Closed MSP-Greg closed 4 years ago
Would it make more sense to call #to_io
indiscriminately?
I haven't retested it, but my notes to the commits for it state that the reason for not using to_io
in the Ruby code is to align with the c code. I said of the code before adding the SSL conditional:
When using SSL sockets, in the c extension code, monitor.io is an SSLSocket, but in Ruby code, monitor.io is a TCPSocket.
Assuming that non Windows MRI is the most prevalent use (using the c code), I thought changing the Ruby code was best. Maybe there's a better solution... IDK
Sorry I missed that, I just woke up so I'm still ingesting coffee to get things started.
I'm fine with this PR. #register
is a reasonably hot path though.
Sorry I missed that
No problem
I just woke up so I'm still ingesting coffee
Have the same problem. I try to not write anything until I've had enough...
register is a reasonably hot path though.
I benchmarked the new conditional using both defined?
& Object.const_defined?
, and both were doing like 10_000_000 per second on a slow notebook, but I didn't have a lot of things loaded. I assumed it won't be an issue...
Currently, errors are raised if OpenSSL is not loaded. Fixed, along with small spec change.
JFYI, I benchmarked
defined?
vsObject.const_defined?
, and, most of the time,defined?
was a little faster. Only tested on Windows mingw, 2.2 & head.For GitHub Actions
runs-on:
, seems to prefer 'macos-latest', current workflow used 'macos'. Changed so CI passes on macOS MRI.Closes https://github.com/socketry/nio4r/issues/238