dtrace(1) handle SIGPIPE, and handle signals when listing probes

[markj]

Summary

It's occasionally useful to do something like

dtrace -lv -P 'pid$target' -p <target PID> | less

to browse the pid provider probes of the target process. However, quitting less(1) will cause dtrace to receive SIGPIPE and exit, which in turn causes the target process to exit since it's being ptrace(2)ed. It's also conceivably useful to run the output of a dtrace script through a pipe.

This change does two things. First, it installs a handler for SIGPIPE - the same handler that's currently used for SIGQUIT and SIGINT. If a dtrace program is executing, it'll cleanup and exit if it sees SIGPIPE. Second, in list mode (-l), it'll break out of the for-each-probe loop and clean up.

Test Plan

The above example would reliably kill the target process. Now dtrace exits cleanly, and the target process is able to continue running. Similarly, ctrl-Cing dtrace when it was attached to a process and listing a large number of probes would cause the target process to be killed; that no longer happens.

[markj]

Closed by commit rS272455 (authored by @markj).