Hi, if invalidation occurs, it is recorded in

invalidateStack

, and if this stack is not empty, it is popped from this stack and returned from

endRestartGroup()

. Recomposition is performed through the returned

RecomposeScope

. Right? However, if there is no invalidation during the composition process, null is returned, and if invalidation is requested later, how does

endRestartGroup()

detect this and recompose again? If look at the implementation of

endRestartGroup()

, there is no code to track invalidations. Only request pop once.

val scope = if (invalidateStack.isNotEmpty()) invalidateStack.pop() else null // RecomposeScopeImpl?

EDIT: I found some clues. Suspend until invalidation is added through

awaitWorkAvailable()

in

Recomposer.runRecomposeAndApplyChanges()

. If there is a waiting task(

hasSchedulingWork

), it immediately resumes

Continuation

and proceeds to recompose. Otherwise, it maintains the suspend state. The maintained suspend(

CancellableCoroutine

) is resumed when the operation is completed through the

effectJob

, but as I traced the places where the

effectJob

is used, there are only

rememberCoroutineScope()

and

LaunchedEffect

. Where will this suspend resume? I think it’s time to record invalidation, but it wasn’t. They are recorded regardless of

Job

.

well, it seems like your puzzlement might be because you're looking primarily at the invalidation code in

Composer.kt

, but it's only responsible for a portion of the invalidation logic. invalidate calls flow up to the

Recomposer.kt

, which is the parent of the root CompositionContext.

Thank you so much 🥰 I will investigate again with your comments.