release/3.4.12 minor

[djspiewak]

• Implement SleepCallback
• Implement SleepersQueue
• Add a cancelation mechanism to SleepCallback
• Complicate the parking mechanism a bit
• Pretend to handle sleeping fibers
• Add a sleep method on WorkerThread
• Implement the sleep method on WorkStealingThreadPool
• Simplify canceled sleepers handling
• Worker threads need to fight for their awakening
• Wire up the new sleeping mechanism in the run loop
• Restore the correct thread pool state after sleeping
• Rename the internal sleep method
• Add a sleep cancelation unit test
• Add a link to the openjdk source code
• Add SleepBenchmark
• Check ownership of WorkerThread before sleeping
• WorkStealingThreadPool extends Scheduler
• Switch the default scheduler to the compute pool
• Add a unit test for foreign execution contexts
• Directly complete the sleep async callback
• Re-use the RightUnit instance
• Highlight uncancelable behavior in MonadCancel#onCancel scaladoc
• Tweak Async#async and Async#async_ scaladocs about uncancelable behavior
• Add Async#asyncPoll for Consider an Async utility to encapsulate double-check registration #3087
• Format sources with scalafmt
• Add three more AsyncLaws tests and spec for basic Async implementation
• Fix scalafix
• Fix Async#async scaladoc
• Update sbt-scalafix to 0.10.4 in series/3.x
• Taking a quick stab at changing async_'s cancelation semantics
• Update scalafmt-core to 3.6.0 in series/3.x
• Update sbt-java-formatter to 0.8.0 in series/3.x
• Add IOLocal#scope
• Update sbt to 1.7.3 in series/3.x
• Update scalafmt-core to 3.6.1 in series/3.x
• Introduce a BatchingMacrotaskExecutor
• Rename 'asyncPoll => asyncCheckAttempt', sync 'Async/IO' scaladoc
• Update sbt to 1.8.0 in series/3.x
• Run scalafmtAll
• never needs to be explicitly cancelable
• This definitely bumps the base version...
• Update specs2-core, specs2-scalacheck to 4.19.0 in series/3.x
• a Defer instance for Resource without Sync
• Defer Spec
• Replace IO with Almost-IO in Defer test
• fix comment
• organize imports
• Use PureConc for Resource Tests
• remove excess import (somehow missed by scalafix)
• hacking around to find instances for PureConc in 2.12
• Update sbt-scalajs, scalajs-compiler, ... to 1.11.0 in series/3.x
• Update tests/shared/src/test/scala/cats/effect/ResourceSpec.scala
• rename var
• Redefinition of never
• Fixed cancelation leaks in Resource.cont
• Scalafmt
• Sigh... scalafix
• Removed references to external IORuntime from sleep unhappy path
• Fixed memory leak in sleep cancelation
• Fixed reference to global executor
• A combinator to flatten a modification of a Ref.
• sbt prePR
• This definitely bumps the base version
• Disable starvation checker in SleepDrift test
• Removed errant println
• Fixed blockOn mechanism for integrated timers
• Moved WSTP tests into platform spec
• Removed unused import
• Update nscplugin, sbt-scala-native, ... to 0.4.9 in series/3.x
• Update scala-java-time to 2.5.0 in series/3.x
• Use batching only for IOFiber#schedule
• Fix BatchingMacrotaskExecutor implementation
• Wip test spec
• Rearrange platform checks
• Make batching ec fiber-aware
• Restore deque-based implementation
• Fix bugs
• Add comments to spec
• Fix "limit batch sizes" test
• Organize imports
• Add MiMa filter
• Comments and scaladocs
• Another mima filter
• More efficient fiber monitoring
• Update comment
• Add IO#voidError
• Fix compile
• Add IO#supervise and IO#toResource.
• Add IO#metered.
• Ensure flatModify is atomic in that if the Ref updates then the effect will be run.
• Improve documentation.
• Improve formatting.
• Revert to the existing implementation and improve the documentation.
• Use asyncCheckAttempt in IODeferred#get
• Add 'IOLocal.lens' method to produce lens 'IOLocal[B]' from getter and setter
• fix tests under Scala 3
• add 'IOLocal.lens' to core MIMA exceptions (IOLocal is sealed)
• Generalize IOLocalSpec tests for IOLocal[A] instances
• comment about lens composition impl details
• move common methods to 'IOLocal', check 'IOLocal.lens' stack safety
• Update sbt to 1.8.2 in series/3.x
• Remove flatModify entirely due to the controversy it causes pertaining to cancelability.
