Launching the Lock Poisoning Survey

[KodrAus]

We've been talking about running a survey to see how the standard library's locks are used in the wild. This post announces that survey along with a bit of a description of what poisoning is all about.

We started an FCP for the survey over here: rust-lang/libs-team#5

r? @sfackler @withoutboats

[bjorn3]

You are only talking about poisoning as a way to maintain invariants, while in my opinion poisoning is just as much about easily propagating panics between threads, even when not taking panics into consideration, as .lock().unwrap() is easier to write than any way that doesn't propagate panics. This means that it is much harder to remain in a broken state where one thread has panicked and isn't restarted while all other threads keep going.

[KodrAus]

@bjorn3 The post comes around to that same conclusion, that lock poisoning is more a tool for propagating panics, rather than a tool for directly protecting from broken invariants.

[bjorn3]

Missed that part.

[KodrAus]

It's a nice bit of validation to know you independently have a similar impression on what lock poisoning does and doesn't do 😁

[KodrAus]

I’ve just pushed a small update to the end of the post to be less strongly worded towards removing poisoning, since part of the survey will be telling us how useful it actually is for everyone.

I’ve also updated the publish date so I think we should be good to merge now!

[KodrAus]

@Mark-Simulacrum I think this one's just been waiting for some Core team approval for a while (but that might not have been very clear from the thread here). Would it be easier to publish if I targeted the insiders blog instead of the main one?

[Mark-Simulacrum]

If you want to bump the date to Friday I think it'd be fine to merge this - left a few additional comments too

[KodrAus]

Thanks @Mark-Simulacrum 🙇 I’ve updated the post.