Switching between work environments

Table of contents

1. The .env file is what selects the environment
2. bash: …/venv/bin/flask: bad interpreter: No such file or directory
   1. Solution 1: recreate the venv folder in local environment
   2. Solution 2: use a name for the local environment different from the conventional one
3. error: Cannot update time stamp of directory ‘rosemary.egg-info’
4. Stale bytecode after switching interpreters

The .env file is what selects the environment

Local, Docker and Vagrant share the same working tree, and each of them expects a different .env. The variables that actually differ are WORKING_DIR and MARIADB_HOSTNAME:

Environment	Example file	WORKING_DIR	MARIADB_HOSTNAME
Local	.env.local.example	""	localhost
Docker	.env.docker.example	/workspace/	db
Vagrant	.env.vagrant.example	/vagrant/	localhost

Every Rosemary command builds its paths from WORKING_DIR, so a stale value is what produces errors such as FileNotFoundError: [Errno 2] No such file or directory: '/workspace/app/features'. When you move to a different environment, copy the matching example file first:

cp .env.local.example .env
source .env

Beware of custom variables

Overwriting .env discards the variables contributed by your features. Run rosemary compose:env afterwards to merge every .env file found under app/features/ back into the root .env.

bash: …/venv/bin/flask: bad interpreter: No such file or directory

A virtual environment records the absolute path of the interpreter that created it. If that path does not exist in the environment you are currently in, every script inside venv/bin/ fails this way.

Docker and Vagrant are set up so that they no longer collide with your local environment:

• Docker installs the dependencies into the image itself. docker/images/Dockerfile.dev does not create a virtual environment at all, so nothing of it lands in the bind mount.
• Vagrant creates its environment as vagrant_venv, not venv, precisely so that it does not overwrite yours through the synced folder.

So a broken venv today normally means it was created by a different interpreter on your own machine, or it is a leftover from an older checkout.

Solution 1: recreate the venv folder in local environment

rm -r venv
python3.13 -m venv venv
source venv/bin/activate
cp .env.local.example .env
pip install --upgrade pip
pip install -r requirements.txt
pip install -e ./rosemary

Solution 2: use a name for the local environment different from the conventional one

If you want to be extra safe, give your local environment its own name:

python3.13 -m venv myenv
source myenv/bin/activate

venv/, .venv/ and vagrant_venv/ are all in .gitignore, so remember to add your own name to it if you pick a different one.

error: Cannot update time stamp of directory ‘rosemary.egg-info’

The development container reinstalls Rosemary from inside the bind mount, running as root:

pip install -e ./rosemary

That writes rosemary/src/rosemary.egg-info/ into your working tree with root ownership. Back on the host, pip cannot update it. Remove it and reinstall from the environment you are actually using:

sudo rm -r rosemary/src/rosemary.egg-info/
pip install -e ./rosemary

Stale bytecode after switching interpreters

__pycache__ directories and .pyc files are also written into the bind mount. They are keyed by interpreter version, so they are usually harmless, but if you see import errors that do not match the source you are reading, clear them:

rosemary clear:cache

This removes the build directory and every __pycache__ directory and .pyc file under the project root.