More fhandler_serial fixes.

[Sergey Okhapkin]

>
> I could see changing this to size_t but not adding the extra vmin_.
>
> >@@ -423,20 +426,21 @@ fhandler_serial::ioctl (unsigned int cmd
> >                                               0, &mcr, 4, &cb, 0);
> >                if (!result)
> >                  {
> >-                   __seterrno ();
> >-                   res = -1;
> >-                   goto out;
> >+                   modem_status |= rts | dtr;
>
> This is saying that if a function (DeviceIoControl) fails we fallback to
> Win95 behavior.  Why wouldn't we want to reflect the failure of the
> function?  Otherwise we could be masking an actual failure.
>
> Anyway, I've checked in a variation of this change minus the
> DeviceIoControl part.

The call to DeviceIoCtl is
     BOOL result = DeviceIoControl (get_handle (), 0x001B0078, NULL,
                           0, &mcr, 4, &cb, 0);

I heve no ideas what does this call means, it's undocumented. The call
always fails on my W2k when
attempting to get the status of laptop's built-in modem. It seems to me it's
better to ignore the error and to fall back to Win95 behavior rather than
return the error.

Sergey Okhapkin
Somerset, NJ