Not sure this is feasible, as the docker container environment is set when container is started, and subsequent commands ran with `docker exec` can't amend it.

Can't you set environment before checkout ? What's your use-case ?

We are facing the same problem. As I can see the environment present and used during container start is the host environment present during Jenkins node start (may be days or weeks ago) including environment variables configured for this node (before node startup).
There is no way to add additional environment variables (e.g. using EnvInject) as it seems to be only working for build steps and the Docker container start isn't such a dedicated step but "fancy work in the background".

ndeloof: In JENKINS-29677 you mentioned priority of concurrent Build Wrappers. If you find a solution, this may be the trick for this issue as well.

skahlhoefer no, issue here is that you can't change environment for a container after it has been started, so only variable set before the build start can be used.

I start using the CloudBees docker build environment plugin a couple of weeks ago and all was going well until I install the EnvInject plugin. After installing the plugin my shell build step can't find ls (or npm or a number of other things which are in the container)

/tmp/hudson4662764981604235729.sh: line 2: ls: command not found
Build step 'Execute shell' marked build as failure

After uninstalling the EnvInject plugin, all works again.

The problem occurs in the build when the EnvInject plugin options are not being used. I'm glad I don't have to have the EnvInject plugin