Don't use -o pipefail for plugin #907
Conversation
|
@michaelshobbs thoughts on this? I feel as though if a plugin fails, then the deploy should also fail, but I'm not strong on that opinion. |
|
Seeing as core is fairly religious about using |
|
@vincentfretin do you mind updating the example and making it apparent as to what pipefail does? |
|
@michaelshobbs @josegonzalez I updated the PR. |
docs/plugins.md
Outdated
There was a problem hiding this comment.
Should still have a space :)
37780ca to
b81a6c3
Compare
|
Oups, fixed |
|
Yeah @vincentfretin you would need to do that for every command that could fail. I'm not saying its the right choice for your plugin. Just an option. |
Don't use -o pipefail for plugin
I don't think it's a good idea to document the use of -o pipefail by default.
Plugin author blindly follow this instruction and we have issue with the plugin after that.
A simple command like:
elasticsearch:create)
# Check if Elasticsearch container is installed
IMAGE=$(docker images | grep "jezdez/elasticsearch" | awk '{print $3}')
if [[ -z $IMAGE ]]; then
echo "Elasticsearch image not found... Did you run 'dokku plugins-install' ?"
exit 1
fi
will completely stop dokku right before the if because grep returned exit code 1. (jezdez/elasticsearch image not build)