From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 15089 invoked by alias); 1 Nov 2010 09:56:53 -0000 Received: (qmail 15081 invoked by uid 22791); 1 Nov 2010 09:56:52 -0000 X-SWARE-Spam-Status: No, hits=-1.9 required=5.0 tests=AWL,BAYES_00,T_RP_MATCHES_RCVD X-Spam-Check-By: sourceware.org Received: from eddie.linux-mips.org (HELO cvs.linux-mips.org) (78.24.191.182) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Mon, 01 Nov 2010 09:56:46 +0000 Received: from localhost.localdomain ([127.0.0.1]:37570 "EHLO localhost.localdomain" rhost-flags-OK-OK-OK-OK) by eddie.linux-mips.org with ESMTP id S1491172Ab0KAJ4n (ORCPT ); Mon, 1 Nov 2010 10:56:43 +0100 Date: Mon, 01 Nov 2010 09:56:00 -0000 From: "Maciej W. Rozycki" To: Richard Sandiford cc: binutils@sourceware.org Subject: Re: [PATCH] MIPS/GAS: Fix NewABI reloc handling with the LD/SD macro In-Reply-To: Message-ID: References: <87wrp6m03j.fsf@firetop.home> User-Agent: Alpine 2.00 (LFD 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-IsSubscribed: yes Mailing-List: contact binutils-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: binutils-owner@sourceware.org X-SW-Source: 2010-11/txt/msg00002.txt.bz2 On Mon, 1 Nov 2010, Richard Sandiford wrote: > > Given the current state of affairs I'll post the testcase I have in mind > > separately later on. Note that the LD failures spotted by Alan are > > indirect only -- we seem to be lacking a direct test, so one will be good > > to have IMO. > > It cuts both ways. I see the binutils testsuite as effectively a unit, > rather than as a collection of separate component testsuites (gas/, ld/, > binutils/). So while you could say that assembler-only relocation tests > are more "direct" than ld tests, you could also say that they're less > complete than the assembler+linker tests in ld/. You're right, but that's not what I had in mind writing that -- these tests (be they from the binutils or GAS or LD subset -- as you say that really does not matter here) have not been specifically designed to test compound relocations, but some other features and they trigger and fail by chance only. That is of course a good and desired side effect and as you may have noticed I even try to induce some side effects deliberately (or just notice their presence ;) ) with some test cases I make, but I think that's a feature that deserves an explicit check guaranteeing a better coverage. I have a rough idea how to do that, but I'll need some time to implement it properly and I simply ran out of last weekend. > > gas/ > > * config/tc-mips.c (macro)[M_LD_OB, M_SD_OB]: Use the offset > > reloc supplied. > > (mips_ip)['o']: Initialise offset_reloc. > > OK. Applied now, thanks. Maciej