• Use Concurrent for AtomicCell
• Improve Mutex implementation
• Update sbt-typelevel to 0.5.0-M9 in series/3.x
• Attempt to improve LockChain
• Add a Concurrent based implementation of AtomicCell
• Re-Add Concurrent based Mutex implementation
• Add 'fromFutureCancelable` and friends
• Fix test
• Make executeBatchTask an object
• Make link-time condition elidable
• Expose config for BatchingMacrotaskExecutor
• First try at JSArrayQueue
• Add JSArrayQueueSpec
• isEmpty() is side-effectful
• Bug fix
• Use JSArrayQueue in batching EC
• Directly use queueMicrotask
• Rollback Scala.js bump
• Update scalafmt-core to 3.7.1 in series/3.x
• Reformat with scalafmt 3.7.1
• Add 'Reformat with scalafmt 3.7.1' to .git-blame-ignore-revs
• Sketch of thread blocking detection
• Add TODO
• Don't select current thread for block detection
• Account for other blocked thread states and intentional thread parking
• Better scaladoc
• Typo
• Fix benchmarks
• Add mimia exclusion
• Make stub test trivially pass on JS
• Cache executeBatchTaskJSFunction
• Protect timers against Long overflow
• Update Resource.scala comments to clarify that release is only run when acquire succeeds
• Fix bincompat warnings (I think...)
• Register JS test
• scalafmt once again
• Apparently I can't hide old version yet
• Deprecate old workstealing threadpool constructor method
• Don't run test on JS
• scalafix :eye-roll:
• Don't run blocking check if pool is of size 1
• IOLocal - generalize scope function
• Update Scala 3, Scala.js, Scala Native
• Make helpers private
• Remove AsyncAtomicCell
• Make helpers private
• Re-add AsyncAtomicCell
• Fix AtomicCell example
• Add AtomicCell benchmark
• Add Mutex benchmark
• Remove IOLocal#scope, revert Add IOLocal#scope #3214
• Add Mutex cancellation benchmark
• Fix benchmarks
• Add circle ci config
• Rename job
• Fix config
• Run testsNative/test specifically
• Set no_output_timeout
• Try setting up Cirrus
• More memory
• More generic CI detection
• Fix sbt command, try less memory
• Add 'flatModify', 'flatModifyFull' and corresponding 'State' methods to 'Ref'
• Fix formatting
• Try moving everything to Cirrus
• Try installing clang
• Try a custom Dockerfile
• Less CPUs
• Run apt-get update
• Extract UnsafeUnbounded
• Implement async mutex with UnsafeUnbounded
• Add test for canceled waiters
• Fix mutex release bug
• Add test for deadlock under contention
• Fix release race condition
• Tweak mutex happy path benchmark
• MiMa filters
• Swap parameter order
• Simplify branching logic for enabling thread blocking detection
• Avoid looping while looking for another thread to check blocked status
• Fix benchmarks
• Revert access change to workerThreads field
• Fix naming ambiguity for scala 3
• Restore bounds check
• Make FailureSignal an object, prefetch it
• A slightly wider margin on real-time tests
• Use shadowing
• Update sbt-mdoc to 2.3.7 in series/3.x
• Update specs2-core, specs2-scalacheck to 4.19.2 in series/3.x
• Update sbt-jmh to 0.4.4 in series/3.x
• Add docs highlighting when to use thread blocking detection
• Note that thread blocking detection is relatively expensive
• Add missing import
• Update headers
• Fix git merge fail (???)
• Formatting
• Unified scheduler init defaults with IOApp
• Nix shell -> Nix flake
• Fixed loop when timers expire with an empty external queue
• Fix NPE in blocked thread detection
• Further optimize IODeferred
• Added some extra test comments
• Update flake.lock to get singleton LLVM_BIN
• Corrected handling of self-cancelation within timeout
• Updated Resource#race to match corrected semantics
• Expand Ref#flatModify doc and use where appropriate
• Use Ref#flatModifyFull in some places
• Expand flatModifyFull doc
• Update scalafmt-core to 3.7.2 in series/3.x
• Added cancelable
• Added FAQ entry describing canceling delay/blocking
• Added missing parentheses
• Added a few more words to cancelable faq
• Added caveat regarding "proper" approach to FileInputStream
• Swapped IO(...) for IO.blocking(...)
