From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cagney To: Daniel Jacobowitz Cc: gdb-patches@sources.redhat.com, Joel Brobecker Subject: Re: [rfc] Fixes for sim and gdb gdb_mbuild.sh failures Date: Wed, 12 May 2004 20:05:00 -0000 Message-id: <40A283A0.8050801@gnu.org> References: <20040511022048.GA2172@nevyn.them.org> <40A0F14A.6090609@gnu.org> <20040511173249.GB19487@nevyn.them.org> X-SW-Source: 2004-05/msg00379.html On Tue, May 11, 2004 at 11:29:14AM -0400, Andrew Cagney wrote: >The others were all warnings. Two are dead fix_call_dummy's, one is a >printf format string warning (sizeof is size_t, which is not necessarily >long), and three are long 64-bit constants without suffixes. The last I'm >least sure about. ia64-tdep.c already used the LL suffix, but alpha-tdep.c >and amd64-tdep.c didn't. I'm slightly worried that a native Alpha (Compaq) >compiler will complain about the LL syntax (since CORE_ADDR may be just a >long in that case, and I remember Compaq's compiler as being remarkably >pedantic). Joel (or someone else), I don't suppose you could test this >patch by building GDB on OSF/1? There must be something in src/include to portably handle `LL' - what does bfd do? It isn't ISO-C '90. Only elfxx-ia64.c and opcode/d30v.h (and related files in opcodes/) use long long constants; everything else seems to avoid needing them, as far as I can tell. I could decompose most of them into shifts and casts to CORE_ADDR instead if you think it's worth it. Let binutils know they should delete the d30v code. I guess we live with LL on 64-bit platforms. Andrew