Say I've started program and it continues writing some output to stdout. Is there some way to add a pipe so that I can, for example, count the number of lines it has produced with wc -l?

  • 1
    It may be possible, but how that would be done is probably dependent on what Unix you are using. – Kusalananda May 25 '18 at 20:41

You can do this in Linux. Say there's a script s:

echo my pid: "$$"
while true; do
    echo "$i"; ((i++)); sleep 1; 

Run it:

$ ./s
my pid: 8815

And so it goes... Now you need gdb to attach to the process.

gdb -p 8815

Now you're in gdb. Here do this:

p close(1)

This has closed the STDOUT file descriptor. Now open a new one:

p creat("/tmp/8815.out", 0600)

In another terminal you can already set the reading process:

tail -f /tmp/8815.out

Nothing's coming yet, as you need to detach from gdb in the other terminal. Either Ctl+D, or:

Quit anyway? (y or n) 

Confirm with y and Enter. And now you should see (where the tail is running) something like this:


If you prefer to use an already existing file or pipe, do this in gdb:

p open("/path/to/your/output", 1)

This illustrates redirecting STDOUT. It's similar with STDERR, only it has number 2.

I used these two answers:


You can use tee when you start the program. If you're using bash, you can do:

program > >(tee -a log.txt)

That will redirect the stdout of program to a process substitution running tee. That essentially writes a copy of the stdout of program to log.txt. tee will continue to send the stdout to wherever it was going before, probably the terminal depending on how program works.

Not the answer you're looking for? Browse other questions tagged or ask your own question.