@orzeh Adding all files to the build context could be problematic if there are huge files. I really don't know how docker CLI handles this, but seems the builder inspects metadata of the base image (http://docs.docker.com/engine/reference/builder/#onbuild):

At the end of the build, a list of all triggers is stored in the image manifest, under the key OnBuild. They can be inspected with the docker inspect command.

Later the image may be used as a base for a new build, using the FROM instruction. As part of processing the FROM instruction, the downstream builder looks for ONBUILD triggers, and executes them in the same order they were registered. If any of the triggers fail, the FROM instruction is aborted which in turn causes the build to fail. If all triggers succeed, the FROM instruction completes and the build continues as usual.

@orzeh After thinking about again probably you're right. So uploading all files and ignore only those that matches patterns in .dockerignore. Seems our original implementation that adds only files mentioned in the Dockerfile was some kind of premature optimization then.

@marcuslinke I agree with you.

If one has to deal with huge files, he can use .dockerignore or - to take full control over what is send to docker daemon - provide own InputStream via BuildImageCmd.withTarInputStream().

Wouldn't better place Dockerfile in separate directory?

@KostyaSha which Dockerfile do you mean?

Any :) from my experience i always placing Dockerfile in clean directory to not have accidental uploads to build context.

In general PR LGTM 👍 even if i not fully understand ONBUILD feature and all possible cases :)

Ok, now I get it.

I use ONBUILD when I have several Spring Boot microservices and I add them to the image as a exploded JAR to leverage Docker caching (see this blog post). Then I create base image with number of ONBUILD ADD ... instructions and just put FROM my-base-image in microservices Dockerfiles.

@orzeh Thanks for contributing!