Missing timeout in SendSync

[webfrank]

Bug description
PID.SendSync has a fixed timeout of 20 seconds.

How to reproduce it?
Send a message to a mailbox which will not reply.

Expected behavior
This should be a configurable parameter, with a 20 secs default value.

Screenshots

Library Version:

• Go-Akt version: 2.7.1
• Go version: 1.23

Additional context
I would like to know how to send a ctx.Response() which is an error. Now I'm using this code:

resp, err := d.getDevices()
if err != nil {
	ctx.Err(err)
} else {
	ctx.Response(resp)
}

but the receiver will timeout if ctx.Err(err) is used, I expect to have an error instead to signal an issue.

[Tochemey]

@webfrank I will take a look at the issue and get back to you. Also can you point me to the hard-coded 20 seconds path?

[Tochemey]

The reply timeout is set when creating the actor system. It is a timeout that is applied to any ask-pattern message in the system. It can be set https://github.com/Tochemey/goakt/blob/v2.7.1/actors/option.go#L68. The default value is 20 seconds as you stated.

[Tochemey]

@webfrank kindly take a look at the example in the cluster code to see how the ask-pattern works: https://github.com/Tochemey/goakt-examples/blob/main/actor-cluster/static/actors/actors.go. I will be glad to provide further explanation.

[webfrank]

The reply timeout is set when creating the actor system. It is a timeout that is applied to any ask-pattern message in the system. It can be set https://github.com/Tochemey/goakt/blob/v2.7.1/actors/option.go#L68. The default value is 20 seconds as you stated.

Hi, I think it could be useful to set it on a per message basis, having more control on it.

[webfrank]

@webfrank kindly take a look at the example in the cluster code to see how the ask-pattern works: https://github.com/Tochemey/goakt-examples/blob/main/actor-cluster/static/actors/actors.go. I will be glad to provide further explanation.

Hi, I've already implemented the examples. What I was trying to achieve is to return back an error instead of waiting timeout.

If a receiving actor wants to reply with an error the only thing could do is calling ctx.Err() but this does not cancel the context and does not returns the error. This way the caller should wait for the timeout.

[Tochemey]

@webfrank fair point. I can look into that.

[Tochemey]

@webfrank are you free for a quick call?

[webfrank]

@webfrank are you free for a quick call?

Hi, sorry, dinner time. I'm CET timezone. I'll send a DM with my contacts.

[Tochemey]

@webfrank I have refactored the code a bit in this PR to address the timeout issue. You may have to use the latest commit hash when I merge it because I am not planning to cut a release tag soon until the christmas period because I am planning some changes in the library.

[Tochemey]

• ctx.Err(err) is recommended instead of calling panic. This will trigger the supervisor to handle the faulty actor.
• ctx.Response is used to send the response to a request. This is used when using the ask-pattern of messaging

[Tochemey]

@webfrank kindly use this commit hash 9760b03 for the meantime. I will cut a release tag a bit later.

[Tochemey]

@webfrank kindly use the latest release 2.8.0