tcdrain is always blocking ...

[veeso]

tcdrain is always blocking ...

[veeso]

Sorry if I haven't answered you for a few days, but I'm very busy these days.
Well, I'm not sure at 100% it works like that, but from the termiwin documentation:
"tcdrain() waits until all output written to the object referred to by fd has been transmitted."

So, isn't it supposed to be blocking?

[Apollon77]

Yes, But you initialize the "CreateFile" in a way (as I understand it) that a Write-call only returns after write is complete ... you use it "non-overlapping" = "Synchron".
On Linux it seems to be "overlapping = asynchronous" by default (?) and so in Linus you do the write call and then wait using "drain" till it was written.

But on windows the Write returnsonly after the write ewas complete, so a drain wait forever because No other write can happen ...

[veeso]

Ok, understood.
Have you tried to replace the last NULL argument with FILE_FLAG_OVERLAPPED ?
I can't try by myself at the moment.

[Apollon77]

Yes, got an compile error then for some other code part, but not started to look deeper into it :-)
So it seems that you also need to consider other things when using overlapping mode.

If you want I can try to reproduce?!

[Apollon77]

Hm ... tried it (it is the parameter after "OPEN_EXISTING") ... no compile error anymore ... hm ...
Will try later if it still works

[veeso]

yeah sorry, was the 6th argument, not the 7th

[Apollon77]

test bresults:

with

		com.hComm = CreateFile(com.port, GENERIC_READ | GENERIC_WRITE, 0, NULL, OPEN_EXISTING, FILE_FLAG_OVERLAPPED, NULL);

I get Error 87 on write to that serial socket.
According to https://msdn.microsoft.com/en-us/library/windows/desktop/ms681382(v=vs.85).aspx it is:

ERROR_INVALID_PARAMETER

    87 (0x57)

    The parameter is incorrect.

Also tried "FILE_ATTRIBUTE_NORMAL | FILE_FLAG_OVERLAPPED" as in one Windows example from https://msdn.microsoft.com/de-de/library/windows/desktop/bb540534(v=vs.85).aspx

More infos an overlapping I found on https://msdn.microsoft.com/en-us/library/windows/desktop/aa365683(v=vs.85).aspx ...

for now I remove tcdrain usage with ifndef _WIN32 ... :-(

[Apollon77]

Ok, digged deeper :-)

When you do Overlapping communication then you also need to provide an OVERLAPPED struct to Read/Write operations and they become Really async. means: Directly after write you only get an info back but no info on bytes written. This you need to query with other commands and such.

But I had an other idea: When I use this:

int tcdrain(int fd) {
    if(fd != com.fd) return -1;
    return FlushFileBuffers(com.hComm);
}

it works as expected or ?!

See also: https://www.winehq.org/pipermail/wine-patches/2013-August/126115.html

what you think ?!

[Apollon77]

PS: I already use it that way in my version and it works :-)

[veeso]

Ok, so it's enough to add FILE_FLAG_OVERLAPPED to createFile and change tcdrain in the way you did, right?
Does FILE_FLAG_OVERLAPPED have any effect if I use it but I don't need an asynchronous communication (I mean, should it be used only if someone needs an asynchronous communication, or it can be used everytime?)

[Apollon77]

NO!
FILE_FLAG_OVERLAPPED at all means higher effort. So leave all as is and only use FlushFileBuffers in the drain method.
Normally all data are alreyd written when Write returns, and if there are some left then FlushFileBuffers will block until they are written (as drain means) ... if everything is already written FlushFileBuffers simply does nothing. :-)

[veeso]

Ok, thank you