Date: Fri, 27 Feb 1998 13:38:59 -0500 From: "Kaleb S. KEITHLEY" <k.#nojunk#keithley@opengroup.org> To: hackers@FreeBSD.ORG Subject: Re: symbols in libc_r not in libc Message-ID: <34F70843.2781E494@opengroup.org> References: <199802271257.NAA24819@ws6423.gud.siemens.at>
next in thread | previous in thread | raw e-mail | index | archive | help
marino.ladavac@siemens.at wrote: > > > > > On a 3.0-971225-SNAP system, when I build Xlib thread-safe I use > > -D_REENTRANT and -D_THREAD_SAFE, and all references to errno are > > converted to *__error(). No problem there. > > > > Now when I link a non-threaded program, e.g. xterm, to the mt-safe Xlib, > > when I try to run it I get unresolved externals for ___error. > > > > What that tells me is that libc (not libc_r) needs a #pragma weak > > __error so that when I link non-threaded programs against libX11, > > they'll work, and when I link a threaded program with libX11 and libc_r, > > it'll get the right version of __error. > > > > As far as I can understand weak symbols, you can add your own version of > __error() in Xlib, as a weak symbol. In my not so humble opinion, that's a hack. > That way, the programs that do not > use libc_r get the __error() from Xlib, and the ones that do from libc_r > (I take it that the libc_r __error() is a strong symbol). This should > give you exactly what you need exactly where you need it. > > The Xlib implementation of __error() should naturally just return the > address of errno. Don't forget to undefine errno when you implement > the Xlib __error(). > > Hope I didn't miss anything. Well, just that Xlib isn't in the business of providing libc functions or putting a band-aid over a broken libc. The weak __error() function belongs in libc. -- Kaleb S. KEITHLEY 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?34F70843.2781E494>