exec: "bad file descriptor" occurs in tests

[cyphar]

This appears to happen on random tests (sometimes in rootless sometimes in normal mode). I started noticing it a week or so ago. The error is basically just EBADF. I'll see if I can reproduce it, it seems to only happen on runc execs.

not ok 33 mask paths [directory]
# (in test file tests/integration/mask.bats, line 46)
#   `[[ "${output}" == "" ]]' failed
# runc list (status=0):
# ID          PID         STATUS      BUNDLE      CREATED     OWNER
# runc spec (status=0):
# 
# runc run -d --console-socket /tmp/console.sock test_busybox (status=0):
# 
# runc exec test_busybox ls /testdir (status=0):
# time="2017-08-24T07:59:50Z" level=error msg="bad file descriptor" 
# runc list (status=0):
# ID             PID         STATUS      BUNDLE             CREATED                          OWNER
# test_busybox   20467       running     /tmp/busyboxtest   2017-08-24T07:59:50.815627018Z   root
# runc kill test_busybox KILL (status=0):
# 
# runc delete test_busybox (status=0):
# 

[cyphar]

Aha! I think I figured it out, this looks like it was caused by #1560 (I still need to double check this is the case). In particular, the added handling of tty.resize() could conceivably race with the stdout being closed by the terminal -- which would result in runc outputting an error message where it doesn't make sense.

In retrospect the tty.resize part of that change doesn't make sense.