We could probably use upx for the Gitlab CI, but I still think it would be nice to find a solution that doesn't require shipping the entire source.
Okay, I don't have any (other) good idea tbh, I'll probably leave it for later if it's an issue and hopefully using the test source + line + message would be enough for Gitlab to not mess the results up.
I've considered doing some hashing in the test code, but the only context we can get in `winetest_ok` (and others similar) calls is the test expression, possibly with a rolling buffer of previously executed tests. That doesn't feel like a very useful or unique information about the test being executed. Capturing the test full expressions would also increase the exe size.
Also considered other things like function-relative (or at least relative to the first test of a function) line numbers, which doesn't feel very robust to code addition/removal.
Overall nothing felt as satisfying as being able to hash the surrounding source lines (actually the preceding lines are the most meaningful, but including a few lines after can be useful to capture ok calls on multiple lines).