Draft: RFC initial cut of error redesign
This is a start at #247 (closed).
So far this:
- Provides the new
TorError
type - Provides a handful of conversions into
TorError
- Abolishes one existing combined error type,
GuardMgrError
, in favour ofTorError
If you like this, my plan is to:
- Make every crate's error type convertible to
TorError
- Abolish the nested error enums in the higher-level crates and just use
TorError
Precisely where the specific enums should be retained is going to be case-by-case specific. Sometimes the specific enums will need an Other(TorError)
. As we go on we are probably going to want to have quite careful case-specific conversations about the precise set of Kind
s and so on.
There are some XXXs still here. Personally I don't propose to merge this until I have fixed them, but they are easily findable and #247 (closed) is still open so there would not be any serious problem with doing so if you think that's better.