Date: Sun, 31 Jan 1999 10:37:10 -0800 (PST) From: John Polstra <jdp@polstra.com> To: robert+freebsd@cyrus.watson.org Cc: hackers@FreeBSD.ORG Subject: Re: Correct way to write a thread-safe library Message-ID: <199901311837.KAA07195@vashon.polstra.com> In-Reply-To: <Pine.BSF.3.96.990130120203.1285A-100000@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <Pine.BSF.3.96.990130120203.1285A-100000@fledge.watson.org>,
Robert Watson <robert+freebsd@cyrus.watson.org> wrote:
>
> So here I am, chugging away at a library I'm writing, and I notice that
> the spec requires me to report errors via errno. I immediatly recall a
> long discussion of errno changes a long time ago and macro definitions of
> it. So my question is this--what is the correct way for me to write a
> thread-safe library that can be used by threaded and non-threaded code,
> and that also uses errno to report errors?
The general idea is that in <errno.h> or <sys/errno.h> you define
"errno" as something like this:
#define errno (* __error())
and then make "__error()" return a pointer to where the possibly
thread-specific value is stored. You can look at our existing libc
and libc_r code for an example.
The C standard explicitly allows for this by requiring "errno" to be
an lvalue but not necessarily a simple variable. It also requires
compilation units that reference "errno" to include <errno.h>.
John
--
John Polstra jdp@polstra.com
John D. Polstra & Co., Inc. Seattle, Washington USA
"Nobody ever went broke underestimating the taste of the American public."
-- H. L. Mencken
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901311837.KAA07195>