State of Dochandle should be "ready" when change event is triggered

[paulsonnentag]

When you subscribe to changes on a doc handle the first time the document loads the state on the handle is still "loading" in the change callback even though the document is already loaded.

handle.on("change", () => {
   console.log(handle.state) // logs "loading" when initially loading instead of ready
})

Why is this happening? On each state transition we call #checkForChanges this function which has two responsibilities:

• checks if we need to emit a change event
• transitions the state machine from "loading" to "ready" when the document was loaded for the first time

Right now when the document is loaded for the first time #checkForChanges emits a change event and then transitions the state machine from "loading" to "ready". That's why in the callback the state field on the handle is still loading. This pull request fixes this that we first transition into the ready state and then trigger a change event.

[msakrejda]

Thanks for the bug report and the fix! I have some minor comments, but overall it looks great.

[msakrejda]

Thanks for adding the comment: this definitely makes it easier to follow.

[msakrejda]

I'm pretty sure the CI failure is unrelated: https://github.com/automerge/automerge-repo/actions/runs/16046435722/job/45280996045?pr=453#step:6:797 -- it's an OOM. I'll try restarting the test and see if it's a recurring issue.

[msakrejda]

@paulsonnentag are you interested in continuing this patch? It looks like the CI OOM is persistent, so we'd need to figure that out, but other than that and the minor cleanup I suggested, I think this is ready to merge, and it seems like a worthwhile change. If not, I can look into getting it over the finish line.

[paulsonnentag]

Thanks for catching that. If you'd take over this PR I'd be very happy

[msakrejda]

Hmm, so, I think this does break some existing assumptions somewhere. I was able to reproduce the CI failure, partly: the tests appear to hang. I'm guessing maybe we're leaking memory somehow during the hang and eventually we OOM. I do see a slow leak locally. I can take a look at it next week.