←back to thread

650 points clcaev | 1 comments | | HN request time: 0.203s | source
Show context
Noumenon72 ◴[] No.45063996[source]
They should be saving every crash as a unit test to ensure it never happens again.
replies(2): >>45065502 #>>45067030 #
1. inetknght ◴[] No.45065502[source]
More than a unit test -- a whole system test. But, as a software engineer with experience in robotics and drones with a focus on software safety, yes I 100% agree.

The unfortunate thing is that the state of the industry (or, my experience in it) currently is not set up to be able to do that cheaply nor at scale. Imagine you have tens of thousands of various unique problem scenarios to run through, and some might take several minutes of simulation to run the test. Even if your release cadence is slow, but especially if you have continuous deployment with dozens of micro-releases every day: how exactly do you cheaply scale such that simulation testing doesn't become a massive bottleneck?