Fix Makefile bare export leaking all variables to subprocesses

[tuxerrante]

Summary

• Remove the bare export directive from the Makefile that was exporting every Make variable into the environment of every subprocess, causing potential slowness and unexpected behavior
• Consolidate the redundant .env file (containing only PYTHONPATH) into the existing env file and env.example

Why some contributors experienced multi-minute make invocations

The bare export exported all Make variables — including ones defined with recursive expansion (=, ?=) containing $(shell) calls (git describe, git rev-parse, go env, etc.) — to every subprocess. Contributors with heavier shell startup (zsh + oh-my-zsh/compinit, nvm, conda) paid that cost on each $(shell) invocation. On fast setups with minimal bash the overhead was imperceptible, masking the issue.

The tunnel target also used $(shell az ...) which ran az at Makefile parse time on every make invocation, not just when make tunnel was called. This is now a recipe-level $$(az ...) call.

Test plan

• make commands expect you to manually source the env file as per the docs

[mrWinston]

Couple of things:

1. I believe sourcing the env file automatically in the Makefile is conceptually flawed. It just increases complexity of our make-targets. Targets can't assume anymore that the caller has already sourced the required env file. Also, using different env files will be more complex.
2. It looks like this PR reorders stuff in the env file without reason. This makes reviewing harder and actually introduced a bug. Please make sure that the changes are actually minimal.
3. This RP seems to add additional configuration to the default env file. I would like to see this in a separate PR to address these changes separately. They don't seem related to the issue ourlined in the PR title and make reviewing harder
4. Adding the env file to git can cause issues overwriting existing local configuration. Also, having a file both ignored and added is pretty confusing and surprising. And both are bad for our dev tooling.

[tuxerrante]

A few clarifications:

• .env file was not part of .gitignore, which seems a pre-existent bug
• env /env.example file needs some ordering and linting in my opinion, something should fail when values are not the ones expected (eg: Rename CLUSTER env var to CLUSTER_NAME for consistency #4656 in my local tests the cluster creation was failing because in the util pkg we expect a variable that doesn't exist in the dev.example)

Tried with shellcheck but doesn't catch it. It sees that LOCATION is assigned somewhere in the file and considers it valid. It doesn't do order-sensitive analysis because the variable could also come from the environment.

[github-actions]

Please rebase pull request.

[Copilot]

Pull request overview

This PR aims to reduce make overhead and unexpected behavior by stopping GNU Make from exporting all Make variables into every subprocess environment, and by consolidating the Python PYTHONPATH setup into the standard env workflow.

Changes:

• Removed the bare export directive from the Makefile and replaced it with an explicit PYTHONPATH export only.
• Added PYTHONPATH to env.example so contributors get the Python path when creating/sourcing their local env file.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
Makefile	Removes bare export and explicitly exports only PYTHONPATH with a default value.
env.example	Adds PYTHONPATH so local env files created from the example include the needed Python module path.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.