From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 10800 invoked by alias); 6 Apr 2005 22:27:51 -0000 Mailing-List: contact gdb-patches-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sources.redhat.com Received: (qmail 10770 invoked from network); 6 Apr 2005 22:27:48 -0000 Received: from unknown (HELO nevyn.them.org) (66.93.172.17) by sourceware.org with SMTP; 6 Apr 2005 22:27:48 -0000 Received: from drow by nevyn.them.org with local (Exim 4.50 #1 (Debian)) id 1DJIzz-0006jr-JG; Wed, 06 Apr 2005 18:27:47 -0400 Date: Wed, 06 Apr 2005 22:27:00 -0000 From: Daniel Jacobowitz To: gdb-patches@sources.redhat.com, Richard Earnshaw Subject: Re: RFA: Update ARM architecture selection, add set arm abi Message-ID: <20050406222747.GA25866@nevyn.them.org> Mail-Followup-To: gdb-patches@sources.redhat.com, Richard Earnshaw References: <20050330203140.GA10953@nevyn.them.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050330203140.GA10953@nevyn.them.org> User-Agent: Mutt/1.5.8i X-SW-Source: 2005-04/txt/msg00063.txt.bz2 On Wed, Mar 30, 2005 at 03:31:41PM -0500, Daniel Jacobowitz wrote: > This patch redoes the way the ARM target handles the architecture's ABI and > floating point model. Right now we adjust the gdbarch in place when the FP > model changes; but gdbarches are supposed to be read-only after their > creation. What we should do instead is use gdbarch_update_p for this > purpose. This also separates the ABI from the OSABI, for consistency with > other architectures; this fixes assorted problems, including the core dump > consistency discussed on gdb@ this week, and the use of the default OSABI > for untagged binaries. If you've ever had GDB set the wrong breakpoint > instruction when debugging a GNU/Linux ld.so, this is why; the patch should > fix it. > > When creating an architecture we need to honor settings in this order: > - Global settings set by the user. > - Settings inferred from the current binary. > - Settings inherited from the last ARM architecture created. This > is usually equivalent to the settings it inferred from the binary; > when gdbarch_update_p is used the new architecture is created without > a BFD. > - Defaults chosen by the architecture, in absence of anything else. > > OK? I've tested this as much as I can by inspection, on both APCS and AAPCS > binaries, but another pair of eyes would be appreciated. Revised versions > of two of the csl-arm-branch patches will use this machinery. Hi Richard, Have you had a chance to look at this? -- Daniel Jacobowitz CodeSourcery, LLC