runtimetest: Fix check for device nodes

[mrunalp]

Signed-off-by: Mrunal Patel mrunalp@gmail.com

[wking]

Drive by?

[mrunalp]

/dev/console is only setup if we set terminal to true.

[wking]

On Mon, Jul 18, 2016 at 03:41:25PM -0700, Mrunal Patel wrote:

-ocitools generate --output config.json "${TEST_ARGS[@]}" --rootfs '.'
+ocitools generate --tty --output config.json "${TEST_ARGS[@]}" --rootfs '.'

/dev/console is only setup if we set terminal to true.

Can we stop adding “runC isn't compliant out of the box” workarounds
now that we're post 1.0.0-rc1? If there's really a reason to not
setup /dev/console when terminal is false, can we PR the spec with
that information?

[mrunalp]

@wking I have created a runtime-spec PR opencontainers/runtime-spec#518

[wking]

On Mon, Jul 18, 2016 at 03:54:39PM -0700, Mrunal Patel wrote:

@@ -78,7 +78,7 @@ tar -xf rootfs.tar.gz -C ${TESTDIR}
cp runtimetest ${TESTDIR}

pushd $TESTDIR > /dev/null
-ocitools generate --output config.json "${TEST_ARGS[@]}" --rootfs '.'
+ocitools generate --tty --output config.json "${TEST_ARGS[@]}" --rootfs '.'

@wking I have created a runtime-spec PR opencontainers/runtime-spec#518

+1. In the meantime, can we land this PR without this change?

[mrunalp]

I have taken out that change for now pending upstream PR.

[wking]

7fd96c1's ‘& os.ModeDevice’ addition looks good to me, but I'm not sure what the --tty addition is for.

[wking]

6f9de0a looks good to me.

The --tty bit should be easy to add back in if we still need it after
opencontainers/runtime-spec#518 gets sorted.