I suppose I should read errata sheets, but I expect them to contain obscure edge cases, not information to the effect that a basic function of the chip won't work when configured in the documented way, and that the work around is to change the configuration on an unrelated pin.
What caught me was item 9 in
The effect is that the UART RX (on RB2) won't work unless one changes RA2 to be non-analogue input (it defaults to analogue).
Haven't these people heard of testing?
Looks like a revision without this fault has been out for at least a year, so how come RS Components sold me an earlier revision?
Sylvia.