We recently updated our gitlab-runner, and found that the new version (9.2.0) failed on build jobs that previously succeeded. After some troubleshooting, we found that the problem was that the new gitlab-runner (or some unidentified change in the way that we use it!) exports a number of environment variables to the job shell. One of these is the generically-named CONFIG_FILE. This conflicts with a Makefile variable used in the Intel IGB driver in the Linux kernel, causing the compile to fail. We ended up working around the problem by adding environment = [ CONFIG_FILE= ] to the runners section of config.toml. The compilation succeeds now.
My questions are:
- Is this an acceptable solution? Is there any potential side effect to what I have done?
- Most other exported environment variables prepend CI_, and in my opinion CONFIG_FILE should as well. That’s a remarkably generic name to be exporting. Is this worth submitting an issue over?