tar: prefix hard link destination with extraction target

[cgonyeo]

If the extraction is not running in a chroot, the file to hard link
against must be prefaced by the target for the extraction, otherwise the
file that's being linked to will not exist and the link will fail.

Fixes containers/build#139.

[jonboulle]

Egads. Great catch. LGTM. Can we test, though?

[jonboulle]

Also s/extration/extraction/

[cgonyeo]

Test added, and as a bonus all tests now run against both ExtractTar and ExtractTarInsecure

[jonboulle]

This isn't quite right because you're skipping both test cases (insecure && !insecure)

[cgonyeo]

whoops

[cgonyeo]

No longer unintentionally skipping parts of tests.

[jonboulle]

return err

[jonboulle]

I think splitting these out is preferable, they're quite distinct paths conceptually:

func TestExtractTarFoldersChroot(t *testing.T) {
  if !sys.HasChrootCapability() {
    t.Skipf(...)
  }
  testExtractTarFolders(t, true) 
}

func TestExtractTarFoldersInsecure(t *testing.T) {
  testExtractTarFolders(t, false)
}

func testExtractTarFolders(t *testing.T, chroot bool) 

(The bool is kinda icky, ideally testExtractTarFolders could have a nicer signature with some kind of func() but I'm not sure what it looks like off the top of my head)

[cgonyeo]

Someone should kick semaphore for me :/

[cgonyeo]

Tests split.

[krnowak]

Semaphore decided to remove the branch from testing as seen on https://semaphoreci.com/coreos/rkt/

I guess you'll need to create a new PR. :/

[alban]

It needs to be rebased too.

[alban]

After closing/reopening the PR, we cannot kick Semaphore anymore... For next time, Semaphore can be kicked by pushing the branch again ;)

$ git checkout pr-branch
$ git commit --amend --no-edit
$ git push --force my-repo pr-branch

[alban]

LGTM if it is rebased and if Semaphore is green.