Re: [boost] detail::atomic_counter

From: Jens Maurer (Jens.Maurer_at_[hidden])
Date: 2001-11-18 13:17:05

Mattias Flodin wrote:
> Speaking of atomic access, I've been thinking about the sig_atomic_t type
> discussed earlier. Wouldn't it be useful with an portable atomic-access
> integer type? In systems where sig_atomic_t exists as an atomic type, this
> could be simply a typedef.

I'm against overloading the semantics of sig_atomic_t.  The type
sig_atomic_t serves a specific purpose:  Communications between signal
handlers and the "main" program.  We should not put threading semantics
on top of it.

Jens Maurer