While investigating fixes for #8682 I discovered that rbByteEncode does not properly handle the case of source and destination encoding being the same.

As described in #8682 (comment):

IOOutputStream calls RubyIO.write and passes only the incoming bytes, offsets, and encoding:

This path was added as an optimization in 236f7ba to avoid constructing string objects just to immediately unwrap them for IO writes.

Unfortunately it hits a bug eventually in EncodingUtils.rbByteEncode where two identical encodings will not no-op, but instead will trigger an encoding error because it rejects encoding from and to the same encoding.

The short-circuit checks here:

...were never updated when the sister logic for RubyString was updated in 621369c. The correct logic, with additional refactoring, looks like this:

Ignoring the problems with IOOutputStream only being able to specify a single encoding for all String writes, we need to fix this issue in rbByteEncode and ensure same-encoding calls no-op the same way as rbStrEncode.