-
Notifications
You must be signed in to change notification settings - Fork 22
Description
Hello, awesome library! Was combing through it's internals and noticed you currently use thread-local/atomic pointer addresses for seeding the RandomState/SeedableRandomState types. What are your thoughts on providing a getrandom feature to toggle delegating that to the OS-provided RNG?
Most pressing issue I can see this raising is how to deal with the error handling, but it looks like the tracking issue for something like this being introduced in std is trending towards just panicking on failure. This is also the approach go settled on a while back.
It may be worth it to investigate how the API could be altered to allow for a seeding process that might fail, since that should make it easier to move to whatever failing RNG API core will eventually provide (assuming that's something you plan to do).