[Bug 11339] New: serial.c:wait_on() sanity checks seem to be too strict

http://bugs.winehq.org/show_bug.cgi?id=11339

Summary: serial.c:wait_on() sanity checks seem to be too strict
           Product: Wine
           Version: unspecified
          Platform: Other
        OS/Version: other
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P2
         Component: ntdll
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=10448)
 --> (http://bugs.winehq.org/attachment.cgi?id=10448)
Trivial patch to fix the problem

In its present state, wait_on() function implemented in ntdll/serial.c fails
more often then it should. This means that in some cases serial events are not
reported to the calling Windows program (even if in practice they could be)
leading to application malfunction.

The problem is in the sanity check:
  if ((status = get_irq_info(fd, &commio->irq_info)) ||
      (status = get_modem_status(fd, &commio->mstat)))
       goto out_now;

which bails out and silently returns 0 if low-level Unix kernel driver for
serial device doesn't implement TIOCGICOUNT or TIOCMGET ioctl()s. This is, in
fact, intended behavior if the calling program wants to monitor events like
EV_BREAK, EV_ERR or EV_CTS and friends (otherwise, Wine will loop forever in
wait_for_event()) but is useless and even harmful if the event in interest is
EV_RXCHAR or EV_TXEMPTY. In latter case the check just prevents the program
from working correctly.

I've came across this bug when trying to make control software for AnyData CDMA
modems (EasyWirelessNet 1.00.xx) working. In Linux, those devices use "option"
driver, which lacks TIOCGICOUNT ioctl().

The trivial patch is attached. It's not against git clone of Wine so it is not
meant for applying directly.