SMSC serial0

[Matt Evans]

Vincent Sanders wrote:

> serial 1 could have been nobbled?

Can't parse - you mean physically screwed-up?  (yes, could have been, and
double-checked this last night to make sure. ;)

*Thanks Vincent for the flake, and moves on to....*

Andrew Flegg wrote:

> > I had a fiddle with this last month and it was a pain in the
> > arse.  Peculiarly, *TYPE devicefs:$.serial1 returned /nothing/ when I sent
> > data down serial port 1, but returned gibberish (as expected) from the IR
> > remote via serial2.
> 
> Did you solder the port directly onto the SerialIO lines, might a MAX232 
> be needed?

*grin*  Yes yes yes.  I'm utterly barbaric with my hardware, but not
/that/ barbaric ;)  (mind you, my Sun's RS232 port is so weak, it probably
wouldn't kill it.. urgh.)

> > (The port physically works, in non-RISC OS land.)
> 
> What other checks did you do?

We've discussed this separately but for the benefit of the list:

My own code running from ROM talks down serial 0  [note serial 0 = Serial1
in RISCOS, serial 1 = Serial2 ;]  to a workstation with no problems.

[pretty rom hack at http://www.cs.man.ac.uk/~evansm7/pics/BushSTBrom.jpg -
never got round to buying them there posh 16bit flash chips ;]

I verified this last night; set up the box with my gateway PC, did 'cat
/dev/cuaa0' and could see the banner of my serial monitor.  Then I
replaced the NCOS ROMs.  *serialinfo says Serial1 is 9600, 8N1, no rtscts.

So then I do *SPOOL devices:$.Serial1, type some rubbish, and *SPOOL.

It 'hangs' at this point (ESCapeable).  I see nothing on my GW :(

Also doing 'cat > /dev/cuaa0' on the PC, typing some junk, and doing *TYPE
devices:$.Serial1 shows nothing (whilst the same command on Serial2 shows
chars from the IR port).

Likewise, *PPPConnect devices:$.Serial1 9600 <etc>  (can't remember
syntax) claims (in *PPPLog) that it's sending LCP stuff, but I see nothing
at the other end.


AAARRRRGH!  So my question... I'm assuming that because the serial port
*serialinfo output says it's not using handshaking, that it ISN'T using
handshaking.  I assume that if I *TYPE <serialport> it will open it with
the same parameters (just as though I'd done "stty < /dev/ttyS0;  cat
/dev/ttyS0" on a unix box).

Is there some weird default that it opens the port with every time that is
trying to use the (unconnected) RTS CTS lines, thus blocking on send?  I
don't know enough about RISCOS serial stuff, I think I tried something
like opening devices#9600baud;nocrts:$.Serial1 too (last month).

I don't see why the device would appear in RISC OS yet be unusable, so
there must be some other tweak that needs doing.  (I repeat: the hardware
works.  (that's the one bit I'm confident about))


Cheers,


Matt

ps: (anyone want to start a BushSTB demo compo?  Some of my little gfx
ditties for the RISCPC fly on this thing ;D)