AVR Freaks

I tested on a 3217 board. Although I did't spend a lot of time trying to figure it out, the errata seems to apply although worded a little odd. It makes it sound like the pin override (for the output value, which is all it controls) is not released. Doesn't sound so bad since are setting the dir to input anyway (whatever it is overriding on the output should not make a difference). But it seems there is more to it, or I just am not understanding, since I can disable/enable the tx and can then no longer tx. If a write to the baud reg is done after turning off the tx, then it works ok again (one fix in the errata).

Probably the simple thing to do in your case (and what I initially did below), is just leave the usart on. There is no need to disable tx, so just switch the pin to an input and back to an output when wake. There also may be an advantage to enabling the tx output before enabling the rs422 so the tx line isn;t floating when rs422 turns on.

...

int main(){
    Cpuint::vectorFunction( board.sw.portVector(),
        []{ board.sw.inputMode( INTDISABLE ); }
    );
    u0.setBaud1x(115200);
    u0.txOn(); //also sets tx to output
    for(;;){
        u0.write('X');
        wait( 1_sec );
        Port<USART0_TXD>::input(); //(not a feature in Usart, so do through Port)
        board.sw.inputMode( BOTHEDGES ); //also clears flag
    u0.txOff(); //if I do this (but no need to do)
    u0.setBaud1x(115200); //then any write to usart register 'fixes' (else no more tx)
        Slpctrl::sleep( Slpctrl::PDOWN );
        Port<PINS::USART0_TXD>::output(); //tx back to normal
    u0.txOn();
    }

}

It is a little confusing, but I think I understand it better now, although not fully.

What is is a little unusual is the override that seems to be in play is the DIR override, which I think may be only used in open drain mode (the errata does not specify which override they are talking about). Also unusual is it seems the override signal makes it way to a register which we can see (DIR- you can read and see what is going on to some degree). Normally these override signals are not visible since the override is after the register.

We are required to set the DIR to output for most tx uses, so the DIR override is not used. But it seems when turning off TXEN the tx pin DIR will be set to input in any case (when TXEN 1->0, then tx pin DIR=0). If the tx pin needs to be an output for any use after tx disabled, then you have to first release the override the usart is still applying (DIR=0), by writing to a usart register. If the RXEN is 1, then this is not required, but the tx pin DIR still goes to 0.

So, a fix for turning off tx would be something like-

void txOff(){

    TXEN = 0;

    TXEN = 0;

    //but note that tx pin is now an input

    //any time txen is disabled

}

and then also deal with the pin direction again next time enabling tx.

This may still not be totally correct, but its close.