my own sleep program vs the sleep command

**Skaperen** · December 15th, 2019

i wrote my own sleep program and bash behaves different with it than it does with the sleep command. could it be a secret built-in command? the bash man page has no references to "sleep" whatsoever. the different behavior is when my program (named "nsleep") runs before another on one command line ("nsleep 999999;date") and ctrl-C is pressed. with my nsleep the date command is run anyway. with the standard sleep command, date is not run. bash is doing it different. what causes this? is bash getting a special signal? is bash treating sleep differently?

Code:

lt2a/forums /home/forums 2> sleep 999999;date
^C
lt2a/forums /home/forums 3> nsleep 999999;date
^CSat Dec 14 20:17:40 EST 2019
lt2a/forums /home/forums 4>

**sudodus** · December 15th, 2019

At least in 18.04.x LTS sleep is an executable program (compiled, ELF can be seen at the head of the file /bin/sleep. You find some details via

Code:

man sleep

and it is possible to find its source code. I guess there is nothing to trap the interrupt. You can compare with the shell built-in read

Code:

read -st5 -n1 ; date

that behaves like sleep för ctrl+C but differently if you stop it gracefully with for example Enter.

**Skaperen** · December 15th, 2019

sleep has always been a binary executable in ever Unix/BSD/Linux system i have ever used on every architecture. my nsleep is also a binary executable that has been built successfully, and run OK, on a few Linux distros and also on OpenBSD. i have seen this behavior on Slackware Linux (32 bit and 64 bit), Timesys Linux on an ARM based eval board (32 bit only), and on Ubuntu Linux (64 bit only) (every LTS since 14.04. i have not tried it on others for various reasons.

if i use && instead of ; then date does not run for both programs being interrupted. i am considering writing a new version in Python3.