Important BugFix in v0.5.9 - Restoring State Access in Graph Callbacks​

[hi-pender]

Hi everyone,

We want to highlight an important bug fix included in the latest release, [v0.5.9], which resolves a regression introduced in [v0.5.4]. If you've encountered issues accessing state within your root Graph Callbacks recently, this post is for you.

What Went Wrong?

In our effort to bring the Graph Cancel Interrupt feature in v0.5.4, we made an internal change to the execution order of Graph Callbacks. Unintentionally, this change broke a key behavior: the ability to use ProcessState to retrieve the current state within a callback.

After the [v0.5.4] update, if your code tried to access the state inside a callback like this:

err := ProcessState[*State](ctx, func(_ context.Context, state *State) error {
    // Your logic here
    return nil
})
if err != nil {
    // You would consistently get this error:
    // "get state from context fail: have not set state"
}

The call to ProcessState would fail with the error "get state from context fail: have not set state", even though the state was correctly set.

The Fix in [v0.5.9]

This release, v0.5.9, directly addresses this issue. We've corrected the callback execution point, ensuring that ProcessState works exactly as it did prior to v0.5.4.

Key improvement:

• State access is restored: You can now reliably use ProcessState inside your Graph Callbacks once again.

How to Check if You're Affected

You are likely affected by this bug if:

• You are using a version between v0.5.4 and v0.5.8.
• Your application uses GraphCallback and relies on ProcessState within those callbacks to read or modify the graph's state.
• You have seen the "have not set state" error in your logs.

Solution: We strongly recommend upgrading to v0.5.9 immediately.

Our Commitment to Stability

We take stability extremely seriously. To ensure this specific regression cannot happen again, we have added comprehensive unit tests that specifically validate the interaction between Graph Callbacks and state access. This extra layer of testing will help us catch similar issues early in the future.

A Sincere Apology

We sincerely apologize for the disruption this bug may have caused to your projects. Introducing a regression is never our intention, and we are sorry for the inconvenience and frustration it has created. We are continuously improving our testing and release processes to prevent such issues in the future.

Thank you for your patience and continued support. If you have any questions or run into any issues after upgrading, please feel free to comment on this discussion or open a new issue.

Best regards,
The Team