Getting from tested to battle-tested
18 points by winter
18 points by winter
Obligatory disclaimer: I work for Antithesis (and am happy to answer any questions), but independently, I think this post has some really great points outside of “we used Antithesis.”
Thanks for sharing. I've had a somewhat difficult time explaining the potential of DST with colleagues and I found this post to be a really solid concrete example.
Because we were leveraging existing code, we didn’t think to write a new test, especially for this situation.
This is such a good point.
What I’m most interested to see is how much of the existing test suite ends up being replaced with Antithesis. I think there’s still room for some of the hand-written tests, especially for well known and high risk scenarios.
But I do wonder which categories presented here will become obsolete over time. They call this out, and say Antithesis won’t replace their suite. But, who knows how it will play out.
I think the goal is greater coverage of test cases, not to replace manual or hand written tests entirely. I don't think the folks at Antithesis would say unit tests are longer needed, or even certain kinds of integration testing.
It's another tool in the toolbelt. Albeit, an extremely powerful one.
Totally agree. But I think it still influences the composition of which and how many tests to write, even if it doesn’t replace them.
Like, unit tests will always be useful because you have such fine grained control. But will you write as many if you’re confident that many of the edge cases are caught by Antithesis?
I don’t think you’d write exactly the same tests, and just add it on top if you were starting from scratch and it’s further along in maturity.
One change could be that some of the flaky tests become useful: Antithesis can force and record their failures, if they are genuine hopefully-fixable races.
I think an integration test with demanding validation conditions is exactly what you want to write under Antithesis tooling, isn't it?