Re: Left Bliz1230mkIV compiling, found it had crashed.

[Roc Vallès <vallesroc%gmail.com@localhost>]

On Wed, Dec 11, 2013 at 9:13 AM,  <is%netbsd.org@localhost> wrote:
> Hi Roc,
>
> May I answer this in public? Because it might be useful to somebody
> else.
>
Feel free. I've just replied to the list instead.

> On Tue, Dec 10, 2013 at 09:33:33PM +0000, Roc Vallčs wrote:
>> No luck, crtscts both sides, it still won't be reliable above 38400.
>
> what sort of cable? null-modem?
>
Yes... A DB9 one. And a DB9 to DB25 adapter. So many possible points of failure!
> For DB25, you need:
>
> 2-3, 3-2, 4-5, 5-4, (6+20)-8, 8-(6+20), 7-7.
>
> (To connect to a modem, you need 2-2,3-3,4-4,5-5,6-6,7-7,8-8,20-20).
>
> Looking up the DB9 pin numbers is left as an exercise for the reader.
>
>> I'm assuming
>
> No good assuming. *Check*.
>
I've enabled hardware handshake. If it wasn't wired, it'd not work at
all, would it?

>> the cable does have hardware handshake wired. I'd expect
>> it to not work at all with hardware handshake on if it did wasn't
>> wired...
>
> Ahem. I'd expect it to not work reliably at all with ser0 over 38400.
>
38400 is too much for an A1200 with 030? Ouch.

See:
https://en.wikibooks.org/wiki/Aros/Platforms/68k_support#Serial_Debugging

These people are wrong, then? Or it's just a netbsd limitation?

> (That's why there are modern hardware solutions with 16, 64 or even more
> bytes of hardware FIFO for the RX.)
>
> But you shouldn't see any more silo overflows at 38400 or 19200, or
> at least only very occasional.
>
> Looking at the code to verify:
>
> ser0: silo overflow
> -> hardware detected that the interupt routine didn't read out the
>    data & status word before the next character input was finished.
>
> ser0: N ring buffer overflows.
> -> the low priority interupt didn't manage to read out the RX data
>    from the software maitnained ring buffer fast enough.
>
> With proper handshaking and not too fast speed, both shouldn't happen,
> or at least not a lot.
>

Situation seems to improve with crtscts but it still logs silo
overflows, so it might be placebo. I'm lazy but will eventually get
around grabbing both ends and checking continuity with a multimeter
while looking at some diagram. You've done all you can on your end.

> Regards,
>         -is