On Fri Nov 28 20:34:17 2025 +0000, Paul Gofman wrote:
I can add similar interface change (adding a possibility to complete async) throughout the other instances of async.c if you think it is better. I can also change windows.speech thing to use the same. Will that help? But honestly I don't want to add the wrong way of completing request, it is a way to not immediately obvious (as sync-related) bugs. If we have the choice between old behaviour and new one we should be prefering newer. It is not really more complicated in this case, rather the opposite, the only complication is that existing wrapper doesn't support it and I believe we'd be better off fixing that instead.
FWIW there are currently only 5 related async.c files in the codebase (the other bunch are thankfully unrelated, like ws2_32/async.c). And one of those in windows.media.speech already needs the same if we go the route of firmly reusing async.c's IAsyncInfo implementation (the test there firmly assumes synchronous completion and an app already depends on that).