CIRRUS_BASE_SHA is, as we determined earlier this year, worthless.

I need to understand a lot more of this, like, what happened, what changed, why is main missing, what's the deal with DEST_BRANCH, and it's gonna take me a long time because I'm not familiar with this repo. Blocking until I get some time to understand.

What a rabbit hole! TL;DR this appears to have been broken from the very beginning, and was only caught by happenstance because the error was not (and still is not) causing a CI failure.

Furthermore, the reason this works in podman is because of superconvoluted clone_script and other setup that has presumably been fine-tuned and tested over the years. We no longer have local expertise in such arcana.

Proper solution is too complicated to get into here. It'll involve doing a git clone with proper branch tags for the given branch, and testing it under different branches, and under different conditions (CIRRUS_PR vs TAG vs MERGE vs PURPLEBUTTERFLIES or whatever) and different merge bases. I don't see a good ROI.

Good-enough solution: this approach, but please change as follows:

- if -z CIRRUS_CI
+ if -z EPOCH_TEST_COMMIT

Reason: yes, there's an implicit connection between CIRRUS_BASE_SHA and CIRRUS_CI, but it's much cleaner safer better to just actually test the actual actual, not some sort of distant cousin.

Nice find!