I think we have a few issues that need to be fixed upstream (in GHC itself, or cabal, etc.)
Detecting versions of what we're instrumenting
Sending contexts through to threads that we didn't create
Cleaning up thread local state more reliably
I could see us providing some nix utilities in the meantime that update wrapped dependencies to use patched versions of async, expose their Paths_ module, etc. This would give us much deeper integration abilities into the broader ecosystem at the expense of tooling being less useful outside of the nix ecosystem.
This is a bit of a brainstorm:
I think we have a few issues that need to be fixed upstream (in GHC itself, or cabal, etc.)
I could see us providing some nix utilities in the meantime that update wrapped dependencies to use patched versions of
async, expose theirPaths_module, etc. This would give us much deeper integration abilities into the broader ecosystem at the expense of tooling being less useful outside of the nix ecosystem.