Sometimes when starting Tachyon on a Linux machine it will terminate immediately, before all the initialization is complete, with the message "I/O Possible". This is coming from the default action of SIGIO on Linux which is to terminate the process. If I catch these signals they happen continuously.
It doesn't happen every time I start Tachyon, but seems to happen consistently if I start Tachyon from bash as sh from inside a fresh screen window.
This seems to have been caused by a non-sensical fcntl(0, F_SETFL, 1) call which set nonsensical file values. I'm not really sure why that code was there in the first place. Perhaps it was meant to be F_SETFD? In any case I don't need to set close-on-exec here and the code as is is broken. Removing it fixes this bug.
Sometimes when starting Tachyon on a Linux machine it will terminate immediately, before all the initialization is complete, with the message "I/O Possible". This is coming from the default action of SIGIO on Linux which is to terminate the process. If I catch these signals they happen continuously. It doesn't happen every time I start Tachyon, but seems to happen consistently if I start Tachyon from bash as sh from inside a fresh screen window.
Date: 2013-11-30 22:08:01
User: travisb
Possible related email thread from a long time ago: http://lkml.indiana.edu/hypermail/linux/net/0210.0/0022.html
Date: 2013-11-30 23:13:40
User: travisb
This happens in xterm in /bin/sh on Linux 3.2.30 on ARM as well as via ssh on Linux 2.6.24 on PPC
Date: 2013-12-01 02:28:56
User: travisb
This seems to have been caused by a non-sensical fcntl(0, F_SETFL, 1) call which set nonsensical file values. I'm not really sure why that code was there in the first place. Perhaps it was meant to be F_SETFD? In any case I don't need to set close-on-exec here and the code as is is broken. Removing it fixes this bug.
Tracked by Nitpick