Re: Review Request 22528: Rebalance server gets stuck in windows when client2 is not stopped
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/#review45610 --- proton/trunk/tests/python/proton_tests/messenger.py https://reviews.apache.org/r/22528/#comment80488 Question for Rafi - why is this 'manual flushing' necessary? It implies that messenger will never reach 'stopped' unless the application manually monitors incoming/outgoing and keeps cleaning up. I didn't expect this to be needed. proton/trunk/tests/python/proton_tests/messenger.py https://reviews.apache.org/r/22528/#comment80489 Doesn't this suffer from the original race condition too? Specifically, if the rebalancer kicks in during the previous pump, these asserts will fail, no? - Kenneth Giusti On June 12, 2014, 8:38 p.m., Chug Rolke wrote: --- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/ --- (Updated June 12, 2014, 8:38 p.m.) Review request for qpid, Kenneth Giusti and Rafael Schloming. Bugs: PROTON-606 https://issues.apache.org/jira/browse/PROTON-606 Repository: qpid Description --- On windows (python 2.6.1) the server in testCreditRebalance won't shut down when client2 in the test is not stopped. Since the test may assert at a random place there's no way to guarantee client2's state. * Promote test client2 messenger to be a full class member * In teardown() when a messenger's outgoing 0 then call settle() * In teardown() while incoming 0 call get() * Add a new test that formerly hung the server but is now fixed Diffs - proton/trunk/tests/python/proton_tests/messenger.py 1602016 Diff: https://reviews.apache.org/r/22528/diff/ Testing --- works on linux and windows. Thanks, Chug Rolke
Re: Review Request 22528: Rebalance server gets stuck in windows when client2 is not stopped
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/#review45612 --- proton/trunk/tests/python/proton_tests/messenger.py https://reviews.apache.org/r/22528/#comment80490 On retest without the .outgoing and .incoming flush the windows works just fine. I put random asserts into original test and windows stops and the random assert messages are printed. No cleanup needed. I'll submit another patch, without the annoying red trailing white space. - Chug Rolke On June 12, 2014, 8:38 p.m., Chug Rolke wrote: --- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/ --- (Updated June 12, 2014, 8:38 p.m.) Review request for qpid, Kenneth Giusti and Rafael Schloming. Bugs: PROTON-606 https://issues.apache.org/jira/browse/PROTON-606 Repository: qpid Description --- On windows (python 2.6.1) the server in testCreditRebalance won't shut down when client2 in the test is not stopped. Since the test may assert at a random place there's no way to guarantee client2's state. * Promote test client2 messenger to be a full class member * In teardown() when a messenger's outgoing 0 then call settle() * In teardown() while incoming 0 call get() * Add a new test that formerly hung the server but is now fixed Diffs - proton/trunk/tests/python/proton_tests/messenger.py 1602016 Diff: https://reviews.apache.org/r/22528/diff/ Testing --- works on linux and windows. Thanks, Chug Rolke
Re: Review Request 22528: Rebalance server gets stuck in windows when client2 is not stopped
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/ --- (Updated June 13, 2014, 3:17 p.m.) Review request for qpid, Kenneth Giusti and Rafael Schloming. Changes --- This patch does away with the .incoming and .outgoing flush code. Also, the main issue fixed here is the pump loop. Pulling client2 into the loop makes the original test behave itself[1] and print the assert messages from the original assert without having a broker/client stopped assert overwrite it. [1] However, inserting an assert between a put() and pump() causes a core dump on both windows and linux. Calling settled and get works around that issue and that's where that code came from. A new bug is forthcoming. Bugs: PROTON-606 https://issues.apache.org/jira/browse/PROTON-606 Repository: qpid Description --- On windows (python 2.6.1) the server in testCreditRebalance won't shut down when client2 in the test is not stopped. Since the test may assert at a random place there's no way to guarantee client2's state. * Promote test client2 messenger to be a full class member * In teardown() when a messenger's outgoing 0 then call settle() * In teardown() while incoming 0 call get() * Add a new test that formerly hung the server but is now fixed Diffs (updated) - proton/trunk/tests/python/proton_tests/messenger.py 1602016 Diff: https://reviews.apache.org/r/22528/diff/ Testing --- works on linux and windows. Thanks, Chug Rolke
Re: Review Request 22528: Rebalance server gets stuck in windows when client2 is not stopped
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/#review45620 --- Ship it! Ship It! - Kenneth Giusti On June 13, 2014, 3:17 p.m., Chug Rolke wrote: --- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/ --- (Updated June 13, 2014, 3:17 p.m.) Review request for qpid, Kenneth Giusti and Rafael Schloming. Bugs: PROTON-606 https://issues.apache.org/jira/browse/PROTON-606 Repository: qpid Description --- On windows (python 2.6.1) the server in testCreditRebalance won't shut down when client2 in the test is not stopped. Since the test may assert at a random place there's no way to guarantee client2's state. * Promote test client2 messenger to be a full class member * In teardown() when a messenger's outgoing 0 then call settle() * In teardown() while incoming 0 call get() * Add a new test that formerly hung the server but is now fixed Diffs - proton/trunk/tests/python/proton_tests/messenger.py 1602016 Diff: https://reviews.apache.org/r/22528/diff/ Testing --- works on linux and windows. Thanks, Chug Rolke
Re: Review Request 22528: Rebalance server gets stuck in windows when client2 is not stopped
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22528/ --- (Updated June 12, 2014, 8:38 p.m.) Review request for qpid, Kenneth Giusti and Rafael Schloming. Changes --- the actual diff Bugs: PROTON-606 https://issues.apache.org/jira/browse/PROTON-606 Repository: qpid Description --- On windows (python 2.6.1) the server in testCreditRebalance won't shut down when client2 in the test is not stopped. Since the test may assert at a random place there's no way to guarantee client2's state. * Promote test client2 messenger to be a full class member * In teardown() when a messenger's outgoing 0 then call settle() * In teardown() while incoming 0 call get() * Add a new test that formerly hung the server but is now fixed Diffs (updated) - proton/trunk/tests/python/proton_tests/messenger.py 1602016 Diff: https://reviews.apache.org/r/22528/diff/ Testing --- works on linux and windows. Thanks, Chug Rolke
