You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
One thought here - why don't we simply chain an entrypoint script into the actual ENTRYPOINT for the container image as a part of the image build for the feature step? We can get the existing entrypoint (via docker inspect) and invoke it afterwards as an arg set. (Using exec "$@"). That also removes the need to reference the feature to get the entrypoint to apply... and it would automatically use whatever the container user was.
The difficulty is with us temporarily setting USER root for the feature install scripts and then having to go back to USER <previous>. When features were built in a separate image (FROM <user image>), we did inspect the user image and pass its user as a build argument. Now that we build the user image and add the features in a single Dockerfile (for caching checksums and multi-arch builds), I haven't yet found a way to determine the user before we change it to root and then go back to it with USER.
E.g., if ENV would support command substitution (which it doesn't):
#70 follow-up
cli/src/spec-node/utils.ts
Line 344 in c67f5c8
Multi-stage builds could result in the current approach returning the wrong user.
Also: Variables are not supported.
Considering the possible cases an alternative approach might work better (if one exists).
The text was updated successfully, but these errors were encountered: