> Here's a more specific approach following feedbacks from deraadt@.  I'm
> also looking into something that can be committed because there's no way
> to test this as syzkaller has no reproducer (yet) for this bug.
> 
> Comments?  Oks?

Not directly related to these changes, but we might want to change the
various ddb routines which take a printf-like function, and make them
take an snprintf-like function, a buffer and its size.

This would allow you to use a local buffer to assert_not_managed() or
whatever routines you inline it into.

But that might up ending in more churn than reasonably acceptable.