Code Review: sockets-1

tfpt review /shelveset:sockets-1;REDMOND\jflam

I’ve cleaned up Peter Bacon D.'s socket patch from last week now
that we have our new MutableString implementation in. I’m running a test
only job through SNAP right now.

There are a few nits that I have with some new conversion protocols,
some missing overloads etc. in this shelveset, but I’d rather see this
go through sooner rather than later.

  1. Many methods and constructors don’t have the strongly-typed
    overloads. Here’s one in TCPSocket.cs:

[RubyConstructor]
public static TCPSocket/!/ CreateTCPSocket(CodeContext/!/ context,
object remoteHost, object remotePort) {

This is not critical since it will pass the tests correctly, but we
really should resolve these things at some point. Not a high priority
now (this is vastly preferable to having the strongly typed overload
only, and missing the object overload) since we’re in a bit of a time
crunch.

  1. Not sure what the status of the implementation is - but we can’t
    create instances of Socket today since there isn’t a ctor defined for
    it.

  2. A large # of PortNames are statically created at start-up (in
    BasicSocket.cs). Would be nice to change “tcp”/“udp” to an enum since
    there are only 2 values.

  3. ConvertToPortNum: consider a faster search than linear search for
    PortName. Also, the conversion protocol at the start (port is int) looks
    a bit strange?

  4. // Ruby uses Little Endian whereas .NET uses Big Endian IP values
    byte[] bytes = new byte[4];
    for (int i = 3; i >= 0; --i) {
    bytes[i] = (byte)(iHostname & 0xff);
    iHostname >>= 8;
    }

Is this code platform-independent for MRI as well?

Thanks,
-John

This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.

| Privacy Policy | Terms of Service | Remote Ruby Jobs