Tagging subscribers to this area: @dotnet/area-microsoft-win32
See info in area-owners.md if you want to be subscribed.

Thinking, perhaps it would look something like: an environment variable to disable the finalizer thread (does that exist?); then manually running a set of unit tests (presumably we have unit tests for invalid reg keys e.g.), forcing GC, and inspecting a dump to see objects pending finalization, perhaps focusing on certain types. Or simply events/tracing of finalizer activity .

Just curious, do we make any attempt in tests of some sort to look for unintended finalizations?

We've done things in the past like this:

I'll play around with what that would look like in the base type. That of course would only handle safe handles, only with a debug/checked runtime, and only if enabled.

Curious how many hits we'd get if we ran our unit tests now with that pasted into SafeRegistryHandle, e.g.. I'm guessing leaky tests would hide such bugs. Would it be worth creating an issue in case a community member is interested in such an experiment? If it was not too painful they could then try with SafeFileHandle E.g.

I'm already working on it. It's a lot.