If we update the maximum value for "infinite" in the future, it is possible
that we forget to update this comments and thus leads misunderstanding.
Signed-off-by: Hu Keping <hukeping@huawei.com>
The documentation already says the cache miss happens only at `ARG`
variable usage, not declaration, but there is a very common implicit
usage: `RUN`, which this commit documents even more, improving on #21790.
Also, use `definition` instead of `declaration`: it's the same thing, and
`definition` is already used in this documentation, contrary to
`declaration`.
Also, distinguish between "instructions" and "variables defined by `ARG`
instructions".
Signed-off-by: Thomas Riccardi <riccardi@systran.fr>
ignores buildtags, for instance, and somehow didn't pick up some code in
the signer.
This calls golint on every go file in the repo and also fixes some linting
issues, which involves renaming two yubikey functions to avoid stuttering.
Signed-off-by: Ying Li <ying.li@docker.com>
Implements a `CachedPath` function on the volume plugin adapter that we
call from the volume list function instead of `Path.
If a driver does not implement `CachedPath` it will just call `Path`.
Also makes sure we store the path on Mount and remove the path on
Unmount.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
3 out of 4 KeyStore implementations have already been returning this
type or a pointer to it; document this as a requirement, and modify
implementations to comply.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
- Remove Sign, no longer needed for CryptoService implementations
- Remove unused testKeyID1
- Drop the intermediate MockCryptoService with no users; rename
StrictMockCryptoService to MockCryptoService
- Actually be strict in the new MockCryptoService, and fix key ID
use
- Clearly document what FailingCryptoService and MockCryptoService
are supposed to do
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
keystoremanager.certsToRemove unexpectedly returns an empty map of old
certificates when there are no new ones. While there is nothing really
great to do in that case (we don't want rotation to make a repo
completely unusable), this already can not happen so just fail instead
of having this weird case.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
allValidCerts implies validity, and certsFromCN is neutral, when the
roles are at least reversed. So, talk about neutral certsFromRoot,
and trustedCerts.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Updates the readme to mention running the tests using golem.
Also provides instructions for making test development easier.
Signed-off-by: Derek McGowan <derek@mcgstyle.net> (github: dmcgowan)
