From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2005-11-07 02:03:29
> Is there any defined behavior of the execution_monitor if the controlled
> function starts new threads? Does every thread need its own monitor or
> is one for the main thread from where the others are started sufficient?
> Thanks in advance,
Sorry for delay in reply. My understanding is that you are trying to use
boost::execution_monitor - part of Boost.Test library. Unfortunately this
component is not thread safe. I've made sure that it reentrant safe. But to
access it from multiple threads require adding guards for static data and/or
use of TLS. I did not have time to do this and frankly this is not in my
immediate to do list. There are some non trivial issues to consider. So you
only options is single monitor in main thread. And even in this case I am
not sure if it will work. Especially in regards of exceptions and signals
handling. You probably will need to go extra mile to get what you want.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk