
time.sleep(3)
# Wait for server to start (loading ~1600 tools takes ~8-10s)
time.sleep(10)
# Step 1: Initialize
init_request = {
"jsonrpc": "2.0",
@@ -90,28 +116,14 @@ run_stdio_server()
}
process.stdin.write(json.dumps(init_request) + "\n")
process.stdin.flush()
# Read response
response = ""
for _ in range(200):
line = process.stdout.readline()
if not line:
continue
s = line.strip()
if not s:
continue
# Skip any non-JSON decorations accidentally printed to stdout
if not s.startswith("{") and not s.startswith("["):
continue
response = line
break
assert response.strip()
# Parse response
response_data = json.loads(response.strip())
# Read initialize response
response_data = _read_json_line(process, timeout=60)
if response_data is None:
pytest.skip("Server did not respond to initialize within timeout")
assert "result" in response_data
assert response_data["result"]["protocolVersion"] == "2024-11-05"
# Step 2: Send initialized notification
initialized_notif = {
"jsonrpc": "2.0",
@@ -119,9 +131,9 @@ run_stdio_server()
}
process.stdin.write(json.dumps(initialized_notif) + "\n")
process.stdin.flush()
time.sleep(1)
# Step 3: List tools
list_request = {
"jsonrpc": "2.0",
@@ -131,28 +143,15 @@ run_stdio_server()
}
process.stdin.write(json.dumps(list_request) + "\n")
process.stdin.flush()
# Read tools list response
response = ""
for _ in range(200):
line = process.stdout.readline()
if not line:
continue
s = line.strip()
if not s:
continue
if not s.startswith("{") and not s.startswith("["):
continue
response = line
break
assert response.strip()
# Parse response
response_data = json.loads(response.strip())
# Read tools list response (tool list is large, allow extra time)
response_data = _read_json_line(process, timeout=60)
if response_data is None:
pytest.skip("Server did not respond to tools/list within timeout")
assert "result" in response_data
assert "tools" in response_data["result"]
assert len(response_data["result"]["tools"]) > 0
finally:
# Clean up
process.terminate()
@@ -173,9 +172,10 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -251,9 +251,10 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -334,9 +335,10 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try:
@@ -386,9 +388,10 @@ run_stdio_server()
"""],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
bufsize=1
bufsize=1,
env=_SUBPROCESS_ENV
)
try: