[QPID-2294] The Unix python client can erroneously throw exceptions from select due to interrupted system call

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.6
• Fix Version/s: 0.27
• Component/s: Python Client

Environment

Red Hat Enterprise Linux 5.4

• Assignee: Rafael H. Schloming
• Reporter: Andrew Stitcher
• Created: 17/Dec/09 22:31
• Updated: 03/Jun/14 15:34
• Resolved: 15/Jan/14 18:41

Description

When running autotools "make check"

You can cause the python_tests to fail in the qpid.tests.messaging section (and perhaps elsewhere) by resizing the window that is running the tests.

You get something like this:
qpid.tests.messaging.SenderTests.testSendAsyncCapacityUNLIMITED ............................ fail
Error during test:
Traceback (most recent call last):
File "/home/astitche/bld-working/src/tests/python/commands/qpid-python-test", line 307, in run
phase()
File "/home/astitche/bld-working/src/tests/python/qpid/tests/messaging.py", line 796, in testSendAsyncCapacityUNLIMITED
self.asyncTest(UNLIMITED)
File "/home/astitche/bld-working/src/tests/python/qpid/tests/messaging.py", line 777, in asyncTest
drained = self.drain(self.rcv, timeout=self.delay())
File "/home/astitche/bld-working/src/tests/python/qpid/tests/messaging.py", line 84, in drain
contents.append(rcv.fetch(timeout=timeout).content)
File "<string>", line 6, in fetch
File "/home/astitche/bld-working/src/tests/python/qpid/messaging.py", line 668, in fetch
msg = self.session._get(self._pred, timeout=timeout)
File "<string>", line 6, in _get
File "/home/astitche/bld-working/src/tests/python/qpid/messaging.py", line 360, in _get
timeout):
File "/home/astitche/bld-working/src/tests/python/qpid/messaging.py", line 294, in _ewait
result = self.connection._ewait(lambda: self.error or predicate(), timeout, exc)
File "/home/astitche/bld-working/src/tests/python/qpid/messaging.py", line 142, in _ewait
result = self._wait(lambda: self.error or predicate(), timeout)
File "/home/astitche/bld-working/src/tests/python/qpid/messaging.py", line 131, in _wait
return self._waiter.wait(predicate, timeout=timeout)
File "/home/astitche/bld-working/src/tests/python/qpid/concurrency.py", line 59, in wait
self.condition.wait(timeout - passed)
File "/home/astitche/bld-working/src/tests/python/qpid/concurrency.py", line 96, in wait
sw.wait(timeout)
File "/home/astitche/bld-working/src/tests/python/qpid/compat.py", line 53, in wait
ready, _, _ = select([self], [], [], timeout)
error: (4, 'Interrupted system call')

The cause is that python is receiving the SIGWINCH signal from the window size change and this is interrupting the select system call. The exception that is being thrown by select is not being caught.

This needs to be fixed as interrupted system calls are a fact of life when running on Unix - you could argue that the python run time should just restart the select, but it isn't doing that here.

Note that allowing for interrupted system calls screws up the timeout calculation. As you'll need to figure out how much of the wait is left and then wait for less time when restarting the system call.