rtems-980923 is available

[joel at oarcorp.com (joel at oarcorp.com)]

On Wed, 23 Sep 1998, Eric Norum wrote:

> I think the best solution here is to have  
> rtems_termios_dequeue_characters return 0 if the last characters have  
> been dequeued and return 1 if they have not.
> 
> That's the way the return value is being used now.
> 
> In fact, I am sure I sent a patch with this change several weeks ago  
> -- it must have gotten lost in the shuffle.

There have been so many changes to the return value from this function
that I have lost track of which is in now or not. :)

I think the libchip tree has yet another variant on this.  The dequeue
function returns nothing and it checks the state afterwards.  This was not
that clean since the state was a private value.

> The change is so small that I'm including it in this message.
> =============================================================
> diff -u -r rtems-980923.orig/c/src/lib/libc/termios.c  
> rtems-980923/c/src/lib/libc/termios.c
> --- rtems-980923.orig/c/src/lib/libc/termios.c  Wed Sep 23 07:20:29 1998
> +++ rtems-980923/c/src/lib/libc/termios.c       Wed Sep 23 13:13:42 1998
> @@ -928,8 +928,5 @@
>                 tty->rawOutBufState = rob_busy;
>         }
>         tty->rawOutBufTail = newTail;
> -
> -       return nToSend;
> +       return 1;
>  }
> -
> -
> ========================================================
> 
> BTW -- Sorry about this, Joel.  I realize that keeping all these  
> submissions straight is tough enough without the rest of us getting  
> into disagreements like this.

:)  This one has flip-flopped a lot.  I think I personally argued about
this a while back.  It started with two things:

+ making this routine safe to call when nothing was enqueued, and
+ getting a return value/indicator back.

I personally like the "is there something there?"

But in all fairness, returning the count is an indicator of that as well. 
You have to treat it as zero or non-zero then which is technically how a
lot of C routines are defined anyway.  The problem would come when you use
it as a loop counter. 

<stepping up on the podium>
RTEMS routines should have precise semantics and the return values should
have accurate meanings.  If 0 or 1 is all that is reliable, then that is
what it should be.  A count is more useful but if it is not accurate, then
someone will eventually lose.
<stepping down>

:)

--joel