gcc will interpret the constant value as a uint32 but
the port's set_property() was taking it as a uint64. Thus
the top 32 bits were probably garbage, and messed up
on big-endian architectures leading to random large
probe delays.
The standard dictates CSQ response strength value to be [0 - 31]
inclusive, and 99 means "unknown" or "no service". Make that
apparent and don't treat 99 as 99% which it clearly isn't. Also,
allow spaces in the CSQ response.
Phones especially don't seem to consistently implement this. For now,
we'll hack it out, but later, we'll want to have a class method for
power-on instead of just a property so that subclasses can decided for
themselves (since they know their hardware better) whether failure
of the power-on command is fatal or not.
Full references prevented destruction of the modem object if
it was unplugged or somehow removed. To fix that using full
references on the modems would require that all usage of
MMCallbackInfo to be aware of the validity of the modem and to
ensure the callback was called whenever the modem became invalid.
That, needless to say, would suck. Since any in-progress calls
can't complete when the modem is invalid anyway, just have the
MMCallbackInfo object return a generic error when the modem goes
away and the call is still in-progress.
Like UMTS vs. GSM, EVDO and 1x are separate networks and technologies
and have separate registration state. You can even be roaming on
EVDO while in your home 1x network. Handle that.
rfcomm devices seem to be created as 'virtual' devices first, without
any parents, then moved to the right place in the device tree. So
handle moves too; if the modem was already found in the 'add' phase
it'll be ignored in the move phase.
Two hacks here:
1) rfcomm ports don't have an easily accessible driver name, so we just
match the parent's subsystem to 'bluetooth' and use that
2) libgudev doesn't seem be be able to get the rfcomm device's device file,
which would normally be /dev/rfcommX. Oh well, we don't use the device file
yet anyway