ecr: switch implementation to use the AWS SDK#126
Conversation
Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>
Signed-off-by: CrazyMax <crazy-max@users.noreply.github.com>
Flydiverny
left a comment
Flydiverny
left a comment
There was a problem hiding this comment.
Nice!
I tested docker/login-action@42ba9a1dc8ea3ab85b1c32bdcb2bff18ba03c94f (before last force-push) on our self-hosted (linux) runner without providing any input credentials and it worked as expected for that use case 👍
I was thinking maybe we could use the v3 aws-sdk lib instead of the v2 one, ie @aws-sdk/client-ecr and @aws-sdk/client-ecr-public. I tried to make these changes locally but for some reason docker buildx bake validation fails saying there's diffs to output. Seems to be CRLF vs LF issue 😅
It seemed to reduce the dist size a bit since v3 sdk is modular.
|
#128 if you're curious |
|
FYI from ~1 hour when this PR is merged our pipelines started failing to login to AWS ECR, it could be related to the changes here.
Using exact version v1.10.0 is working. |
|
@aramalipoor Can you add a comment with your workflow in #129 please? |
3 participants
Follow-up #114 (review)
Fixes #125
@Flydiverny Now using the AWS SDK so we don't need to rely on the AWS CLI.
With this we are also able to authenticate on Windows runners.