Subj : Re: double-checked locking in C
To   : comp.programming.threads
From : Charles Bryant
Date : Sun Jul 10 2005 08:43 pm

In article <zaudnbhMuPTj_1DfRVn-rg@comcast.com>,
Joe Seigh  <jseigh_01@xemaps.com> wrote:
>So what kind of value could PTHREAD_ONCE_INIT have that couldn't be
>memcpy'd from an initialized pthread_once_t object never otherwise
>used?

Part of it might contain the address of another part of it. Of course, that
would have to be achieved with the assistance of the compiler, as a plain
#define wouldn't be enough. There's no obvious reason *why* it should
contain such a thing, but I assume the intent in the standard is to leave
the greatest freedom for implementations.

.