Integration tests don’t really help if you just push the wring build to production.
This is like designing a deadbolt that tells you that the key doesn’t work, but it allows you to open the door anyway. Why would anyone have a process in place where you can push to production with failing integration tests?
I actually work in automotive testing, and the honest truth is that there likely is no real automated pipeline.
Automotive software testing is much more complex than simple software unit or integration tests. You need to run on actual hardware, accompanied by all the other ECUs you are interfacing with. And the tools that slow you to do so are specialized tools, which often are not yet integrated into CI/CD processes (they’re pretty much all working on it though). I.e. getting test results for a build involves manual labor, which makes it prone to errors.
This is like designing a deadbolt that tells you that the key doesn’t work, but it allows you to open the door anyway. Why would anyone have a process in place where you can push to production with failing integration tests?
I actually work in automotive testing, and the honest truth is that there likely is no real automated pipeline.
Automotive software testing is much more complex than simple software unit or integration tests. You need to run on actual hardware, accompanied by all the other ECUs you are interfacing with. And the tools that slow you to do so are specialized tools, which often are not yet integrated into CI/CD processes (they’re pretty much all working on it though). I.e. getting test results for a build involves manual labor, which makes it prone to errors.
You don’t think this is a smell? Surely your company can improve on this.
I do, and we are actively working on it.