Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 12 Apr 2003 18:26:36 -0700
From:      "David O'Brien" <obrien@freebsd.org>
To:        Loren James Rittle <rittle@latour.rsch.comm.mot.com>
Cc:        gmarco@giovannelli.it
Subject:   Re: gcc iussue or ... ?
Message-ID:  <20030413012636.GA76030@dragon.nuxi.com>
In-Reply-To: <200304120241.h3C2fQCc061882@latour.rsch.comm.mot.com>
References:  <5.2.0.9.2.20030411082040.02604e90@194.184.65.4> <200304120241.h3C2fQCc061882@latour.rsch.comm.mot.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Apr 11, 2003 at 09:41:26PM -0500, Loren James Rittle wrote:
> In article <5.2.0.9.2.20030411082040.02604e90@194.184.65.4> you write:
> > The problem is very simple.  I mantain a unix/linux port of a UO
> > emulator which works really fine natively under FreeBSD.
> 
> Thank you for this report.  Please consider this question: What
> versions of g++/binutils were used on the various platforms?  Please
> know this: There is a per-platform and per-binutils version influence
> on mapping of program data/BSS to ELF section.  g++ 3.3 with recent
> binutils on FreeBSD *will* allocate data as you saw on debian 3.0
> (with a compiler version you didn't report).
> 
> Created with g++ 3.3 (prerelease) with binutils 2.13.2.1:
> 
> ; size lonewolf
>    text    data     bss     dec     hex filename
> 1792215   49768 21850592        23692575        169851f lonewolf
> ; ls -l lonewolf
> -rwxr-xr-x  1 rittle  wheel  1992568 Apr 11 20:18 lonewolf*

Hi Loren,

Are you setup to easily try this with your g++ 3.3 binary + stock system
binutils?  It looks like this is due to the fact BSD's don't define
either ASM_OUTPUT_BSS or ASM_OUTPUT_ALIGNED_BSS.  From cp/decl.c:

  #if ! defined (ASM_OUTPUT_BSS) && ! defined (ASM_OUTPUT_ALIGNED_BSS)
    /* Tell the back-end to use or not use .common as appropriate.  If we say
       -fconserve-space, we want this to save .data space, at the expense of
       wrong semantics.  If we say -fno-conserve-space, we want this to
       produce errors about redefs; to do this we force variables into the
       data segment.  */
    DECL_COMMON (tem) = flag_conserve_space || ! TREE_PUBLIC (tem);
  #endif

I'm not sure we should change FreeBSD to do something that purposfully
produces wrong semantics.  From your test it sounds like either GCC 3.3
or Binutils 2.13.2.1 changes things for us.  I haven't updated FreeBSD to
2.13.2.1 as the changes from 2.13.2 don't really change any of the
FreeBSD architectures.  The behavior change is probably due to improved
g++ utilization of ELF features, or your FreeBSD-related bug fixes in
3.3.

-- 
-- David  (obrien@FreeBSD.org)


Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030413012636.GA76030>