Open obulkin opened 1 year ago
Thanks for the report. Yes, that makes sense, a PR would be helpful.
Excellent. I'll get started on one.
@eregon PR is up now: https://github.com/ruby-concurrency/concurrent-ruby/pull/1009
The CI failure seems like a flaky test since the test case doesn't seem related to my changes and passed under JRuby.
@eregon Checking in on this. I recommitted my changes to force a new test run, and everything is green this time.
Checking in. Please let me know if I can do anything to move this along. The PR is very minimal and should be quick to review.
@eregon Bumping again
Merged. Do you need a release?
Thanks! No urgent need for a release if there's something else in the pipeline that you'd prefer to bundle it with.
Background Info
concurrent-ruby
version:>= 1.1.10
, earlier versions may be affected as wellconcurrent-ruby-ext
installed: noconcurrent-ruby-edge
used: noIssue
Concurrent::DaemonThreadFactory#newThread
callsdefaultThreadFactory()
each time it generates a new thread, which creates a new JavaThreadFactory
object each time. This is clearly at odds with intendedThreadFactory
usage and is (potentially) problematic in a few ways:pool-X-thread-Y
, where Y is always 1 while X is different for each thread. This is confusing as it suggests that the Concurrent Ruby thread pool is not actually pooling and reusing threads.ThreadFactory
instance is retained somewhere, the result will be a memory leak.ThreadFactory
instance normally belong to the same Java thread group, which has permissions implications and may cause errors in code that expects all the threads in a Concurrent Ruby thread pool to belong to the same Java thread group.Proposed Solution
Call
defaultThreadFactory()
once inConcurrent::DaemonThreadFactory#initialize
, store the resulting factory in an instance variable, and reuse it as needed inConcurrent::DaemonThreadFactory#newThread
.Happy to create a PR with the fix if this seems like an acceptable solution.