Join GitHub today
GitHub is home to over 36 million developers working together to host and review code, manage projects, and build software together.Sign up
Switch IO reactor to nio4r #1728
Sorry, I should have looked at it more. Passed on Appveyor.
nio4r is aware of the issue, but they thought switching to libuv is (understandably) something they're not interested in changing at this time. I'm not a c type, so...
From https://github.com/socketry/nio4r/wiki/Selectors#selecting-io-objects-for-readiness , if it's a nice-to-have.
Overall lit seems reasonable. Looks like mostly a re-mapping of the behavior that was there before. I'll have to update all the docs on this class to not be based on
Is there a test that fails if you remove this code or is it an intermittent failure? Basically, I want to make sure that we know where to start looking if we decide we want to debug this in the future.
The other comment about switching to use a block instead of allocating an array would be a nice perf bump, but we can also leave that as a TODO.
Looks like there's still a failing test on 2.5.3 for now:
And a few failing against jruby and windows.
Mar 20, 2019
As the maintainer of nio4r, I'd be happy to evaluate other options. Or even build something entirely new on top of
The actual problem space is pretty large though. For example,
If your only issue is blocking network IO, then