Infinite Loop when Git Permission Denied

[danger-ranger]

https://media.giphy.com/media/fExunDejohBMQ/giphy.gif

[danger-ranger]

Please make sure you have the correct access rights
and the repository exists.
From github.com:pivotal-cf/s3-image-resource
 * branch            master     -> FETCH_HEAD
HEAD is now at 8069372 adds test tag name file
ERROR: Permission to pivotal-cf/s3-image-resource.git denied to concourse-olm-bot.
fatal: Could not read from remote repository.

[RaniSputnik]

We are also seeing this same issue, but we get no error message.

From github.com:Organisation/Repository
 * branch            develop    -> FETCH_HEAD
HEAD is now at 5732f88 Add versions directory (#22)
From github.com:Organisation/Repository
 * branch            develop    -> FETCH_HEAD
HEAD is now at 5732f88 Add versions directory (#22)
From github.com:Organisation/Repository
 * branch            develop    -> FETCH_HEAD
HEAD is now at 5732f88 Add versions directory (#22)
From github.com:Organisation/Repository
 * branch            develop    -> FETCH_HEAD

I think I've found the issue in the git driver.

This code assumes that an error will be returned OR a write will occurr. If neither of those two things happens, the loop is never broken.

semver-resource/driver/git.go

Lines 93 to 107 in 0cb9fc0

	for {
	err = driver.setUpRepo()
	if err != nil {
	return err
	}
	wrote, err := driver.writeVersion(newVersion)
	if err != nil {
	return err
	}
	if wrote {
	break
	}
	}

When calling setupRepo we occassionally return no error and no write

semver-resource/driver/git.go

Lines 339 to 349 in 0cb9fc0

	if strings.Contains(string(pushOutput), falsePushString) {
	return false, nil
	}
	if strings.Contains(string(pushOutput), pushRejectedString) {
	return false, nil
	}
	if strings.Contains(string(pushOutput), pushRemoteRejectedString) {
	return false, nil
	}

I think this is incorrect, we should probably return an error for each of the three cases above.

[RaniSputnik]

This is what our plan looks like;

 - name: assign-version-image
    plan:
      - aggregate:
        - get: source
          resource: git-develop
          trigger: true
          passed: [run-tests]
        - get: image-version-temp
        - get: image-version-dev
          params:
            bump: patch

        attempts: ((attempts))
        timeout: ((timeout))

      - put: image-version-temp
        params:
          file: image-version-dev/number

In our case the failure is occurring in the put: image-version-temp step.

[renbeynolds]

We are seeing the same behavior as well:

Concourse Version: 4.2.1

Config:

- name: version
  type: semver
  source:
    driver: git
    uri: ((versions_repo))
    branch: master
    file: ((version_path))
    private_key: ((private_key))

- name: release-patch
  plan:
  - aggregate:
    - get: snapshot
      passed: [deploy]
    - get: version
      params:
        bump: patch
    - get: pipeline-repo
    - get: release-repo
  - put: release
    params:
      file: updated-snapshot/*
      version_file: version/version
  - put: version
    params: { file: version/version }

Log:

Cloning into '/tmp/semver-git-repo'...
Warning: Permanently added 'gitlab.domain' (ECDSA) to the list of known hosts.
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0
From gitlab.domain/product-versions
 * branch            master     -> FETCH_HEAD
HEAD is now at 2e88ca1 bump to 0.37.0

[vito]

gonna close this in favor of #89 which is admittedly a dupe of this report, but has a PR associated to it