V2.4.0 Not using k8s config

[dkoshkin]

Ran ./helm init

➜  out git:(helm) ✗ kubectl get nodes
NAME        STATUS                     AGE       VERSION
master001   Ready,SchedulingDisabled   17m       v1.6.2
worker001   Ready                      17m       v1.6.2
worker002   Ready                      17m       v1.6.2
➜  out git:(helm) ✗ export KUBECONFIG=~/.kube/config
➜  out git:(helm) ✗ stat ~/.kube/config
16777220 22173086 -rw-r--r-- 1 dkoshkin (1535981906 0 6228 "May  2 20:42:16 2017" "May  2 20:36:45 2017" "May  2 20:36:45 2017" "May  2 16:12:41 2017" 4096 16 0 /Users/dkoshkin/.kube/config
➜  out git:(helm) ✗ ./helm version
Client: &version.Version{SemVer:"v2.4.0", GitCommit:"1ec2c04be8d4d69c15b85df97394bfa2f5919f03", GitTreeState:"clean"}
^C
➜  out git:(helm) ✗ ./helm init
$HELM_HOME has been configured at /Users/dkoshkin/.helm.
Error: error installing: Post http://localhost:8080/apis/extensions/v1beta1/namespaces/kube-system/deployments: dial tcp [::1]:8080: getsockopt: connection refused
➜  out git:(helm) ✗

Expected to run successfully, the cluster is not running locally but looks like Helm is not picking up the kube config file. The same cluster setup works fine with v2.3.1

[rodcloutier]

After a git bisect the following commit 4d8f23e introduced the break

[yuvipanda]

A strace tells me it does open and read $HOME/.kube/config, but still tries to hit localhost.

[yuvipanda]

This isn't just helm init - all helm commands fail with similar reason (even when there's a tiller running already)

[rodcloutier]

Changing the github.com/imdario/mergo dependency to the previous version fixes the problem.

darccio/mergo@6633656 is fine
darccio/mergo@3e95a51 is breaks

[rodcloutier]

Kubernetes is also using mergo Maybe a clash in versions?

[rodcloutier]

I suggest to pin the version of mergo to a known working version for both sprig and Kubernetes. Would this be fine?