Eric Frias
aa6882b3b7
Allow us to safely cancel tasks that are executing asynchronous network reads
...
and writes. This was previously unsafe because we almost always passed read/write
buffers to boost that were on the stack. Canceling the task deleted the stack and
therefore the buffer, but couldn't reliably prevent boost from writing to the buffer
if data came in after the cancel. This commit adds variants of the read and write
functions that take a shared_ptr<char> instead of a raw char* as the buffer, and
these variants will ensure the shared_ptr will outlive the boost::asio read/write.
2014-09-09 11:10:37 -04:00
Eric Frias
db8eb2e5d4
Revert "Instead of canceling the read_loop, close the socket and wait for the read_loop to finish as a result. This prevents the receive_from from writing into the buffer (which is on the stack!) after the task is canceled. This caused nasty intermittent exception stack unwind errors."
...
This reverts commit c79cbdae8e
2014-09-09 11:08:27 -04:00
dnotestein
c79cbdae8e
Instead of canceling the read_loop, close the socket and wait for the read_loop to finish as a result. This prevents the receive_from from writing into the buffer (which is on the stack!) after the task is canceled. This caused nasty intermittent exception stack unwind errors.
2014-09-08 17:23:03 -04:00
dnotestein
10fdbcf5b3
Dequeue a thread from a blocking promise's notify list when the all the thread's tasks that are waiting on that promise are canceled. This prevents a crash when the promise is fulfilled after the thread is destroyed.
...
Re-organize read-loop execution to avoid crashes in read_loop on Win32 when ntp object destructs. Call quit on ntp_thread when ntp object destructs to free up thread (eventually we need to make fc::threads call quit in their destructor, but more work is required to make that work properly).
~fc::udp_socket now closes socket on destruction (consider doing this for tcp sockets as well).
2014-09-08 15:31:13 -04:00
Eric Frias
ac385d1f6b
Allow the user to supply a reason string when canceling a task (useful for debugging)
2014-08-27 14:07:44 -04:00
Eric Frias
d188b138d6
Allow fc::canceled_exception to pass through places where we were catching and ignoring all fc::exceptions
2014-08-25 18:43:12 -04:00
Eric Frias
92d29ac657
Prevent crash at shutdown rescheduling NTP task
2014-08-12 15:19:44 -04:00
dnotestein
7cc69f3bb7
Set canceled flag in task's context object so that we cancel out of active tasks when they try to resume. Wrap a try-catch block around tcp and udp resolve requests to convert any non-fc exceptions to fc exceptions. Minor change to clarify design intent that request_time_task should always be running in _ntp_thread.
2014-08-01 11:31:36 -04:00
dnotestein
9e320a3db8
Add descriptions for a bunch of async tasks, remove logging during thread::quit to avoid crashes for now until we cleanup thread quit code
2014-07-27 00:09:15 -04:00
Eric Frias
b460fd6b41
Cleanup all async tasks on exit, thread safety, work in progress.
2014-07-26 18:22:38 -04:00
Eric Frias
c208a968b2
Fix ntp on win32
2014-07-25 14:40:47 -04:00
Daniel Larimer
456c81df27
fix potential errors in NTP caused by running in main thread
2014-06-25 23:32:09 -04:00
Eric Frias
331e6aac7d
Convert networking exceptions to fc::exceptions to properly catch them in NTP code (fixes at least some of the shutdown crashes)
2014-06-20 12:22:37 -04:00
Daniel Larimer
bb3c3ec118
fix NTP flooding issue
2014-06-16 14:04:36 -04:00
Eric Frias
6898484321
Remove a few annoying unreferenced variables
2014-06-16 12:03:20 -04:00
Daniel Larimer
b7ad720647
updating ntp service
2014-06-16 11:17:29 -04:00
Eric Frias
a1604e3708
Fix NTP on Win32
2014-06-12 08:40:24 -04:00
Eric Frias
8ef1c4e0a7
Add include for ntohl() on linux
2014-06-11 13:56:03 -07:00
Eric Frias
59d3e80fb7
Add include to fix win32 build
2014-06-11 16:32:47 -04:00
Daniel Larimer
75ec020227
fix ntp
2014-06-11 15:30:30 -04:00
Daniel Larimer
0c64d208d9
adding ntp support to fc
2014-06-11 15:17:28 -04:00
