rgbds
rgbds copied to clipboard
gbdev
Improve tests and testing infrastructure
The current way of doing regression tests is quite crude, it has to be improved. Also, there are not enough tests, there are many things of the toolchain that aren't tested and could potentially break.
I think that ideally the CI scripts should run all tests individually (maybe each test can be in a folder) and it could keep track of the number of passing/failed tests, for example.
I don't know, the infrastructure is still kind of crude—the amount of failed tests is not reported, and their output tends to be drowned in the reports of passing tests.
I used to think that bash wasn't the right language here, but now I think it actually makes sense because of how much the output is compared to reference data in files. At most, you could replace it by some python scripts, I guess.
That much I agree with, though for rsgbds I think I will instead make a test runner in Rust that walks through all of the test cases.