Listener callback is invoked multiple times when message is published and when calling subscribe multiple times

[puopg]

Edit: Changed the title to be more representative of the bug at hand.

Seeing some interesting behavior differences after upgrading to version 4.34.0.

Services currently enabled on my key:

• Storage and playback
• Stream controller
• Access Manager

Context here is a simple chat application. On message publish, a client receives that message via subscribing to the channel. However, a special situation here is we are subscribed to a channel group.

• Upon first publish, the client correctly hears a single event, and the handler is called once.
  • Also push a signal to the subscribed channel group. Observe still the handler is called once. (We have attached a channel for this signal onto the channel group).
• Publish a 2nd message now, and also publish another signal to the subscribed channel group.
• Observe that the handler is called multiple times.
  • I've seen a varying amount of invocations, sometimes its 4 calls to the handler, sometimes 3, sometimes its just 1.

Affected version: 4.34.0
Unaffected version: 4.33.1

Actual outcome:
Multiple invocations to the handler.

Expected outcome:
A single invocation to the handler.

[samuellawerentz]

I'm having the similar issue where the handler is invoked multiple times.
v4.33.0 is working fine.
After updating to v4.34.1 its firing multiple times.

For one message, the handler is firing 4 to 5 times depending on the number of times the pubnub.subscribe() method was called.

[puopg]

I have attached 2 videos, one on 4.33.1 the other using 4.34.0.

Code sandbox also attached. This is a simple counter to visually see how many times the handler is invoked. Notice I only call addListener once, and the only change is calling subscribe.

Codesandbox: https://codesandbox.io/s/dazzling-cannon-y71vi

4.33.1

Screen.Recording.2021-11-30.at.10.14.07.AM.mov

4.34.0

Screen.Recording.2021-11-30.at.10.14.45.AM.mov

[puopg]

@raipubnub Have you guys noticed this?

[ghost]

Thanks for the detailed report @puopg. We are looking at this and will provide an update once we finish the analysis.

[ghost]

This was a result of a change in superagent that affected chaining the end method. We have a fix ready to go out, should be released soon.

[github-actions]

@puopg this issue is addressed in v4.34.2

[puopg]

Cool thanks!