Const & ?

const ?

It's better to add braces to the while.

Using pure pointers isn't safe, are you sure that it can't be replaced by std::unique_ptr for example?

Wrong order.

Do you know what's going on here?

@Itms it's also the question for you.

I suppose Enable != Start. Enable doesn't mean that it starts something.

The brace should be on the new line.

The brace should be on the new line.

ScenarioConfig is related only to RL. Maybe use namespace instead?

No need to have all enums in the one line.

GameMessageType is also related only to RL.

Are objects of this class supposed to be copied?

This Step in which space? Simulation, real time or something else?

Why const ScenarioConfig* and not const ScenarioConfig& here? Especially in case you formally "move" it to the method.

Why const char* and not std::string or CStr?

Should it be public?

Should it be public?

What does this method do?

I'd like to add some empty lines to separate members with a similar context.

Is it lock for all methods?

I am setting m_GameMessage to reference msg (which is on the stack). This doesn't result in any issues since m_GameMessage only needs to be available until m_MsgApplied.wait(msgLock) unblocks which certainly is within the lifetime of the previous stack frame (either Step or Reset).

I will probably update this to use an rvalue ref (and use std::move) to make ownership more obvious... Thoughts?

This was following the convention used in CProfiler2 but I don't have a preference myself either way. Let me know if you would like it to be changed to something like StartHTTP

This is stepping the simulation (stepping the game engine is exposed via the HTTP endpoint).

Nope. Thanks for the catch!

Nope. Fixed this locally and will update soon.

This applies any existing message (ie, instance of GameMessage) which has been received by the RL interface to the current game (or resets the game).

Yeah, at least it is for all methods used by the HTTP endpoint (ie, Step, Reset, GetTemplates).

What's about the visual case?

You don't need to add RL for each method, you can just wrap the whole code in namespace RL.

Yes, and that should be avoided. We shouldn't store pointers/references to local variables as class members until we explicitly guarantee the lifetime.

Additional variables isn't needed here, it doesn't give more information about what's going on here and the lines are pretty short.

Braces on new line.

extern void EndGame(); shouldn't be in the header. It should be in the implementation on in the GameSetup header.

In the visual case, the game messages are applied in the Frame function similar to when the game simulation step is called when running w/o --rl-interface (should be line 397)

Yeah, that would've been a good idea to make it cleaner. Updated now.

I moved this to source/ps/GameSetup/GameSetup.h

Add newline between the two

It should be solved.

What's happening if the val is incorrect integer?

Incorrect type of std::unique_ptr<char>, it should be std::unique_ptr<char[]>.

Braces on new line.

It's unclear where does it get the message from.

My bad. I forgot to make it an rvalue ref here, too.

Hmm... I was assuming that it was a properly formed HTTP request. Would you like me to be more pessimistic about malformed HTTP requests?

So far, I haven't worried too much about it since "Content-Length" is often handled by http request libraries and the RL interface requires both a flag to be set and then only accepts requests from localhost (unless you change the address to 0.0.0.0). I also figured that there wasn't much motivation for malicious agents to attack the game via the RL interface as it is likely being used for development of an AI rather than a hosting the game engine as a server of sorts. (In that context, I had actually imagined there would still be an orchestrator of sorts collecting and applying the commands as all actions at a given step are part of a single call to Step - and it likely wouldn't make sense for both players to control the stepping of the game arbitrarily.)

Regardless, I can certainly update it if preferred. Let me know what you think!

r-value doesn't make things easier here.

The issue is still here. You're storying reference to a local variable.