ASCO 0.3.0 Pre-release

A new remake version

Full Changelog: https://github.com/pointertobios/asco/commits/0.3.0-pre

ASCO 0.2.1

Fixes

• Segfault when destroy cloned coroutines
• File IO options sometime blocked
• SLUB sometime allocated nullptr

Full Changelog: 0.2.0...0.2.1

ASCO 0.2.0

scheduled unstable

Features

• inline_future.dispatch()
• Type safety flags
• True void type support for future
• placement_list
• block_read_writer
• SLUB memory pool

Fix

• Worker maps use unordered_map
• File IO randomly crash
• Some other randomly crashing

Improvement

• Use spin instead of mutex in inner::channel
• Add noexcept for spin and rwspin
• Remove customizable runtime
• Normalize file API
• New member in buffer
• On performance, compactibility and also correcting some misusing
• More template extern specialization
• Improve many techical debt

Full Changelog: 0.1.1...0.2.0

ASCO 0.1.1

Features

• Exception asco::coroutine_abort
• New future.exceptionally definition
• Target for choice of async main function or sync main function
• SSO for zero copy buffer

Improvements and Fixes

• Monopolized timer and io thread
• Add multi producer single consumer channel for inner util, and use it for task dispatching

Full Changelog: 0.1.0...0.1.1

ASCO 0.1.0

Features

• Timer library: timeout and sleeps
• Exception with stacktrace and call chain trace
• Synchronous primitives: condition_variable, barrier and rwlock
• Simple performance recording
• Preparation for IO module: Zero copy buffer and stream_base
• Asynchronous task combination: future.then() and future.exceptionally()

Improvements and Fixes

• If select spawned to the same worker, the tasks will block because barrier. We don't need all other coroutine exited when the first selected coroutine return. Just remove the barrier.
• select went into exception and not correctly suspend and awake task.
• select didn't correctly remove caller coroutine.
• Use channel to make sure the return value of coroutine didn't destructed before caller co_await it.
• Processing of exception improved.
• Use spin lock instead of mutex in runtime and scheduler. Their critical sections are short enough.
• semaphore::acquire() improvement: break out of CAS loop after suspend.
• If sub coroutine had not been co_await, once the caller exit, the stacktrace in sub coroutine will crash due to segfault. Now move trace info to coro_local_frame.

License change to MIT from GPLv3

Full Changelog: 0.0.3...0.1.0

ASCO 0.0.3

Features

• Interval
• select<N>: Fetch return value of the first returned coroutine and abort rest coroutines
• Recursing abort

Improvements & Fixes

• Exception passing correctly

Full Changelog: 0.0.2...0.0.3

ASCO 0.0.2

Features

• Channel(Pipe): Simple Producer Simple Consumer
• Coroutine Local Frame Lazy Destroy

Fixes

• Semaphore incorrectly sleep and never been awaken.

Full Changelog: 0.0.1...0.0.2

ASCO 0.0.1

Basic asynchronous environment

• Peer coroutines: Crated by scheduler and work as partner with each other.
• Inline coroutines (Non-peer coroutines): Calling the callees and returning to caller with symmetrical transformation.
• Abortable coroutines: future<T>::abort() function with recover semantic. The aborted coroutine must recover the affected object to the state before the coroutine started.
• Coroutine local variables: Variables with its lifetime spreading with the coroutine calling chain.

Synchronization primitives

• Semaphore