This is the mail archive of the mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Cygwin-Windows pipe communication

I have this scenario:

The parent process forks and execs a child process which is a Windows
program. The child's stdin is connected (via a dup2) to a pipe that
it uses to get input from the parent and writes it to a file. 

The parent writes data to the pipe and then closes its end of the
pipe, but I find that the child hangs around waiting for more data. I
don't see this problem if I exec a Cygwin process to do the same
thing instead. 

It seems that a pipe close by the parent is not recognized by a child
Windows process. So, I tried a little experiment. 

Let's assume the parent writes to fd[1]. Before the child execs, it
closes its copy of fd[1] and does dup2(fd[0], 0). Now, if I set the
close_on_exec flag on fd[1] IN THE PARENT before the fork, all is
fine -- the child knows that the parent has closed the other end of
the pipe. However, if I set that flag IN THE CHILD before it execs,
the problem persists.

Are there any known limitations in the Cygwin pipe implementation
when the pipe is between a Cygwin process and a Windows one? Why does
setting the close_on_exec flag in the parent get rid of the problem?
And why doesn't setting it in the child work?


Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!

Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]