README.md

Dockerfile for running the evaluation

The Dockerfile in this directory contains all the software-requirements necessary to run the evaluation.

For conveniently running this, simply run make in this directory, which will automatically build the Docker image, and drop you into a bash shell inside the container, connecting in the base repository into /trex-usenix25. Exiting out of the container shell and running make again will automatically drop you back into the same container, so any files or extra files set up inside are persisted across make invocations. To completely close the container (which will erase such data), use make kill. This does not remove the underlying built image however, to do so, use make destroy-image.

Important notes:

• We use podman rather than Docker in the Makefile. This means that root inside the container directly corresponds to the regular user inside, and no permission fixes are necessary if moving files in and out.
• The trex repository must be cloned within the /trex-usenix25 directory inside the container. You can do this by running just trex from within the container.
• For evaluating against ML-based systems (specifically, for the last subsection of the evaluation: Section 5.3 of the paper), due to GPU requirements, we recommend setting up ReSym on a remote system with a powerful GPU (as per instructions here); specifically, make sure that within the container the environment variables ENABLE_RESYM, REMOTE_SERVER, and REMOTE_RESYM_BASE_DIR are set as described in the instructions, and all the instructions have been correctly followed on the powerful-GPU server. If your local system has a powerful GPU, then you can simply use the local address as the REMOTE_SERVER environment variable too.