lib: Fix usec comparison in timeout_update_next When tv_usec is exactly 1000000, call to kevent() will fail because tv_sec does not get incremented. Found by Adrian Gonzalez <adrianglz@globalpc.net>

lib: ioloop: Add functions for adding/moving timeouts and ios to a specific ioloop.

Updated copyright notices to include the year 2018.

lib: io_loop_context_deactivate() - Assert that given context is currently active

lib: io_loop_context_new() - deactivate old context Since it changes the current context, it needs to deactivate the old one.

lib: io_loop_destroy() - Assert-crash if ioloop contet is still set It may be too late to try to deactivate it at that point, so it points to a bug.

global: start relying on [io]_stream_unref(NULL) being a no-op Cleanup performed with the following semantic patch: @@ expression E; @@ - if (E != NULL) { - i_stream_unref(&E); - } + i_stream_unref(&E); @@ expression E; @@ - if (E != NULL) { - o_stream_unref(&E); - } + o_stream_unref(&E);

lib: timeout_remove(NULL) should be a no-op

lib: io_remove(NULL) and io_remove_closed(NULL) should be no-ops

lib: Free io_switch_callbacks_free only after other atexit-callbacks Fixes a crash at deinit where io_loop_remove_switch_callback() was called in lib_atexit() callback.

lib: When logging I/O or timeout leak, log also raw backtrace This can be useful when trying to figure out where the io_loop_destroy() was called from.

lib: Add asserts to make sure running ioloop is always current_ioloop We could also switch current_ioloop to the running ioloop temporarily while calling callbacks, but this behavior is probably clearer. All of the existing code should already work this way. Add the asserts after IO or timeout callback is called, so if the assert triggers, we can find out which callback caused the change. The initial assert in io_loop_handler_run() is enough to verify that the initial ioloop is correct.

lib: If CORE_IO_LEAK environment is set, panic on IO/timeout leaks

lib: io_loop_set_current() - do nothing if ioloop doesn't change There's no need to call the switch callbacks if the ioloop doesn't change.

Updated copyright notices to include the year 2017.

global: Replaced all instances of memset(p, 0, sizeof(*p)) with the new i_zero() macro. Used the following script: C_FILES=`git ls-files *.c` H_FILES=`git ls-files *.h` for F in "$C_FILES $H_FILES"; do echo "$F" perl -p -i -e 's/safe_memset\(&\(?([^,]*)\)?,\s*0,\s*sizeof\(\g1\)\)/i_zero_safe(&$1)/g' $F perl -p -i -e 's/safe_memset\(([^,]*),\s*0,\s*sizeof\(\*\g1\)\)/i_zero_safe($1)/g' $F perl -p -i -e 's/memset\(&\(?([^,]*)\)?,\s*0,\s*sizeof\(\g1\)\)/i_zero(&$1)/g' $F perl -p -i -e 's/memset\(([^,]*),\s*0,\s*sizeof\(\*\g1\)\)/i_zero($1)/g' $F done

lib: Added io_wait_timer This allows easier tracking of how much time the io_wait_timer has been spending on (multiple) ioloops.

lib: Code cleanup - move code to ioloop_add_wait_time()

lib: Fix ioloop wait times when io_loop_time_refresh() is used

lib: Added ioloop_global_wait_usecs This is helpful when multiple ioloops aren't used and io_loop_get_wait_usecs() can't be easily used for all of them.

lib: Ensure tv_r gets set in io_loop_get_wait_time If this does not get reset here to 0 it can cause errors or unpredictable behaviour in select/kqueue polling, such as very long delays.

io: Add source filename This will make debugging alot easier when you can see what the filename is as well.

lib: Update ioloop attributes With pending IO, the ioloop attributes still need updating, instead of just returning.

lib: Prevent crash in ioloop If only non-fd sources and no timeouts are in ioloop, prevent crash and unnecessary waiting when there are pending IOs.

lib: Changed t_pop() API to make it a bit more like free()

lib: Created data_stack_frame_t type for data_stack_frame

lib: Fixed memory leak io_add_istream()+io_remove() when the stream didn't have fd.

global: freshen copyright git ls-files | xargs perl -p -i -e 's/(\d+)-201[0-5]/$1-2016/g;s/ (201[0-5]) Dovecot/ $1-2016 Dovecot/'

lib: Added io_loop_find_fd_conditions()

lib: Added io_loop_get_wait_usecs()

lib: Try fixing ec6e672a6e32 (ioloop timeout fixing) Previous code was broken at least when moving a timeout between ioloops. This is now tested and works.

Reverted ec6e672a6e32 for now due to some bugs.

ioloop: Delay actual start of a new normal timeout until the next io_loop_run() cycle. This makes sure that timeouts will not expire before they get a chance to run.

lib: Don't call ioloop context deactivate() if activate() hasn't been called yet. This could have happened immediately after the callback was registered.

global: freshen copyright Robomatically: git ls-files | xargs perl -p -i -e 's/(\d+)-201[0-4]/$1-2015/g;s/ (201[0-4]) Dovecot/ $1-2015 Dovecot/' Happy 2015 everyone! Signed-off-by: Phil Carmody <phil@dovecot.fi>

lib: io_remove() should finish closing the io before unreferencing its istream. This is because the istream unreferencing may close the fd.

lib: Fixed io_loop_move_timeout() to retain the next_run time, so that the timeout is not implicitly reset. This problem became with timeout_add_absolute(), since resetting an absolute timeout causes it to fire immediately (msecs == 0).

lib: ioloop: Fixed absolute timeout removal. Absolute timeouts fire only once, so the timeout should be removed from the priority queue once it fires.

lib: Fixed earlier timeout_add_absolute() commit. I removed too much from Stephan's patch.

lib: Added support for setting timeouts at absolute time, rather than relative to current time.

lib: Code cleanup - check io_loop_context_add/remove_callbacks() callback parameters.

io_add_istream(): Allow the istream to be added even if it doesn't have fd.

lib: add identifying markers to data-stack frames Add a string parameter to t_push() so that in DEBUG mode, misbehaviour inside a stack level can be blamed on someone. Default the T_BEGIN macro to automatigally use __FUNCTION__ or __FILE__:__LINE__ as that identifier, therefore no clients of those macros need to change. ioloop used t_push() directly as it wanted customised diagnostic strings. To preserve this friendliness, also introduce a t_push_named() which takes a format string with paramters. Apart from the unused paramter, a non-DEBUG build should see no changes. Signed-off-by: Phil Carmody <phil@dovecot.fi>

ioloop: Fixed segfault crash in io_loop_call_pending(). When a callback called io_set_pending(), the io_loop_call_pending() loop crossed the end of the ioloop->io_files list. Based on patch by Stephan Bosch

Added io_add_istream() and related functionality for combining the ioloop/istream. The idea here is that it's possible to implement asynchronous istreams when there isn't a file descriptor that can be used to reliably receive updates.

ioloop: Added io_set_pending()

Updated copyright notices to include year 2014.

Added io_loop_add/remove_switch_callback() to call a callback when ioloop is changed.

Use io_loop_set_current() instead of directly setting it.

io_loop_run() now assert-crashes if it's attempted to be used recursively for the same ioloop.

Compiling fix for Sun compilers. I wish gcc/clang warned about these as well, as sometimes they indicate bugs.

Added io_loop_have_immediate_timeouts()

timeout_add_short(1) may have caused infinite looping.

Added io_loop_have_ios()

Oops :) Update copyrights to 2013 without breaking all .c files.

Updated copyright notices to include year 2013.