Remove endpoint protocol according to latest docker specific

[ikikko]

Problem

I couldn't docker.image("xxx").pull() with withRegistry from Amazon ECR when I had another job in which I used aws ecr get-login and docker pull with native awscli & docker cli.

Cause

This plugin updated ~/.docker/config.json in a wrong format. We didn't have a problem when we used only this plugin. However, we had a problem when there was another job with native awscli.

The config file ~/.docker/config.jsonis like a following.

{
  "auths":   {
    "1234567890.dkr.ecr.us-east-1.amazonaws.com":     {  --- (A)
      "auth": ... ,
      },
    "https://1234567890.dkr.ecr.us-east-1.amazonaws.com":     { --- (B)
      "auth": ... ,
      }
 ...
  }
}

A problem procedure was a following way.

1. A job which uses aws ecr get-login
   1. Add auth entry (A) to config.json
2. A pipeline which uses docker.withRegistry()
   1. Refresh auth entry (B) in config.json
   2. A auth entry (A) is used when docker pull in the pipeline above
   3. docker pullfails if a auth entry (A) is expired

Though I'm not 100% sure, according to https://github.com/moby/moby/pull/23100/files#diff-0ec52e4c7165f2740e01649dcef5867b , both docker client should uses hostname key (A), and (B) is for backward compatibility. However this plugin updated auth key (B).

[ikikko]

There is no problem so far after I fixed this plugin and installed that to my Jenkins a week ago.

[jglick]

Same comment as #59: maybe right, maybe not, but anyway all this code would better be rewritten.

[jglick]

Likely unnecessary after #67, though could still apply to corner cases such as users of the Google container registry which cannot be handled by a simple docker login.

[jglick]

I would recommend just running something along the lines of

eval `aws ecr get-login --no-include-email`

from a sh step.