On Thu Sep 4 12:04:00 2025 +0000, Rémi Bernon wrote:
Well, like before I don't really like the idea of putting a lock in the generic implementation, when it's only needed because of one of its user. I think it would be good to have a separate thread safe one wrapping the unsafe one. The serial can still be implemented in the unsafe code, if we really need to return E_CHANGED_STATE.
Makes sense, I made reverted the locking code in `map.c` in the latest revision. The `PropertySet` implementation now guards access to the underlying `IMap/IMapView/IIterable/IIterator` with a SRWLOCK instead (though we can't neatly aggregate `IPropertySet` now). Does this look better?