On Tue May 24 16:57:51 2022 +0000, Rémi Bernon wrote:
Regarding the refcount issue it looks like calling `IWMReader_Close` should be enough to release the callback. As there's no guarantee that it does we could also clear the filter pointer when it is destroyed and leave the callback to be released eventually later. Using the sync reader we would just re-implement the async reader in qasf reader and add another read thread implementation which feels pretty redundant.
FWIW I can see that there're tests to validate that `IWMReader_Release` doesn't leave any outstanding ref and that `IWMReader_Close` also releases the provided callback.