Shell provisioner output all messed up #11047
Comments
|
As a workaround we need to wrap/buffer the command execution with something (which works most of the time, but not always) like: # wrap commands in a way that their output is correctly (most of the time) displayed on the vagrant up output.
# see https://github.com/hashicorp/vagrant/issues/11047
function docker {
set +x; echo "$(/usr/bin/docker "$@")"; set -x
}
function kubectl {
set +x; echo "$(/usr/bin/kubectl "$@")"; set -x
} |
|
More workarounds:
Though as mentioned in the help:
In that case, grep's |
|
this is still happening in latest vagrant (2.2.16) and it makes the whole vagrant experience quite bad as the output is all messed up (even with things like vault). there's a python script after these example that will by default flush stdout after each character. call this from a provisioner step or with vagrant-execute as this is how the output looks: rgl@rgl-desktop:~/Projects/vault-vagrant$ vagrant execute -c 'python3 /vagrant/stdout-partial-lines.py --flush'
==> default: #
==> default:
==> default: l
==> default: i
==> default: n
==> default: e
==> default:
==> default: 0
==> default: 0
==> default: 0
==> default: 1And this is how it should have looked: rgl@rgl-desktop:~/Projects/vault-vagrant$ vagrant execute -c 'python3 /vagrant/stdout-partial-lines.py --no-flush'
==> default: # line 0001 ############################################################################################################
==> default:
==> default:
==> default:
==> default: Here's the python script to reproduce the problem: #!/bin/python3
import sys
import argparse
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
description='write partial lines to stdout to demonstrate vagrant broken output handling')
parser.add_argument(
'--flush',
default=True,
action='store_true',
help='flush after each written character')
parser.add_argument(
'--no-flush',
dest='flush',
default=False,
action='store_false',
help='do not flush after each written character')
args = parser.parse_args()
for n in range(1, 26):
prefix = '# line %04d #' % n
line = '%s%s\n\n\n\n\n' % (prefix, '#' * (120 - len(prefix)))
for c in line:
sys.stdout.write(c)
if args.flush:
sys.stdout.flush() |
partial lines are buffered until they are complete (have a line ending) see hashicorp#11047
… the ui partial lines are buffered until they are complete (have a line ending) see hashicorp#11047
|
Please have a look at #12403! :-) |
… the ui partial lines are buffered until they are complete (have a line ending) see hashicorp#11047
… the ui partial lines are buffered until they are complete (have a line ending) see hashicorp#11047
|
I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues. |
Vagrant version
Vagrant 2.2.5
Host operating system
Ubuntu 18.04.3 LTS
Guest operating system
Ubuntu 18.04.3 LTS
Vagrantfile
You just need to add a normal
config.vm.provisionstep, e.g.:The full Vagranfile is at https://github.com/rgl/kubernetes-ubuntu-vagrant/blob/master/Vagrantfile.
Expected behavior
Actual behavior
The output is all messed up, it seems that vagrant is not correctly interpreting the new lines or just dumps the whole output buffer when its full without regarding whether its tail has a new line or not, and thus, forces a new line. At least it seems that what is happening by looking at the garbled/messed output (the output should be two perfect columns of text like displayed in the Expected behavior section):
This behavior practically makes the vagrant output useless and really hard to interpret.
Steps to reproduce
This can be easily reproducible with the execute plugin:
vagrant plugin install vagrant-executevagrant execute -c 'kubectl get clusterrole --all-namespaces' km1Or by launching the whole environment with
vagrant up.The text was updated successfully, but these errors were encountered: