From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 10080 invoked by alias); 4 Dec 2003 22:45:49 -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 10049 invoked from network); 4 Dec 2003 22:45:48 -0000 Received: from unknown (HELO zenia.home) (12.223.225.216) by sources.redhat.com with SMTP; 4 Dec 2003 22:45:48 -0000 Received: by zenia.home (Postfix, from userid 5433) id 5F3EF20766; Thu, 4 Dec 2003 17:44:13 -0500 (EST) To: Ulrich Weigand Cc: gdb-patches@sources.redhat.com, uweigand@de.ibm.com Subject: Re: [PATCH] S/390 port modernization 4/4 References: <200312042008.VAA07725@faui1d.informatik.uni-erlangen.de> From: Jim Blandy Date: Thu, 04 Dec 2003 22:45:00 -0000 In-Reply-To: <200312042008.VAA07725@faui1d.informatik.uni-erlangen.de> Message-ID: User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-SW-Source: 2003-12/txt/msg00153.txt.bz2 Ulrich Weigand writes: > this is patch 4/4 to update the s390 backend. > > The focus of this patch is frame handling; it completely removes > all old-style frame routines and switches the backend to use the > new frame handling using frame_unwind and frame_base. > > This is still using prologue analysis; in fact the analyser is > made somewhat more generic: it now understands the new long > displacement instructions introduced with z990, and it is also > more flexible in that it can handle unknown instructions > intermixed with the prologue due to instruction scheduling. > Also, literal pool constants are now accepted in 64-bit mode > as well (not just 31-bit mode). > > This patch removes the last (to my knowledge) use of deprecated > features in the s390 backend. > > > Tested on s390-ibm-linux and s390x-ibm-linux with no new regressions. > Fixes (about) 23 unexpected and 4 known failures on s390; > fixes (about) 26 unexpected and 1 known failures on s390x. This looks great. For another target, I made the same optimistic assumption you did about unrecognized instructions, and it worked well. I did run into some problems with the analysis going too far, but those were all exposed by the test suite, so you've probably found everything I did.