diff options
| author | Jakub Stasiak <jakub@stasiak.at> | 2016-01-10 04:01:36 +0100 |
|---|---|---|
| committer | Jakub Stasiak <jakub@stasiak.at> | 2016-01-10 04:01:36 +0100 |
| commit | b2fb240978654950005636442ca20107a96118ec (patch) | |
| tree | 6b724c5263cd4074d6bdcf744d6a1457f6a2eb30 /tests | |
| parent | ff2479fd966811519e4c2e5e4e908bbd326fbaf1 (diff) | |
| download | eventlet-b2fb240978654950005636442ca20107a96118ec.tar.gz | |
tests: Make the test_closure test less CPU intensive
As far as I can see the test already had a tendency to timeout and [1]
pushed it over the threshold of 1 second on Travis (Python 2 builds).
I don't see a point of flooding the socket like this so adding an
artifical delay seems like an easy "fix" here.
From what I understand the logic of the test (introduced in [2]) remains
unchanged - a closer greenthread closes the client socket, client
greenthread finishes because of EBADF and server greenthread finishes
because of broken pipe.
[1] bc4d1b5d362e5baaeded35b1e339b9db08172dd2
[2] 1a8a9fc0512299c8fe944784525a021dbd2fe2a0
Diffstat (limited to 'tests')
| -rw-r--r-- | tests/greenio_test.py | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/tests/greenio_test.py b/tests/greenio_test.py index 5eaf6b5..9a2699a 100644 --- a/tests/greenio_test.py +++ b/tests/greenio_test.py @@ -481,6 +481,9 @@ class TestGreenSocket(tests.LimitedTestCase): while True: try: sock.sendall(b'hello world') + # Arbitrary delay to not use all available CPU, keeps the test + # running quickly and reliably under a second + time.sleep(0.001) except socket.error as e: if get_errno(e) == errno.EPIPE: return @@ -496,6 +499,9 @@ class TestGreenSocket(tests.LimitedTestCase): while True: data = client.recv(1024) assert data + # Arbitrary delay to not use all available CPU, keeps the test + # running quickly and reliably under a second + time.sleep(0.001) except socket.error as e: # we get an EBADF because client is closed in the same process # (but a different greenthread) |
