global: Fix spelling mistakes in comments Original work by @andreasschulze and @jsoref

global: stop including fd-set-nonblock.h & fd-close-on-exec.h directly

lib: ioloop-kqueue - Improve kevent() panic log message

lib: Fix kqueue io_loop_get_wait_time usage The code needs to take into consideration, when there is no IO to be waited, but possibly there is io_set_pending used, or just timeouts.

kqueue: Changed all i_fatal() calls to i_panic()s to make debugging them possible.

ioloop: Added io_set_pending()

ioloop-kqueue: Added extra assert.

ioloop-kqueue: Added assert

Array API changed: ARRAY_DEFINE(name, type) -> ARRAY(type) name Easy way to update your existing code: perl -i -pe 's:ARRAY_DEFINE\(([^,]+), *([^)]+)\);:ARRAY($2) $1;:' **/*.[ch]

Added array_append_zero() to write a zero-filled record to an array. Replaced (void)array_append_space() calls with it.

Renamed lib/*-internal.h files to lib/*-private.h for consistency.

ioloop: Added support for per-io/timeout callback log prefix automation.

ioloop-kqueue: Improved error messages. --HG-- branch : HEAD

ioloop: Added callback for handling time jumping forwards/backwards. The default implementation is now to only log a warning when time moves backwards. The callback is also called if it's detected that time jumps forwards. In both cases existing timeouts' run times are updated so that they're called approximately the intended time. --HG-- branch : HEAD

Added io_loop_set_max_fd_count() to specify how many fds we expect to use. It's used currently only for figuring out how much space should be allocated initially to fds. --HG-- branch : HEAD

Don't break if io_remove() or timeout_remove() are called for non-active ioloops. --HG-- branch : HEAD

Added io_remove_closed(). --HG-- branch : HEAD

Ignore io_loop_get_wait_time() return value, we don't need it. --HG-- branch : HEAD

Use priority queue to implement timeout handling. Added timeout_reset(). --HG-- branch : HEAD

These files are MIT licensed nowadays. --HG-- branch : HEAD

If poll()/etc. timeouts, we didn't handle time timeouts immediately but called poll()/etc again with zero timeout and then handled the timeouts. --HG-- branch : HEAD

Various cleanups to ioloop code. --HG-- branch : HEAD

s/array_get_modifyable/array_get_modifiable/ --HG-- branch : HEAD

Removed compiler warnings with NetBSD. --HG-- branch : HEAD

Replaced ARRAY_CREATE() macro with [ipt]_array_init() macros. The macro has no side effects so it might as well be lowercased. --HG-- branch : HEAD

Removed type parameter from ARRAY_CREATE since it's not needed anymore. --HG-- branch : HEAD

Reference IOs before calling timeout handlers, in case they call io_remove(). --HG-- branch : HEAD

OK, so the original kqueue code wasn't actually broken, but it could have been made much simpler. Updated comments in ioloop.h about IO_READ and IO_ERROR usage. --HG-- branch : HEAD

Fixes --HG-- branch : HEAD

Fixes --HG-- branch : HEAD

Compiling fixes --HG-- branch : HEAD

Rewrote much of the kqueue code. It didn't work correctly if there were both input and output I/O handlers added for the same file descriptor. The code works now very much like epoll code. Did also several cleanups. --HG-- branch : HEAD

Fixed "Unrecognized event" errors that occurred with IO_ERROR. Patch by Vaclav Haisman --HG-- branch : HEAD

Added more debugging information for handling unexpected events. Patch by Václav Haisman --HG-- branch : HEAD

kqueue updates. Patch by Vaclav Haisman --HG-- branch : HEAD

Fix IO_ERROR behaviour. Patch by Vaclav Haisman --HG-- branch : HEAD

Added kqueue support. Patch by Vaclav Haisman. --HG-- branch : HEAD