X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
Message-ID: <45F41F43.7020403@iki.fi>
Date: Sun, 11 Mar 2007 17:24:51 +0200
From: Andris Pavenis <andris DOT pavenis AT iki DOT fi>
User-Agent: Thunderbird 2.0b2 (X11/20070116)
MIME-Version: 1.0
To: djgpp AT delorie DOT com
Subject: Re: BSS problem in DJGPP port of GCC 4 (was: Re: Question about djcross-gcc-4.1.1-2)
References: <BAY119-F2182C875A1CFCF1E182309E3AD0 AT phx DOT gbl>
In-Reply-To: <BAY119-F2182C875A1CFCF1E182309E3AD0@phx.gbl>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Reply-To: djgpp AT delorie DOT com

Pablo Fujii kirjoitti:
>> It seems to be bug in binutils. It does not appear even in C++ when 
>> option -fno-exceptions is specified (so no .eh_frame section).
> 
> Actually, the -fno-exceptions causes files like eh_alloc.o not to be 
> linked into the resulting executable. The stdc++ library is compiled 
> with several compiler flags, one of them being -fdata-sections.
> 
> Consider the following small test program:
> 
> --- cut begin---
> static int int_array[500000];
> 
> int main (void) {
>    return 0;
> }
> --- cut end---
> 
> Compiling this C file *without* the -fdata-sections option gives the 
> desired output:
> 
> $ i586-pc-msdosdjgpp-size bss_test.exe
>   text    data     bss     dec     hex filename
>  44488    3072 2014208 2061768  1f75c8 bss_test.exe
> $ i586-pc-msdosdjgpp-objdump -h bss_test.o
> 
> bss_test.o:     file format coff-go32
> 
> Sections:
> Idx Name          Size      VMA       LMA       File off  Algn
>  0 .text         00000020  00000000  00000000  000000b4  2**4
>                  CONTENTS, ALLOC, LOAD, CODE
>  1 .data         00000000  00000000  00000000  00000000  2**4
>                  ALLOC, LOAD, DATA
>  2 .bss          001e8480  00000000  00000000  00000000  2**2
>                  ALLOC
>  3 .comment      00000014  00000000  00000000  000000d4  2**2
>                  CONTENTS, DEBUGGING
> 
> Compiling this C file *with* the -fdata-sections option gives "wrong" 
> output:
> 
> $ i586-pc-msdosdjgpp-size bss_test.exe
>   text    data     bss     dec     hex filename
>  44488 2017280       0 2061768  1f75c8 bss_test.exe
> $ i586-pc-msdosdjgpp-objdump -h bss_test.o
> 
> bss_test.o:     file format coff-go32
> 
> Sections:
> Idx Name          Size      VMA       LMA       File off  Algn
>  0 .text         00000020  00000000  00000000  000000dc  2**4
>                  CONTENTS, ALLOC, LOAD, CODE
>  1 .data         00000000  00000000  00000000  00000000  2**4
>                  ALLOC, LOAD, DATA
>  2 .bss          00000000  00000000  00000000  00000000  2**2
>                  ALLOC
>  3 .bss.int_array 001e8480  00000000  00000000  000000fc  2**2
>                  CONTENTS, ALLOC, LOAD, DATA
>  4 .comment      00000014  00000000  00000000  001e857c  2**2
>                  CONTENTS, DEBUGGING
> 
> As indicated before, the native GCC 4.1.1 compiler of e.g. Fedora Core 6 
> does not have this problem:
> 
> $ objdump -h bss_test.o
> 
> bss_test.o:     file format elf32-i386
> 
> Sections:
> Idx Name          Size      VMA       LMA       File off  Algn
>  0 .text         00000019  00000000  00000000  00000034  2**2
>                  CONTENTS, ALLOC, LOAD, READONLY, CODE
>  1 .data         00000000  00000000  00000000  00000050  2**2
>                  CONTENTS, ALLOC, LOAD, DATA
>  2 .bss          00000000  00000000  00000000  00000050  2**2
>                  ALLOC
>  3 .bss.int_array 001e8480  00000000  00000000  00000060  2**5
>                  ALLOC
>  4 .comment      0000002e  00000000  00000000  00000060  2**0
>                  CONTENTS, READONLY
>  5 .note.GNU-stack 00000000  00000000  00000000  0000008e  2**0
>                  CONTENTS, READONLY
> 
> 
> Comparison of the generated assembler files, first the one by created by 
> DJGPP GCC 4.1.0:
> 
>     .file    "bss_test.c"
>     .section .text
> .globl _main
> _main:
>     leal    4(%esp), %ecx
>     ...
>     ret
>     .section    .bss.int_array,"w"

I tried to edit .section directive to add flag 'b' (.bss), but it does 
not seem to help. One possible ugly workaround for this problem could be 
moving .bss.* to .data section by modifying linker script.