[Nut-upsdev] telling apart USB devices

[Peter Selinger]

Arnaud Quette wrote:
> 
> 2005/11/6, Peter Selinger <selinger at mathstat.dal.ca>:
> >
> > I just noticed the following curious behavior when using newhidups
> > with upsdrvctl and two different USB devices:
> >
> > ...
> >
> > On the other hand, if I assign two different dummy "ports" to these
> > devices:
> >
> > [belkinusb]
> > driver = newhidups
> > port = auto1
> > desc = "Belkin UPS, USB interface"
> > vendorid=050d
> >
> > [mgeusb]
> > driver = newhidups
> > port = auto2
> > desc = "MGE UPS, USB interface"
> > vendor = "MGE.*"
> >
> > then the two devices will be distinguished correctly. Is this behavior
> > documented? The man page currently says:
> >
> > IMPLEMENTATION
> > newhidups only support 1 UPS at this time. You can put the "auto" value
> > for port in ups.conf, i.e.:
> >
> > [newhidups]
> > driver = newhidups
> > port = auto
> >
> > This is now most definitely outdated. But the fact that the "port" is
> > used in this way, although it does not mean anything, should probably
> > be documented.
> >
> > Arnaud, any thoughts on this? -- Peter
> 
> yep, 2 in facts:
> 1) the above problem is due to the state sockets and driver pidfile naming.
> These are currently build using "drv-name"-"port" (adding ".pid" for the
> pidfile).
> This is, as you stated, not enough now with the newhidups case and its
> "auto" port probing!
> I've thinked about that when starting newhidups, and come to the conclusion
> that using the ups definition name (ie belkinusb and mgeusb) would be
> sufficient. A good way to ensure a unique filename. A special case is when
> the driver is launched manually (ie just with a port, no -a and not with
> upsdrvctl). In that case, we'll use the current behavior.
> I'm preparing that for the cvs Dev. Watch the commit list this evening...

I believe this will not quite work when starting the driver manually
without -a. In this case, there is no way for upsd to know what actual
device is connected to the driver, and therefore, there will be no way
for commands like "upsc upsname at host" to work. On the other hand, I
see no way around this problem. It means, if one wants to use upsd,
one always has to start the driver with "-a" (even for non-USB
devices). It this okay? Or do you want to change the behavior only for
newhidups, and retain the old behavior for all other drivers? 
 
> 2) having the above done, newhidups manpage can be updated, and the problem
> solved. Peter, I let you handle this part.
> 
> Arnaud