Just some tools I've written for running jobs on CI systems like Travis, Shippable, etc.

This is a program similar to the travis_wait function - it runs a long-running command with output logged to a file. It occasionally prints a character to prevent timeout errors. If you hit the maximum runtime in minutes, it send a TERM signal, followed by a KILL signal. It exits with the same exit code or signal that your command exited with.

By default, it prints a period character every 30 seconds, and times out after 30 minutes. The log file name is generated randomly and printed on the first line of stdout. Each part of this can be customized:

runlog [-s sleeptime] [-c sleepchar] [-t timeout] [-l /path/to/log] [-h] -- program

• sleeptime is how often to print your character, in seconds
• sleepchar is the character (or string, if you want) to be printed.
• timeout is how long to wait before sending a TERM/KILL, in minutes
• -- (optional) can signal that you're no longer reading options.

The only thing printed to standard output is the logfile name with a newline, so you can run something like:

LOGFILE=$(./runlog -s 1 -t 1 -- /bin/sh -e -c "echo some output && sleep 90")
ecode=$?

if [ $ecode != 0 ] ; then
    printf "Error, here's the logfile ($LOGFILE)\n"
    cat "${LOGFILE}"
    exit 1
fi

And your output will be:

............................................................
Reached timeout, sending SIGTERM
Process ended with signal 15 (Terminated)
Terminated
Error, here's the logfile (/tmp/runlog-qenYlg)
some output

Another example:

runlog -s 1 -c 💩 -t 5 -- sleep 600

Will run the command "sleep 600" in the background. In the foreground, it will print the "💩" character to stderr every second for a maximum of 5 minutes. Once it reaches 5 minutes, it will send a TERM signal to sleep.

MIT-licensed, see LICENSE for details.