Closed Bug 1344171 Opened 3 years ago Closed 3 years ago
Improve connection management
1) if a connection is read, i.e. socket is connected, and we can attach a pending transaction or a NullTransaction, for https connections, it can happen that the connection is put into idle queue even though it could take a NullTransaction to perform tls. This can happen when the transaction pending queue is not empty but all transaction in the queue are content-blocked. 2) If transaction takes a speculatively open halfOpenSocket, the halfOpenSocket will not backup socket as in other cases. 3) If a halfOpenSocket is opened for a transaction, but the transaction takes another connection, the halfOpenSocket cannot be claimed by another transaction.
Assignee: nobody → dd.mozilla
Status: NEW → ASSIGNED
Comment on attachment 8844421 [details] [diff] [review] bug_1344171_v1.patch Review of attachment 8844421 [details] [diff] [review]: ----------------------------------------------------------------- oddly, this seems simpler. thanks!
Attachment #8844421 - Flags: review?(mcmanus) → review+
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/1f0db7810a84 Improve connection management. r=mcmanus
Backed out for bustage (unused variable 'claimed'): https://hg.mozilla.org/integration/mozilla-inbound/rev/0afa705631ce000b3eef48deef5d618b510a2fc0 Push with failures: https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=8646978d35260c7a796f776d5d03b9d62828bbaf&filter-resultStatus=testfailed&filter-resultStatus=busted&filter-resultStatus=exception&filter-resultStatus=retry&filter-resultStatus=usercancel&filter-resultStatus=runnable Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=85117771&repo=mozilla-inbound [task 2017-03-20T19:25:59.855081Z] 19:25:59 INFO - In file included from /home/worker/workspace/build/src/obj-firefox/netwerk/protocol/http/Unified_cpp_protocol_http1.cpp:110:0: [task 2017-03-20T19:25:59.855394Z] 19:25:59 INFO - /home/worker/workspace/build/src/netwerk/protocol/http/nsHttpConnectionMgr.cpp: In member function 'nsresult mozilla::net::nsHttpConnectionMgr::CreateTransport(mozilla::net::nsHttpConnectionMgr::nsConnectionEntry*, mozilla::net::nsAHttpTransaction*, uint32_t, bool, bool, bool, mozilla::net::nsHttpConnectionMgr::PendingTransactionInfo*)': [task 2017-03-20T19:25:59.856385Z] 19:25:59 INFO - /home/worker/workspace/build/src/netwerk/protocol/http/nsHttpConnectionMgr.cpp:1826:14: error: unused variable 'claimed' [-Werror=unused-variable] [task 2017-03-20T19:25:59.856581Z] 19:25:59 INFO - bool claimed = sock->Claim(); [task 2017-03-20T19:25:59.856760Z] 19:25:59 INFO - ^ [task 2017-03-20T19:25:59.856949Z] 19:25:59 INFO - cc1plus: all warnings being treated as errors [task 2017-03-20T19:25:59.857182Z] 19:25:59 INFO - /home/worker/workspace/build/src/config/rules.mk:1016: recipe for target 'Unified_cpp_protocol_http1.o' failed [task 2017-03-20T19:25:59.857359Z] 19:25:59 INFO - gmake: *** [Unified_cpp_protocol_http1.o] Error 1
Pushed by email@example.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/c040782ca691 Improve connection management. r=mcmanus
This appears to have caused a frequent but intermittent xpcshell crash on Win7VM debug: https://treeherder.mozilla.org/logviewer.html#?job_id=85455754&repo=mozilla-inbound Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/4c45a7779491
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/d11fb28f9b9c Improve connection management. r=mcmanus
Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/552267771708 for pretty much the same Win7 test_altsvc.js crashes, e.g. https://treeherder.mozilla.org/logviewer.html#?job_id=87151902&repo=mozilla-inbound (the rest of the failures after test_altsvc.js are unrelated, just other bustage from another busted patch in the same push).
(In reply to Phil Ringnalda (:philor) from comment #13) > Backed out in > https://hg.mozilla.org/integration/mozilla-inbound/rev/552267771708 for > pretty much the same Win7 test_altsvc.js crashes, e.g. > https://treeherder.mozilla.org/logviewer.html#?job_id=87151902&repo=mozilla- > inbound (the rest of the failures after test_altsvc.js are unrelated, just > other bustage from another busted patch in the same push). Thank you. So sorry this was my mistake, I am on IETF and a bit distracted. sorry!
Pushed by email@example.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/364acbf6a069 Improve connection management. r=mcmanus
You need to log in before you can comment on or make changes to this bug.