From owner-freebsd-stable Mon Apr 12 6:20:30 1999 Delivered-To: freebsd-stable@freebsd.org Received: from dominator.eecs.harvard.edu (dominator.eecs.harvard.edu [140.247.60.28]) by hub.freebsd.org (Postfix) with SMTP id F2ADC15524 for ; Mon, 12 Apr 1999 06:20:28 -0700 (PDT) (envelope-from karp@eecs.harvard.edu) Received: (from karp@localhost) by dominator.eecs.harvard.edu (8.6.12/8.6.12) id JAA18842; Mon, 12 Apr 1999 09:18:03 -0400 Date: Mon, 12 Apr 1999 09:18:03 -0400 From: Brad Karp Message-Id: <199904121318.JAA18842@dominator.eecs.harvard.edu> To: stable@freebsd.org Subject: __deregister_frame_info and egcs-1.1.2, revisited Cc: jdp@polstra.com Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I saw the exchange about the rtld fix to map unresolved weak references to zero, so that one can link with the system cc against egcs-built shared libs. I've been having the undefined symbol lossage, too. But here's the twist: I experience it when I compile *with* egcs against egcs-built shared libs. I admit that I could update my rtld from 3.1-RELEASE, and silence rtld about these symbols. But the obsessive-compulsive part of me is bothered that things don't work even when linking with egcs. As Fritz wrote, egcs' crt{end,begin,endS,beginS}.o reference the *_frame_info functions, and there is text for them in egcs' libgcc.a. I'm wondering if the elf ld needs a tweak, too--it appears that even when egcs is used for the link, the linker fails to bring in the *_frame_info code from libgcc.a. Interested parties might cf. my note to ports, which includes a transcript of the verbose egcs output for the failure case: http://www.freebsd.org/cgi/mid.cgi?db=irt&id=199904110709.DAA12115@dominator.eecs.harvard.edu -Brad, karp@eecs.harvard.edu To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message