Missed update in diamond-derived stores

[igorkamyshev]

What is the current behavior:

$store.update missed in some narrow case. We can see only one watcher's trigger.

Please provide the steps to reproduce and if possible a minimal demo of the problem via https://share.effector.dev, https://codesandbox.io or similar

https://share.effector.dev/2IjuEI7T

What is the expected behavior:

Two watchers' triggers — initial and first update.

Which versions of effector packages, and which browser and OS are affected by this issue? Did this work in previous versions of effector?:

22.1.1 and later.

Notes

This example will be working correctly if we add scope.getState-call or remove combine call.

[AlexandrHoroshih]

I did some research and found out that mapped store $search in the repro is computed twice, while being computed once before 22.1.1

https://share.effector.dev/LhGu9Dlw

Looks like the "flow" of the bug is something like this

1. Computation starts and $search computed first time

2. $search is initialized in scope

3. Then next combined store is computed and during that $search is computed second time for some reason, even though it is already calculated in this run and its dependency was not changed second time 🤔

4. This part i tested locally with effector repo: - internal update filter $search called only one time after all these computations and compares next value with the current - but since store was computed twice in this run, it compares the same values. This can be proved by changing $search mapper function to return a new reference each update (an object for e.g.) - https://share.effector.dev/qZbb1B18 - this way bug is not appearing

Also calling scope.getState($search) somehow fixes $search computations and it is called only one time - probably because it is initializes value in scope before any computations? 🤔
https://share.effector.dev/N4AeCpYX

[zerobias]

Thanks for the report! The fix will be published in next minor release