158
GM stops selling the Chevy Blazer EV to deal with ‘software quality issues’
(www.theverge.com)
This is a most excellent place for technology news and articles.
I mean, software issues exist where there's software and testing on the inside gives you whatever time you pay for, but once your software is out in the world you get more testing done in a day than would have realistically been done over months of paid testing...
I understand this but still think it’s bullshit. A car isn’t like some shitty little app. They cost tens of thousands of dollars, weigh tons, move at insane speeds inches away from one another, and are expected to last decades with daily use and minimal maintenance. I’m a mechanical engineer in product development. Pushing broken shit to production will destroy a company. Just because software updates are easier and cheaper than mechanical recalls shouldn’t excuse releasing broken shit out into the world.
Ok, but you won't know your product is broken in certain ways unless it's tested in every possible ways imaginable and that's impossible to do. As a mechanical engineer you should know that, the end user will always find a way to incorrectly use the product you developed.
But we do though. Maybe not exactly test every possible scenario. Typically when we make a design decision we plan for the worst theoretical condition the part will be exposed to. Then we plan for 5-10 times that. Think about the cost and effort added to everything with that level scrutiny. We design for fringe cases. That’s the point I’m trying to make. It’s insane to me that because it’s software, companies get a free pass on that level of scrutiny. As software takes over more car functions that becomes more concerning. It’s bullshit that I’m part of their beta test.
You'll probably never test every possible way that a user might interact with a given piece of software, but for embedded control software, like what's responsible for managing charging and starting the car, it literally is quite feasible to test every possible input and use case.
And you expect the public to catch edge cases, not mission critical failures.
Exact, they test for the worst cases they can imagine, but once it's in the hands of hundreds of thousands of people or millions of them these people will do things the programmer never could have imagined...
Like what start the car? Turn on the system. Did you read the article? The public isn’t catching edge cases, they’re catching gross system failures.