Web lists-archives.com

Re: Patch request to qt 5.9.4 (Re: [ANNOUNCEMENT] Qt 5.9.4)

Tatsuro MATSUOKA wrote:
----- Original Message -----

From: Yaakov Selkowitz <yselkowitz
To: cygwin
Date: 2019/3/11, Mon 09:53
Subject: Re: Patch request to qt 5.9.4 (Re: [ANNOUNCEMENT] Qt 5.9.4)

On Mon, 2019-03-11 at 09:28 +0900, Tatsuro MATSUOKA wrote:
 > On Mon, 2019-03-04 at 07:43 +0900, Tatsuro MATSUOKA wrote:
 > >  I ask alpply a patch the below which enables to use qt terminal
on gnuplot
 > for Cygwin.
 > >  (cygQt5Network-5.dll is affected.)
 > >
 > >  ---
 > >  +++
 > >  @@ -239,7 +239,7 @@ void QLocalSocket::connectToServer(OpenM
 > >      }
 > >      // create the socket
 > >  -    if (-1 == (d->connectingSocket = qt_safe_socket(PF_UNIX,
 > >  +    if (-1 == (d->connectingSocket = qt_safe_socket(PF_UNIX,
 > >          d->errorOccurred(UnsupportedSocketOperationError,
 > >          return;
 > >
 > It seems we keep going in circles on this point.  If there is a bug in
 > O_NONBLOCK, then please either narrow it down to an STC, or provide a
 > patch to Cygwin.

 Very basic question. What is STC?
 I googled but I cannot find what is it.


Sorry I cannot make STC due to lack of knowledge of QT.
In stead, I explain what gnuplot do.

On gnuplot for qt,
plot is done by child process named gnuplot_qt.

When gnuplot connect with gnuplot with

without the above patch ()
qt_safe_socket(PF_UNIX, SOCK_STREAM, 0, O_NONBLOCK)))
gnuplot cannot connect with gnuplot_qt.

with the above patch ()
qt_safe_socket(PF_UNIX, SOCK_STREAM, 0))
gnuplot can connect with gnuplot_qt.

Perhaps qt with O_NONBLOCK is some wrong with process connection.

Hi Tatsuro,
For the record, what Enrico supplied is what we'd call an STC, a Simple Test Case. Unfortunately his STC dealt with select() on a named pipe.

Your gnuplot issue has to do with "Unix sockets", a form of IPC between processes. Cygwin uses Windows named pipes to implement "Unix sockets" on Windows. This means we are mapping Unix/Linux socket constructs onto a dissimilar feature of Windows, and there are many opportunities for mismatch.

Can you run your gnuplot under strace? You would set up your environment as needed to reproduce the issue, then launch gnuplot as so:
    strace -o gnuplot.trace gnuplot
This will create a trace file of Cygwin syscalls and operations. Let us know how big (how many lines) are in the trace file and we'll see what's next.


Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple