Hacker News Reader

Comment by crazygringo | Hacker News Reader

100% of crashes on my M1 MBA occur in applications when accessing files over SMB network shares. One every couple of days. I have to restart about once every week or two because it borks the system completely.

But when I'm not connected to an SMB drive, I don't think I've had a single crash in years.

I have no idea if it's something in the SMB driver itself that's crashing, or assumptions inside of application logic that fail in certain SMB scenarios.

View on HN

It’s faulty app logic, assuming an io error cannot occur accessing a filesystem. They’ll also crash if you have disk issues or driver problems. The important thing is they may not even realize they’re calling into a function or routine that can access the fs for something!

All io access needs to be recoverable or catchable for your app not to crash. You’d be surprised how much background fs activity there can be in apps (or indirectly triggered by apps!) all the time.

That and sandboxing effects with perhaps improperly accessed bookmark data

I'm building a macos app now, and only because I wasn't catching fs access error did I realize I needed to handle folder access carefully when using bookmarks in core data.

I swear _Finder_ has this bug if you plug in a superannuated external HDD.

This is one reason where I think Rust shines, and has even improved how I write C++.

So many areas where rust makes me deal with a result or option that are easy to take for granted. It’s a strong reminder of the number of things we should be treating as possible error cases

It makes you do something, but if the consumer just does .unwrap() the end result is going to be the same.

Sure, but a developer can also write any number of other code that’s irresponsible. At that point the language has done all it can for them, and the onus is on the developer to not be lazy.

And when the crash happens, it’s obvious why

Stuff like this with NFS was my living nightmare for years. On Linux it would put processes into the D state and you'd have to restart the entire machine to fix anything.

SMB really is hell. I write business software and our admins do what they can to avoid using it because it's very hard to know when it fails.

But from a dev perspective, that's probably a driver level issue because that's where you want to handle or propagate errors instead of crashing.

Reply on news.ycombinator.com