Re: pgbench: new feature allowing to launch shell commands

The threading bug appears when a duration is set for pgbench tests.
Instead of a duration, if a number of xacts is set, this error doesn't
happen.

If i understood the problem well, when the alarm signal comes, all the
threads have to disconnect even the ones looking for a setshell parameter at
this moment, creating the thread error:
setshell: error getting parameter
Client 0 aborted in state 1. Execution of meta-command failed.

I am trying to find an elegant way to solve this, but I can't figure out yet
how to deal with this error as it looks tricky.