Re: [avrdude-dev] patch #5655, again

[Joerg Wunsch]

As Jason von Nieda wrote:

> I'd be interested in following this discussion on list as it's
> something I have had to deal with as well. For the record, I am
> using basically the same code in my local version of avrdude so that
> it will work with my mkII.

Windell sent me the trace files, and I've noticed that the sign-on
command is being sent much earlier than the subsequent commands.
Maybe, somehow the USB driver in MacOS is simply not yet ready for
sending data there.  I've asked him to replace the "return" hack by a
"sleep(1)".  You could do the same if you want.  If that works, we
could try narrowing it down to some amount of usleep(), perhaps even
wrap it up in a loop until it eventually succeeds.  (It's possible the
current loop is executed too quickly still, but it's also possible the
USB driver jams once being presented a request that came in too
early.)

But that's suspicions so far.

> Another note: I have read that one difference between Mac serial
> ports and Linux serial ports is that the Mac automatically sets DTR
> upon opening the port. Could possibly be related?

Are you sure Linux doesn't do this?  I'd only suspected Windows about
that, Unices typically assert DTR when opening a port, but I never
really analyzed Linux for that (I'm sure about FreeBSD).

However, the communication to the AVRISPmkII is not RS-232, not even
an emulation of it over USB (like CDC), so that doesn't matter here.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)