When building a docker image with a dynamic tag, what's the best way to capture this for use by other processes? For context, we want to have 'content-addressed' docker images (managed by pants, rather than docker's caching) with a tag like

image_tags=["{pants.hash}"]

and stick this into a CloudFormation template to have the image running in AWS as a Lambda or Fargate service or whatever. The only approaches I can determine are not so good: 1. capture it out of the logging of a

./pants package ...

call: our system is then doing automatic processing of human facing messaging 2. looking at the most recent row of `docker images --format '{{.Tag}}' $my_repository`: we have to write the repository out in yet another place, and, with caching/content-addressing, there's no guarantee that the most recently created image is the one that's been created as part of the current build 3. (there doesn't appear to be anything recorded in

dist/

, despite this being a

package

call; and something like

peek

gives me seems to give me

image_tags

before formatting)

My team is also interested in thoughts here, currently we do approach 1.

For 3. Since the "artifact" is a docker image there's nothing to write to

dist

. Although the image ID wouldn't be wrong to write 🤔

Ah, interesting. Would it make sense to write a

dist/path.to.target/target_name.json

file containing info like the repositories, tags and image ID?

Slap some extension before the json and you have my vote:

foo.docker_info.json

I filed https://github.com/pantsbuild/pants/issues/16999. Thanks!

It should be pretty simple all things considered to implement. Let me know if you're interested! I/we are happy to help you along