Lecture 28 – Apr 20th, 2020

Setup

1. Log in to clyde.
2. Copy ~steve/ex/sig to your home directory.

Task

1. cd into sig and look at block.c. It installs a signal handler for SIGINT, masks SIGINT, sleeps for three seconds, and then unmasks SIGINT.

   Run $ make to compile block. Run $ ./block and let it sit until it exits. Nothing should be printed.

   Run it again, but this time, press ctrl-C a few times while it is sleeping. See how many times the message printed from the signal handler happens. (Note that although we cannot use printf(3) from a signal handler, we can use write(2).)

2. Now, take a look at sig.c. Modify the code such that when ctrl-C is pressed, interrupt_count is incremented and when ctrl-\ is pressed (which sends SIGQUIT), done is set to 1 to exit the loop. Make sure to uncomment the loop.

   To do this, you’ll need to use sigaction(2) twice to install a signal handler for SIGINT and SIGQUIT. You can use the same function or two different functions. I suggest copying the relevant code from block.c.

3. Build and run ./sig. Try pressing ctrl-C a few times and then press ctrl-\. It should print the number of times you pressed ctrl-C.
4. Run ./block in a short loop

   $ for i in {1..3}; do ./block; done
   

   and press ctrl-C while it is running.

   You’ll notice that after SIGINT is unmasked and the program exits, the for loop keeps running and block will be run again. This is because the signal handler caught and handled the signal so Bash didn’t know that it had exited due to a signal.

5. Change the sa_flags to be SA_RESTART | SA_RESETHAND. This will reset the signal handler to the default behavior as soon as the signal is delivered to the handler.

   Add raise(sig) at the end of the handler to raise the signal again.

   Rerun the loop from part 4 above and press ctrl-C. This time, it should exit immediately.