• Make readLine cancelable via stdin reader thread
• Add a test
• Formatting
• flake.lock: Update
• Test that reads with wrong charset are rejected
• Setup dependabot for Docker CI image
• Don't shadow println in IOFiber
• Fix typo in docs
• Fix IORuntimeBuilder failureReporter config on JS
• Make HotSwap safe to concurrent access
• Add test for concurrent release
• Add MiMa filter
• Add test for concurrent access
• Fix concurrent access to current resource
• Allow that the renamed blocker thread is terminated
• Update sbt-typelevel to 0.5.0-M10 in series/3.x
• Fix "support re-enablement via cancelable" test
• flake.lock: Update
• add console as config in ioRuntimeConfig, pass it to CPUStarvation
• Update scaladoc
• Add TimerSkipList stub
• Use the stub in the WSTP
• Add TimerSkipList (ref 9e5fec4f)
• scalafmt
• Fix bug when parking with an expired timer
• Remove unused code, port some existing tests
• scalafix, scalafmt
• Add skip list tests
• Add parallel skip list tests
• Add test with a lot of IO.sleep
• Add a JCStress test
• Try to run JCStress test in CI
• provide a function to process metrics manually, defaulting to console logging
• fix scaladoc
• Smaller test size in js and native
• Update nscplugin, sbt-scala-native, ... to 0.4.11 in series/3.x
• Stealing timers
• Even smaller test size in js
• Kick CI
• Decrease iterations (CI could just be slow)
• Add more JCStress tests
• Reduce indirections by using AtomicReferenceFieldUpdater
• javafmt
• Try to fix java warnings
• Kick CI
• Kick CI
• Increase durations for ContSpec
• Use realWithRuntime instead of new Scheduler
• flake.lock: Update
• More JCStress tests
• Try to fix JCStress tests
• Use unique names for Cirrus CI jobs
• Update nscplugin, sbt-scala-native, ... to 0.4.12 in series/3.x
• Another JCStress test
• Cleanup
• Remove long stress test (it's too much for CI)
• Additional sleep benchmark
• Handle sleep in a smarter way
• Kick CI
• flake.lock: Update
• Update scalafmt-core to 3.7.3 in series/3.x
• Add tests for Random class
• Formatting
• AtomicCell#get should not semantically block
• Doc updates
• Release gate in evalModify
• Update scaladoc
• Refactor tests to use replicateA and no-op assert
• flake.lock: Update
• Update sbt-scala-native-crossproject, ... to 1.3.0 in series/3.x
• Add test for timer stealing
• Update sbt-scalajs, scalajs-compiler, ... to 1.13.1 in series/3.x
• flake.lock: Update
• Avoid compiler error using randomBoolean, int, long
• Abstract check in range and shuffle Sequence
• Abstract tests
• TestRandom with different Randoms
• Fix bincompat
• Fix get performance regression
• Remove unnecessary cast
• Update sbt-scala-native-crossproject, ... to 1.3.1 in series/3.x
• Move racing sleep test to JVM
• Add failing test
• Fix mutex cancelled acquire
• scalafmt
• Add Apple Silicon to Native CI matrix
• Add high-core JVM Apple Silicon to CI matrix
• Modify test to reproduce In theory, waiters can be lost in Mutex #3554
• Fix mutex cancelled acquire even more
• Fix problem with nextGaussian test
• Decrease test timeout
• Fewer iterations on JS/native
• Formatting
• Make sure a timeout fails the test
• fromFuture => fromFutureCancelable
• 1 minute is not enough in CI
• flake.lock: Update
• Decrease iterations because CI is slow
• Revert Mutex implementation
• Replace isFinite with java.lang.Double.isFinite
• Add Mutex test: handle multiple concurrent cancels during release
• Fix Mutex.AsyncImpl for the new concurrent test
• Add Mutex test: preserve waiters order (fifo) on a non-race cancellation
• maybe ticked won't be flaky?
• Revert "maybe ticked won't be flaky?"
• Add Mutex test: cancellation must not corrupt Mutex
• hear me out: what if we make it differently flaky
• Revert AsyncMutex implementation to the LockChain based one
• Make blockedThreadDetectionEnabled configurable via a system property
• PR feedback
• wip
• Add new AsyncMutex implementation
• Make AsyncMutex.LockQueueCell an AtomicReference itself
• Fix 'discarded non-Unit value of type Boolean' warning
• Reduce the number of iterations on MutexSpec.handle multiple concurrent cancels during release
• Use deferred as the only implementation for Mutex
• Remove wrapper allocation on Mutex.ConcurrentImpl.LockQueue
• Update specs2-core, specs2-scalacheck to 4.20.0 in series/3.x
• Setup CI release
• Make steal timers test less fragile
• dispatcher releasing itself rejects new tasks
• Move action to Typelevel
• Make TimerSkipListIOSpec faster
• only check submit is left if result is false
• Restore a tiny bit of type information for IOFiber#resumeIO
• Remove unnecessary casts to IO[Any]
• If we stole something, let's work on it
• stealTimers returns true iff it stole
• flake.lock: Update
• Reduced iteration count to placate CI
• Remove unnecessary timeoutTo
• Use isJVM
• move onCpuStarvationWarn a method
• fix flakey dispatcher test
• Allow GC of ConcurrentMutexImpl.LockQueueCell during cancellation
• Rename ConcurrentMutexImpl variables
• Update sbt-native-image to 0.3.4 in series/3.x
• Revised Queue.synchronous internals to simplify concurrent hand-off
• Use configure-pagefile-action on Windows CI
• Added test to ensure inner worker awaits are canceled
• Make the new dispatcher release test deterministic
• Moved common test to common
• Update tests/shared/src/test/scala/cats/effect/std/DispatcherSpec.scala
• Add test for global runtime reset
• Reset global runtime when it is shutdown
• Register GlobalShutdown test app on JS
• Adjusted Queue.synchronous to include a two-phase commit
• Scalafmt
• Fixed race condition between offer and canceled take
• flake.lock: Update
• Fixed compilation on 2.12
• Reduce iteration count on non-JVM
• Added tailrec
• Add test for stray exceptions in IO.uncancelable
• Handle stray exceptions in uncancelable body
• Micro-optimize onError
• Add test for negative sleeps
• Treat non-positive sleep durations as cedes
• Extact delay local val
• Fixed issue in which failing uncancelable would remain masked for one stage
• Drafted version update script
• Automation for the minor PR creation
• Update versions for 3.4.12

[djspiewak]

Pls ignore. Testing