[
https://issues.apache.org/jira/browse/THRIFT-1187?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ilya Maykov updated THRIFT-1187:
--------------------------------
Summary: nonblocking_server shutdown race under Ruby 1.9 (was:
nonblocking_server shutdown race under in Ruby 1.9)
> nonblocking_server shutdown race under Ruby 1.9
> -----------------------------------------------
>
> Key: THRIFT-1187
> URL: https://issues.apache.org/jira/browse/THRIFT-1187
> Project: Thrift
> Issue Type: Bug
> Components: Ruby - Library
> Affects Versions: 0.6.1
> Environment: OS X Darwin 10.6 i686 + Ruby 1.9.2-p180
> Reporter: Ilya Maykov
> Priority: Minor
> Attachments: patch-THRIFT-1187.txt
>
>
> While testing some changes today I noticed that nonblocking_server tests pass
> for me under Ruby 1.8.7 but fail under Ruby 1.9.2. Here are the test failures:
> ...
> 1)
> Errno::EBADF in 'Thrift::NonblockingServer should shut down when asked'
> Bad file descriptor
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `select'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `block in serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `loop'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/spec/nonblocking_server_spec.rb:118:in
> `block (3 levels) in <class:ThriftNonblockingServerSpec>'
> 2)
> Errno::EBADF in 'Thrift::NonblockingServer should continue processing active
> messages when shutting down'
> Bad file descriptor
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `select'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `block in serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `loop'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/spec/nonblocking_server_spec.rb:118:in
> `block (3 levels) in <class:ThriftNonblockingServerSpec>'
> 3)
> Errno::EBADF in 'Thrift::NonblockingServer should kill active messages when
> they don't expire while shutting down'
> Bad file descriptor
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `select'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `block in serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `loop'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/spec/nonblocking_server_spec.rb:118:in
> `block (3 levels) in <class:ThriftNonblockingServerSpec>'
> 4)
> Errno::EBADF in 'Thrift::NonblockingServer should allow shutting down in
> response to a message'
> Bad file descriptor
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `select'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:47:in
> `block in serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `loop'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/lib/thrift/server/nonblocking_server.rb:45:in
> `serve'
> /Users/ilyam/src/vendor/thrift-src/thrift-0.6.1-r19/lib/rb/spec/nonblocking_server_spec.rb:118:in
> `block (3 levels) in <class:ThriftNonblockingServerSpec>'
> ...
> I *think* that either the behavior of select() changed between 1.8.7 and
> 1.9.2, or the different threading model in 1.9.2 is tickling a timing bug. In
> any case, shutting down the server causes the @server_transport socket to be
> closed, which raises Errno::EBADF in the main loop inside the serve() method.
> Probably not a big deal because the server is shutting down anyway, but
> should be fixed so the tests pass.
> Will attach patch shortly.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
