Peerplays_Blockchain/peerplays-fc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

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.

Browse source
This commit is contained in:
dnotestein 2014-09-08 17:23:03 -04:00
parent 881903c1be
commit c79cbdae8e
1 changed files with 16 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
src/network/ntp.cpp
View file

@ -30,11 +30,13 @@ namespace fc
fc::future<void> _request_time_task_done;
bool _shutting_down_ntp;
ntp_impl() :
_ntp_thread("ntp"),
_request_interval_sec( 60*60 /* 1 hr */),
_last_ntp_delta_microseconds(0)
_last_ntp_delta_microseconds(0),
_shutting_down_ntp(false)
{
_last_ntp_delta_initialized = false;
_ntp_hosts.push_back( std::make_pair( "pool.ntp.org",123 ) );
@ -143,11 +145,16 @@ namespace fc
}
catch (...)
{
//don't restart read loop if we're destructing
if (_shutting_down_ntp)
return;
//swallow any other exception and restart loop
elog("unexpected exception in read_loop, going to restart it.");
}
_sock.close();
fc::usleep(fc::seconds(_request_interval_sec));
fc::usleep(fc::seconds(1));
if (_shutting_down_ntp)
return;
} //outer while loop
} //end read_loop()
};
@ -165,6 +172,7 @@ namespace fc
ntp::~ntp()
{
my->_ntp_thread.async([=](){
my->_shutting_down_ntp = true;
try
{
my->_request_time_task_done.cancel_and_wait("ntp object is destructing");
@ -177,7 +185,11 @@ namespace fc
{
wlog( "Exception thrown while shutting down NTP's request_time_task, ignoring" );
}
//instead of canceling task, we close the socket and wait for the task to end because receive_from will throw
//if we just canceled the task, the receive_from would likely complete after the recv_buf was gone and we would get a random write into our stack.
my->_sock.close();
my->_read_loop_done.wait(); //wait for socket to close
#if 0
try
{
my->_read_loop_done.cancel_and_wait("ntp object is destructing");
@ -190,7 +202,7 @@ namespace fc
{
wlog( "Exception thrown while shutting down NTP's read_loop, ignoring" );
}
#endif
}, "ntp_shutdown_task").wait();
}