From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 28197 invoked by alias); 8 Oct 2003 07:40:31 -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 28135 invoked from network); 8 Oct 2003 07:40:28 -0000 Received: from unknown (HELO u1sparc.j-son.org) (213.64.105.80) by sources.redhat.com with SMTP; 8 Oct 2003 07:40:28 -0000 Received: (from chj@localhost) by u1sparc.j-son.org (8.11.6/8.11.6) id h987eIq28635; Wed, 8 Oct 2003 09:40:18 +0200 Date: Wed, 08 Oct 2003 07:40:00 -0000 From: Christian Joensson To: Nix Cc: "Tom 'spot' Callaway" , gdb-gnats@sources.redhat.com, gdb-patches@sources.redhat.com Subject: Re: build/955: build failure with GDB-5.3: sparc-nat.c structure redefinition errors with sparc64-linux, glibc-2.2.x Message-ID: <20031008094017.A28615@u1sparc.j-son.org> References: <20030427150800.19859.qmail@sources.redhat.com> <20030427153321.GA3394@nevyn.them.org> <20030510095712.A19164@u1sparc.j-son.org> <20030510153527.A5854@u1sparc.j-son.org> <20030825085704.A14976@u1sparc.j-son.org> <87wubggpby.fsf@amaterasu.srvr.nix> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: <87wubggpby.fsf@amaterasu.srvr.nix>; from nix@esperi.org.uk on Wed, Oct 08, 2003 at 08:29:05AM +0100 X-SW-Source: 2003-10/txt/msg00199.txt.bz2 On Wed, Oct 08, 2003 at 08:29:05AM +0100, Nix wrote: > On Mon, 25 Aug 2003, Christian Joensson uttered the following: > > uhm, just tried gdb cvs HEAD, as of Sun Aug 24 12:01:38 UTC 2003, same thing: > > I'm back, at last. > > > In file included from /usr/include/asm/reg.h:7, > > from /home/chj/src/gdb/sparc-nat.c:38: > > /usr/include/asm-sparc64/reg.h:49: error: redefinition of `struct fpu' > > sparc64? Are you trying to build a 64-bit gdb? uhm, yes, to debug 64-bit binaries.... > This is *not* related to the earlier bug in sparc-nat.c / the Linux > kernel headers / the glibc headers that the earlier patch from Daniel > Jacobowitz was for. That patch works (Debian uses it). ok > Using GDB-6.0-release, I now see > > gcc -c -O2 -mcpu=ultrasparc -mtune=ultrasparc -m32 -g -pipe -D__NO_STRING_INLINES -D__NO_MATH_INLINES -D_FILE_OFFSET_BITS=64 -I. -I. -I./config -DLOCALEDIR="\"/usr/share/locale\"" -DHAVE_CONFIG_H -I./../include/opcode -I./../readline/.. -I../bfd -I./../bfd -I./../include -I../intl -I./../intl -DMI_OUT=1 -DTUI=1 -I./tui -Wimplicit -Wreturn-type -Wcomment -Wtrigraphs -Wformat -Wparentheses -Wpointer-arith -Wuninitialized sparc-nat.c > sparc-nat.c: In function `fetch_inferior_registers': > sparc-nat.c:115: error: structure has no member named `r_ps' > sparc-nat.c:117: error: structure has no member named `r_pc' > sparc-nat.c:119: error: structure has no member named `r_npc' > sparc-nat.c: In function `store_inferior_registers': > sparc-nat.c:266: error: structure has no member named `r_ps' > sparc-nat.c:268: error: structure has no member named `r_pc' > sparc-nat.c:270: error: structure has no member named `r_npc' > sparc-nat.c: In function `fetch_core_registers': > sparc-nat.c:326: error: structure has no member named `r_ps' > sparc-nat.c:327: error: structure has no member named `r_pc' > sparc-nat.c:328: error: structure has no member named `r_npc' > make[1]: *** [sparc-nat.o] Error 1 > > building on an UltraSPARC via the sparc32 personality. ok, but I try to build, using 64-bit versions of binutils, gcc, and ncurses, a 64-bit one... > ... further investigation shows that at some time in the distant past, > something, somehow replaced the sparc32 reg.h in my copy of the kernel > sources with the sparc64 one. (Probably it was me in an earlier > debugging run, by accident.) Because my kernel is sparc64 and nothing > much other than GDB uses this header in userspace, I'd not noticed. > > Putting the sparc32 reg.h back, combined with code to use the right > kernel headers for the target (i.e. sparc32) made everything work. > > If your bug has similar causes to mine --- at least one sparc64 header > being picked up instead of a sparc32 one, in a 32-bit userspace --- I'd > call this user error. I'd also call the sparc64 reg.h header `not ready > for gdb' since symbols present since at least 1989 which gdb relies on > are not there. I see, well, maybe that will be addressed sometime > > (It probably makes sense to replace the kernel headers in > /usr/include/asm with something like the multi-personality > jiggery-pokery normally used to build glibc, i.e. autogenerated headers > looking like > > #ifndef __MULTIUNIVERSE__REG_H__ > #define __MULTIUNIVERSE__REG_H__ > > #ifdef __arch64__ > #include "sparc64/reg.h" > #else > #include "sparc/reg.h" > #endif > > #endif /* !__MULTIUNIVERSE__REG_H__ */ > > and all the sparc64 and sparc32 headers in the appropriate > subdirectories. > > If you've not done that, you could also find yourself picking up sparc64 > headers instead of sparc32 ones, and getting bitten by this. > > If you're using a Linux distro of some kind, I'd imagine they've already > done this.) uhm, a 64-bit gdb is not that easy to accomplish, I guess, but there'll be one in a future Aurora SPARC Linux, a development gdb64-5.2-ish src rpm is available: ftp://auroralinux.org/pub/aurora/scratch Cheers, /ChJ