Kill the previous command in a pipeline

Question

I am running a simulation like this

./waf --run scratch/myfile | awk -f filter.awk 

How can I kill the waf command as soon as filter.awk detects that something happened (e.g. after a specific line is read)?

I cannot change waf or myfile. I can only change filter.awk, and the above command (obviously).

Update after comments:

• waf does not terminated after receiving SIGPIPE (as it should?)
• It spawns child processes, that need cleaning up.

This is my own answer (and challenge).

---

After working on @thatotherguy's ans @Chris's answers, I simplified a bit and got this:

tmp=$(mktemp)
{ ./waf --run scratch/myfile & echo $! > "$tmp"; } | { awk -f filter.awk; pkill -P $(<$tmp); kill $(<$tmp); }

Unfortunately I could not get rid of the tmp file, every attempt to pass the PID as a variable failed.

I won't change the accepted answer (since it was the one that worked when it was really needed), but +1 for anyone that can simplify more.

Answer 1

Use awk's exit statement. waf should exit as soon as the pipe connecting it to awk closes.

Answer 2

What makes this tricky is that waf misbehaves by not exiting when the pipe breaks, and it spawns off a second process that we also have to get rid off:

tmp=$(mktemp)
cat <(./waf --run scratch/myfile & echo $! > "$tmp"; wait) | awk -f filter.awk; 
pkill -P $(<$tmp)
kill $(<$tmp)

• We use <(process substitution) to run waf in the background and write its pid to a temp file.
• We use cat as an intermediary to relay data from this process to awk, since cat will exit properly when the pipe is broken, allowing the pipeline to finish.
• When the pipeline's done, we kill all processes that waf has spawned (by Parent PID)
• Finally we kill waf itself.