Re: [avrdude-dev] Parallel port doesn't keep programmed states

[E. Weddington]

On 24 Nov 2003 at 13:26, Jan-Hinnerk Reichert wrote:

> On Monday 24 November 2003 00:58, idefixx wrote:
> > > You need to disable Windows XP Plug'n'Play for the parallel port.
> > >
> > > I solved the problem with the following file:
> > > http://www.student.dtu.dk/~s991363/avr/avrdude/winxp1.reg
> >
> > But is it really necessary to change the windows registry? What
> > changes are going on during the read out of the AVR flash, as after
> > doing a read out the problem disappears, too.
> 
> I don't know the windows ppi-code. However, there should be no 
> difference in the called functions, since we only transfer 
> SPI-commands.
> 
> I guess there is either some magic sequence occuring on the port while 
> reading or (more likely) it has something to do with the frequency of 
> access. For _unpaged_ memory writing does big pauses when writing. So 
> perhaps windows thinks there is a valid device and stops looking, if 
> there only is enough "action" on the port.
> 
> Have you tried the latest CVS source? Recent changes in the polling 
> code make writes look more like reads in this regard. There is a good 
> chance that this makes your problem go away ;-)
> 
> /Jan-Hinnerk
> 


I'm not necessarily an expert on Windows and drivers for Windows, but from what 
I 
remember, I believe it has to do with the parallel port driver we have to use, 
giveio, 
and Windows XP. I think that the giveio driver is very simple and doesn't "play 
well 
with other" drivers, especially the parallel port driver on WinXP and conflicts 
arise 
causing other signaling on the parallel port pins, i.e. not from avrdude. 
Disabling 
"Plug'n'Play" seems to stop this problem on WinXP. The file that was suggested, 
takes care of disabling Plug'n'Play via the registry, and makes a nice, 
automatic way 
(i.e. from the command line) to do this.

What is really needed is an open source, Windows parallel port device driver, 
that 
works well with Windows NT, 2K, XP just for avrdude (and for uisp, too) that 
allows 
direct writing and reading of the parallel port pins. The giveio driver is just 
too simple 
(IMHO) and not robust enough. However, this will take a volunteer who has 
device 
driver experience and the tools to build it, neither of which I have.

If I'm incorrect about anything, feel free to correct me. Also, if Alex 
Shepherd (NZ) is 
listening in, perhaps he'll have something to say about this.

Eric