Re: Flush serial port

[Alexandre Julliard]

Uwe Bonnes  writes:
> Changelog:
>    wine/server/file.c: create_file()
>    If we open a serial device, flush it

This should be done in serial.c, but could you please explain why this
is necessary?

Flush serial port

[Uwe Bonnes]

> > > > > "Alexandre" == Alexandre Julliard  writes:

Alexandre> Uwe Bonnes  writes:
    >> Changelog: wine/server/file.c: create_file() If we open a serial
    >> device, flush it

Alexandre> This should be done in serial.c, but could you please explain
    Alexandre> why this is necessary?

The flush is not necessary! Probably some error in my test program made me
think it was nesessary.

Flush serial port

[Uwe Bonnes]

> > > > > "Alexandre" == Alexandre Julliard  writes:

Alexandre> Uwe Bonnes  writes:
    >> Changelog: wine/server/file.c: create_file() If we open a serial
    >> device, flush it

Alexandre> This should be done in serial.c, but could you please explain
    Alexandre> why this is necessary?

Testing with the loopback adapter again, it clearly shows that flush is
necessary. I am not sure, if the flush should be done on open or close. I
decided to go for flushing with open().
If you want to test yourself, shortcut RXD to TXD and run the
kernel/test/comm.c with  LOOPBACK_TXD_RXD set to TRUE. The test for cbInQue
== 0 in test_ClearCommErrors() will fail with a newly opended FileHandle
when something was put into the receive buffer before. The test succeeds on XP.

A corrected patch is in wine-patches.

Flush serial port

[James Liggett]

Hi Uwe,
In my opinion, I do think something is amiss with Wine's handling of
serial port data after a program is done with them. Case in point: I've
been able to sync a serial Palm cradle with wine on the first try. After
that, Hotsync doesn't pick up the palm. I have to reboot before it works
again. Maybe a flush is necessary after all. Might the corrected patch
help the problem? ;-)

James Liggett

On Sun, 2005-07-31 at 16:12 +0200, Uwe Bonnes wrote:
> > > > > > "Alexandre" == Alexandre Julliard  writes:
> Alexandre> Uwe Bonnes <[e-mail]> writes:
> >> Changelog: wine/server/file.c: create_file() If we open a serial
> >> device, flush it
>
> Alexandre> This should be done in serial.c, but could you please explain
> Alexandre> why this is necessary?
>
> Testing with the loopback adapter again, it clearly shows that flush is
> necessary. I am not sure, if the flush should be done on open or close. I
> decided to go for flushing with open().
> If you want to test yourself, shortcut RXD to TXD and run the
> kernel/test/comm.c with  LOOPBACK_TXD_RXD set to TRUE. The test for cbInQue
> == 0 in test_ClearCommErrors() will fail with a newly opended FileHandle
> when something was put into the receive buffer before. The test succeeds on XP.
>
> A corrected patch is in wine-patches.

Flush serial port

[Uwe Bonnes]

> > > > > "James" == James Liggett  writes:

James> Hi Uwe, In my opinion, I do think something is amiss with Wine's
    James> handling of serial port data after a program is done with
    James> them. Case in point: I've been able to sync a serial Palm cradle
    James> with wine on the first try. After that, Hotsync doesn't pick up
    James> the palm. I have to reboot before it works again. Maybe a flush
    James> is necessary after all. Might the corrected patch help the
    James> problem? ;-)

You'll never know until you try :-)