[PATCH 1/5] use a correct timeout in test_waittxempty

Wolfgang Walter wine at stwm.de
Thu Sep 5 04:19:01 CDT 2013


Am Donnerstag, 5. September 2013, 17:51:18 schrieb Dmitry Timoshkov:
> Wolfgang Walter <wine at stwm.de> wrote:
> > > > When we send 17 bytes with 150 baud, no parity, one stop bit then
> > > > we need to wait at least 17*10/150 seconds > 1000 ms
> > > 
> > > Under Windows with both real COM-port and USB-serial cable this test
> > > takes
> > > no longer than 890 ms, usually it's even shorter like 875 ms. Testbot
> > > VMs
> > > also don't fail this test. If under Wine it takes > 1000 ms for you then
> > > probably there is a bug somewhere.
> > 
> > How can that be? If you send 17 bytes which expands to 170 raw bits and
> > you
> > only sent 150 bits/s?
> > 
> > Maybe you are using a virtual machine or your com port does not support
> > 150
> > baud and chooses a different speed. Or you use a UART with a large buffer
> > and which already signals TX-EMPTY even if it is still sending from an
> > internal buffer. But one should not rely on that.
> 
> I'm testing under 64-bit Windows 7 on real hardware.

Then it is your UART or it's driver. As I sad: you need 1133 ms to transmit 17 
bytes and I think one should not wait less. Otherwise one depends on hardware, 
driver internals (linux as windows) and scheduling. One should even wait 
longer (that is the reason I add TIMEOUT).

Regards,
-- 
Wolfgang Walter
Studentenwerk München
Anstalt des öffentlichen Rechts
Abteilungsleiter IT
Leopoldstraße 15
80802 München



More information about the wine-devel mailing list