[Haxe 3.3.0] Support of --wait stdio

[SlavaRa]

HaxeFoundation/haxe#5188

[Neverbirth]

This should be configurable, because keeping the old method of working is useful for getting compile information (EDIT: you open a haxe server with verbose compile information, connect to it in FD, and wait for each compile to get the info) and in some minor cases to reuse the same haxe instance.

[SlavaRa]

Okey.
Enable stdtio with false by default or Disable stdio with false by default or another variant?

[Neverbirth]

I'd go with the first one because it's the classical behavior, but I don't care that much really. It would be more for @elsassph and @Meychi to decide maybe.

[elsassph]

We're still using Haxe 3.1.3 here; backward compatibility is important. You should be able to get the compiler version I think.

[elsassph]

And it should probably be possible to disable it in general (who knows what problems people can run into - we also need to disable the compiler server)

[elsassph]

That's very cool though!

[SlavaRa]

@elsassph Enable stdtio with false by default or Disable stdio with false by default or another variant?

[Neverbirth]

@elsassph Unless I miss something backwards compatibility is already kept. And since this is using the compiler server, it can be disabled in the same way as the old server.

[elsassph]

I'd enable by default (bool prop to disable it so default value is false) if you properly detect that it is supported by the compiler.

[SlavaRa]

I added the setting

[Neverbirth]

Will review today (hope not to forget!), but if someone else reviews and merge before than I do I'm OK with it.

[SlavaRa]

@Neverbirth thanks

[Neverbirth]

Not that my English is perfect, but I'd say:

Makes Haxe completion server to use stdin and stderr instead of sockets for communication with the IDE (only applies to CompletionServer Completion Mode and Haxe 3.3+).

[SlavaRa]

ok, I will fix it today

[Neverbirth]

Is this PR really complete? From what I read of this new mode it looks like it isn't. CompletionServerCompletionHandler still behaves the same way, so it's using TCPClient to send queries instead of stdin, and haxeProcess_ErrorDataReceived should be changed to read the compiler results. Isn't it like this?

[Neverbirth]

So what's the plan on this?

[SlavaRa]

I rediscovered PR today or tomorrow