On Fri, Feb 20, 2026 at 09:28:23PM +0100, Marc-André Lureau wrote: > Do you mean that we could drop replay_populated/replay_discarded from > the source? I think we could. replay_by_populated_state() already > handles the aggregation logic by using is_populated() alone. There > isn't much gain in the above implementation narrowing the iteration > from the first source replay.
Yes, then we can also get rid of special casing the first source (where we may invoke other sources' hooks within the 1st source's cb functions), and the skip_source logics. Thanks, -- Peter Xu
