[95] in Pthreads mailing list archive
Re: pthreads porting help!
daemon@ATHENA.MIT.EDU (Ken Raeburn)
Thu Aug 17 17:22:35 1995
Date: Thu, 17 Aug 1995 16:57:10 -0400
From: Ken Raeburn <raeburn@cygnus.com>
To: Christopher Provenzano <proven@MIT.EDU>
Cc: neal@ctd.comsat.com (Neal Becker), pthreads@MIT.EDU
In-Reply-To: Christopher Provenzano's message of Wed, 16 Aug 1995 17:18:52 EDT
From: Christopher Provenzano <proven@MIT.EDU>
Date: Wed, 16 Aug 1995 17:18:52 EDT
> Then we need a lot of fixing up, for example assignments from
> __SIGEMTPYSET.
>
> Yup.
Take a look at machdep/sunos-5.3/__signal.h. Solaris uses a non integral
sigset_t too.
Um, Chris, I just looked at this. Do you really want to write out two
static sigset_t variables into every file that includes <signal.h>?
I think my suggestion of defining a static const variable and
permitting __SIGEMPTYSET and __SIGFILLSET to be any valid initializers
would work better. Actually, since we assume ANSI C, I don't think
the variable has to be static or const. If it's done that way, the
performance should be just as good as the current code when sigset_t
is an integral type, and probably not much worse (if at all) for
aggregate types.
int sigemptyset(sigset_t *set)
{
sigset_t empty = __SIGEMPTYSET;
*set = empty;
return 0;
}
