v1.1.5: Minor future-proofing tweaks

This release just removes some constructs which technically provoke undefined behaviour, includes a tiny performance/memory use optimisation, and a little bit of refactoring.

v1.1.4: Usability and documentation fixes

This is a minor release with some fixes/updates to documentation, and the lambda-based "add_timer" function (previously undocumented) now takes const timespec & arguments (which allows using time_val values).

• document the pre-existing "add_timer() with lambda" function
• change "add_timer()" (with lambda) to accept const timespec & arguments (previously the parameter types were non-const).
• fix broken timer example in introduction (USAGE.md)
• various internal changes.

v1.1.3: Bugfix and CMake support

This is primarily a bugfix release, but includes CMake support files (installed alongside the library with "make install"). Note that CMake is not required to build.

• fix timers not working on non-Linux systems
• Includes CMake files (makes it easier for projects using CMake to use Dasynq)
• build fixes for platforms without kqueue or epoll
• install fixes for BSDs

v1.1.2: Build fixes for FreeBSD/OpenBSD

Unfortunately the previous release had build issues, hence this immediate re-release.

v1.1.1: Bugfix release.

This release fixes a bug which prevented "emulated" fd watches from working properly. It also includes a workaround for a MacOS kernel bug which prevented signals from being reliably detected using the kqueue backend on that platform.

v1.1.0: Now we are POSIX

This release includes bug fixes for multi-thread event loops using the kqueue backend, and also adds a new backend which uses pselect(2). Since pselect is mandated by POSIX this means Dasynq should work on nearly all mostly-POSIX-compliant systems.

(Note however that pselect is probably not going to be great performance wise, and suffers from an inherent inability to deal with file descriptors beyond a certain number).

v1.0.4: "Too early, too often"

This release fixes a silly bug with subtraction of time values which gives the wrong result for times with exactly equal nanoseconds. While this probably would strike only rarely, it's a bug best squashed as early as possible, and so I'm drafting this release.

v1.0.3: "Release early, release often"

This is a bug-fix for an issue that crept into v1.0.2, the previous bug-fix release.

v1.0.2: Now with even more sparkles

This is a bug-fix release, squashing a bug in the heap implementation which could cause watchers to fire out-of-sequence (according to priority) and timers to fire at the wrong times. The test suite is more complete and catches the bugs fixed since 1.0.

v1.0.1: The inevitable bug-fix release

Fix a number of issues discovered since the 1.0 release, some minor, some not-so-minor. Please upgrade.