)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":4257,"name":"Zane Bitter","email":"zbitter@redhat.com","username":"zaneb"},"change_message_id":"e239cf6ef2cddf09aa06e25f9201149f335931bf","unresolved":false,"context_lines":[{"line_number":4,"context_line":"Commit:     Claudiu Belu \u003ccbelu@cloudbasesolutions.com\u003e"},{"line_number":5,"context_line":"CommitDate: 2018-08-21 16:34:05 +0000"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Skips signal handling on Windows"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Setting up signal handling fails on Windows due to missing select"},{"line_number":10,"context_line":"module attributes. This patch skips the setup on Windows."}],"source_content_type":"text/x-gerrit-commit-message","patch_set":3,"id":"3f79a3b5_95715379","line":7,"updated":"2018-08-22 20:23:02.000000000","message":"This is slightly misleading, in that we still set up signal handling in add_handler(). The difference is that we don\u0027t set up the code that raises exceptions with EINTR if a signal handler is called during a previously-interruptable syscall.","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"}],"oslo_service/service.py":[{"author":{"_account_id":4257,"name":"Zane Bitter","email":"zbitter@redhat.com","username":"zaneb"},"change_message_id":"e239cf6ef2cddf09aa06e25f9201149f335931bf","unresolved":false,"context_lines":[{"line_number":204,"context_line":"        a signal from interrupting eventlet\u0027s call to poll() or sleep()."},{"line_number":205,"context_line":"        \"\"\""},{"line_number":206,"context_line":"        self.__force_interrupt_on_signal \u003d (sys.version_info \u003e\u003d (3, 5) and"},{"line_number":207,"context_line":"                                            sys.platform !\u003d \u0027win32\u0027)"},{"line_number":208,"context_line":""},{"line_number":209,"context_line":"        if self.__force_interrupt_on_signal:"},{"line_number":210,"context_line":"            try:"}],"source_content_type":"text/x-python","patch_set":3,"id":"3f79a3b5_daf4daf3","line":207,"range":{"start_line":207,"start_character":44,"end_line":207,"end_character":67},"updated":"2018-08-22 20:23:02.000000000","message":"The condition that\u0027s used inside eventlet when deciding whether to import eventlet.hubs.poll is hasattr(select, \u0027poll\u0027). It might be more consistent to use that - it\u0027s at least potentially more portable.","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"},{"author":{"_account_id":8213,"name":"Claudiu Belu","email":"cbelu@cloudbasesolutions.com","username":"claudiub"},"change_message_id":"136f2b8515d09c511ba63f0ca634554e69751f3f","unresolved":false,"context_lines":[{"line_number":204,"context_line":"        a signal from interrupting eventlet\u0027s call to poll() or sleep()."},{"line_number":205,"context_line":"        \"\"\""},{"line_number":206,"context_line":"        self.__force_interrupt_on_signal \u003d (sys.version_info \u003e\u003d (3, 5) and"},{"line_number":207,"context_line":"                                            sys.platform !\u003d \u0027win32\u0027)"},{"line_number":208,"context_line":""},{"line_number":209,"context_line":"        if self.__force_interrupt_on_signal:"},{"line_number":210,"context_line":"            try:"}],"source_content_type":"text/x-python","patch_set":3,"id":"3f79a3b5_b5046116","line":207,"range":{"start_line":207,"start_character":44,"end_line":207,"end_character":67},"in_reply_to":"3f79a3b5_daf4daf3","updated":"2018-08-24 10:24:04.000000000","message":"yeah, that works too.","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"},{"author":{"_account_id":4257,"name":"Zane Bitter","email":"zbitter@redhat.com","username":"zaneb"},"change_message_id":"e239cf6ef2cddf09aa06e25f9201149f335931bf","unresolved":false,"context_lines":[{"line_number":208,"context_line":""},{"line_number":209,"context_line":"        if self.__force_interrupt_on_signal:"},{"line_number":210,"context_line":"            try:"},{"line_number":211,"context_line":"                from eventlet.hubs import poll as poll_hub"},{"line_number":212,"context_line":"            except ImportError:"},{"line_number":213,"context_line":"                pass"},{"line_number":214,"context_line":"            else:"}],"source_content_type":"text/x-python","patch_set":3,"id":"3f79a3b5_9aac22ee","line":211,"updated":"2018-08-22 20:23:02.000000000","message":"I think this is the issue; when I wrote this code I assumed that we didn\u0027t support Windows. This is a hub that may be used on Linux if epoll is not available, but on Windows it\u0027s eventlet.hub.selects IIUC.\n\nIf the PEP475 issue also affects Windows (no idea if that is the case) then we\u0027d also need to patch sleep in the selects hub. Unfortunately in that module it\u0027s not stored locally, but fetched each time from the (original) time module, so it\u0027d be a slightly different implementation.