#36 Console I/O stalls

[Greg Williams]

While the new console I/O buffering scheme for NppExec is much faster, it has a shortcoming when dealing with small status updates of I/O before completion of the launched external program.

I have a utility that takes about a minute to execute, and outputs short lines of progress text along the way. I think it is because the buffer size limit is not reached that the progress text does not appear until the external application terminates. It would be good to have a periodic timer that will flush the I/O buffer every 500ms or so in order to handle situations like this.

BTW, thanks DV for a great plugin!

[DV]

Logged In: YES
user_id=1468738
Originator: NO

This is an application-specific behaviour. For example, this AWK script:

BEGIN {
for (j = 0; j < 100; j++) {
timeout()

printf( "\b\b%02d", j )

timeout()
}
}

function timeout() {
for (i = 0; i < 100000; i++) {
sin(i*3.14 + 0.0319319)
}
}

shows the same behaviour as you described.
But the following AWK script, which is almost the same as previous, prints the progress correctly:

BEGIN {
for (j = 0; j < 100; j++) {
timeout()

printf( "\b\b%02d", j )
fflush()

timeout()
}
}

function timeout() {
for (i = 0; i < 100000; i++) {
sin(i*3.14 + 0.0319319)
}
}

The only difference - fflush() - forces AWK to write the output string immediately. Whereas the printf() function alone does not guarantee immediate appearance of the output string in the pipe.

[Greg Williams]

Logged In: YES
user_id=1970571
Originator: YES

This is true. I created a Ruby script that flushes I/O before and after some sleeps, and I see the proper behavior. The place where we're seeing the errant behavior in in a Windows C++ console application.