Even when SIGPIPE has the default behaviour, it seems that there is a timing issue if grep has the time to treat EPIPE as an error before the process being killed due to the delivery of the signal.
This seems to happen easily when running on Android (perhaps due to slower signal delivery, or relatively faster output, or how it runs on a multi-cpu setup), as seen from the below run: $ seq 9999 |grep -v xxx |head -1 1 grep: write error: Broken pipe This would motivate silently ignoring EPIPE to avoid having different behaviour depending on timing issues. See also http://superuser.com/questions/554855/how-can-i-fix-a-broken-pipe-error.