\n\nWe also have a similar issue with the kqueue hub on BSD, if we care about BSD. (That has the same implementation as poll though, as far as sleep goes.)","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"},{"author":{"_account_id":8213,"name":"Claudiu Belu","email":"cbelu@cloudbasesolutions.com","username":"claudiub"},"change_message_id":"136f2b8515d09c511ba63f0ca634554e69751f3f","unresolved":false,"context_lines":[{"line_number":208,"context_line":""},{"line_number":209,"context_line":"        if self.__force_interrupt_on_signal:"},{"line_number":210,"context_line":"            try:"},{"line_number":211,"context_line":"                from eventlet.hubs import poll as poll_hub"},{"line_number":212,"context_line":"            except ImportError:"},{"line_number":213,"context_line":"                pass"},{"line_number":214,"context_line":"            else:"}],"source_content_type":"text/x-python","patch_set":3,"id":"3f79a3b5_752329a8","line":211,"in_reply_to":"3f79a3b5_9aac22ee","updated":"2018-08-24 10:24:04.000000000","message":"oslo.service is not 100% Windows-compatible, but some of its consumers are (nova-compute, neutron-openvswitch-agent, cinder-volume, cinder-backup, etc.).","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"},{"author":{"_account_id":4257,"name":"Zane Bitter","email":"zbitter@redhat.com","username":"zaneb"},"change_message_id":"e239cf6ef2cddf09aa06e25f9201149f335931bf","unresolved":false,"context_lines":[{"line_number":236,"context_line":"                poll_hub.sleep \u003d sleep_wrapper"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"            hub \u003d eventlet.hubs.get_hub()"},{"line_number":239,"context_line":"            self.__hub_module_file \u003d sys.modules[hub.__module__].__file__"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"    def _handle_signal_cb(self, signo, frame):"},{"line_number":242,"context_line":"        for handler in self._signal_handlers[signo]:"}],"source_content_type":"text/x-python","patch_set":3,"id":"3f79a3b5_55e09b26","line":239,"updated":"2018-08-22 20:23:02.000000000","message":"The good news is there\u0027s no do_poll() method to be patched in the selects hub, so skipping this entire method is no different in practice to skipping just the part causing the problem.\n\nThe bad news is that if the PEP475 issue also affects Windows then we\u0027d want to detect when the signal handler is being called from inside select.select() as well as time.sleep().","commit_id":"d4914b1f0a59481e002237ed8bc72b5f4d92a80e"}],"oslo_service/tests/test_service.py":[{"author":{"_account_id":6928,"name":"Ben Nemec","email":"openstack@nemebean.com","username":"bnemec"},"change_message_id":"069df2deddf5fc4e65f3e7bc95c9890e3ab00e1b","unresolved":false,"context_lines":[{"line_number":487,"context_line":"        signal_handler \u003d service.SignalHandler()"},{"line_number":488,"context_line":"        self.assertFalse("},{"line_number":489,"context_line":"            signal_handler._SignalHandler__force_interrupt_on_signal)"},{"line_number":490,"context_line":"        service.SignalHandler.__class__._instances.clear()"},{"line_number":491,"context_line":""},{"line_number":492,"context_line":"    @mock.patch(\u0027signal.alarm\u0027)"},{"line_number":493,"context_line":"    @mock.patch(\"os.kill\")"}],"source_content_type":"text/x-python","patch_set":2,"id":"3f79a3b5_89abad1b","line":490,"updated":"2018-08-21 16:28:03.000000000","message":"If the assert above were to fail, this would never get run and I imagine a lot of other tests would fail as a result.  Should we maybe make this part of the class\u0027s setUp and cleanup so we\u0027re always starting with a clean slate?  Or maybe it would be sufficient to just do self.addCleanup(...instances.clear) early in this function.  I don\u0027t know that I\u0027ve seen addCleanup used outside setUp, but I also don\u0027t see anything to suggest it won\u0027t work either.","commit_id":"21eea60c6be9e4451f3b0263bafb0602b66b3318"},{"author":{"_account_id":8213,"name":"Claudiu Belu","email":"cbelu@cloudbasesolutions.com","username":"claudiub"},"change_message_id":"99ca24acffc6e2b3b76efbaa837f714958a8a895","unresolved":false,"context_lines":[{"line_number":487,"context_line":"        signal_handler \u003d service.SignalHandler()"},{"line_number":488,"context_line":"        self.assertFalse("},{"line_number":489,"context_line":"            signal_handler._SignalHandler__force_interrupt_on_signal)"},{"line_number":490,"context_line":"        service.SignalHandler.__class__._instances.clear()"},{"line_number":491,"context_line":""},{"line_number":492,"context_line":"    @mock.patch(\u0027signal.alarm\u0027)"},{"line_number":493,"context_line":"    @mock.patch(\"os.kill\")"}],"source_content_type":"text/x-python","patch_set":2,"id":"3f79a3b5_5460c8a5","line":490,"in_reply_to":"3f79a3b5_89abad1b","updated":"2018-08-21 16:32:33.000000000","message":"good point. self.addCleanup should be sufficient, IMO.","commit_id":"21eea60c6be9e4451f3b0263bafb0602b66b3318"}]}
