[Bug target/44060] New: miscompilation with -Os on x86

[Bug target/44060] New: miscompilation with -Os on x86

[bernhardloos at googlemail dot com]

Gcc 4.5 miscompiles the following code:

char buf[1024];
void add_input_file(int file) 
{
        (*((int **)&buf))[0] = file;
}

it results in this asm:
Disassembly of section .text:

00000000 <add_input_file>:
   0:   c3                      ret    

The file argument is completly ignored.

gcc4.4 does the right thing:
Disassembly of section .text:

00000000 <add_input_file>:
   0:   a1 00 00 00 00          mov    0x0,%eax
   5:   8b 54 24 04             mov    0x4(%esp),%edx
   9:   89 10                   mov    %edx,(%eax)
   b:   c3                      ret    

The command used to produce this output was i486-openwrt-linux-gcc -c asdf.c
-fomit-frame-pointer -Os -o asdf2.o

Openwrt adds a number of patches to gcc, they can be seen here:
https://dev.openwrt.org/browser/trunk/toolchain/gcc/patches/4.5.0
None of them should interfer with the codegeneration for x86.
The problem doesn't seem to happen with the mips target.

This is the output of gcc -v:
Using built-in specs.
COLLECT_GCC=../staging_dir/toolchain-i386_gcc-4.5.0_uClibc-0.9.30.3/usr/bin/i486-openwrt-linux-gcc
COLLECT_LTO_WRAPPER=/home/asdf/openwrt/trunk/staging_dir/toolchain-i386_gcc-4.5.0_uClibc-0.9.30.3/usr/libexec/gcc/i486-openwrt-linux-uclibc/4.5.0/lto-wrapper
Target: i486-openwrt-linux-uclibc
Configured with:
/home/asdf/openwrt/trunk/build_dir/toolchain-i386_gcc-4.5.0_uClibc-0.9.30.3/gcc-4.5.0/configure
--prefix=/home/asdf/openwrt/trunk/staging_dir/toolchain-i386_gcc-4.5.0_uClibc-0.9.30.3/usr
--build=i586-redhat-linux --host=i586-redhat-linux
--target=i486-openwrt-linux-uclibc --with-gnu-ld --enable-target-optspace
--disable-libgomp --disable-libmudflap --disable-multilib --disable-nls
--with-ppl --with-cloog --with-host-libstdcxx=-lstdc++
--with-gmp=/home/asdf/openwrt/trunk/staging_dir/host
--with-mpfr=/home/asdf/openwrt/trunk/staging_dir/host --disable-decimal-float
--disable-libssp --disable-__cxa_atexit --enable-languages=c,c++
--enable-shared --enable-threads
--with-slibdir=/home/asdf/openwrt/trunk/staging_dir/toolchain-i386_gcc-4.5.0_uClibc-0.9.30.3/lib
--enable-lto --disable-tls
Thread model: posix
gcc version 4.5.0 (GCC)


-- 
           Summary: miscompilation with -Os on x86
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bernhardloos at googlemail dot com
GCC target triplet: i486-openwrt-linux-uclibc


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060

[Bug target/44060] miscompilation with -Os on x86

[schwab at linux-m68k dot org]

------- Comment #1 from schwab at linux-m68k dot org  2010-05-10 14:43 -------
Try -Wstrict-aliasing.


-- 

schwab at linux-m68k dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060

[Bug target/44060] miscompilation with -Os on x86

[bernhardloos at googlemail dot com]

------- Comment #2 from bernhardloos at googlemail dot com  2010-05-10 14:57 -------
(In reply to comment #1)
> Try -Wstrict-aliasing.
> 

It does produce a warning about dreferencing a type-punned pointer.
I tried to compile the snipped with both -fstrict-aliasing and
-fno-strict-aliasing and it doesn't produce a difference in the code. Not on
the i486 gcc4.5 nor on the gcc4.4 in fedora.

Busybox uses this in some places and it did work in the past, so it would be
really nice, if this could get fixed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060

[Bug target/44060] miscompilation with -Os on x86

[hjl dot tools at gmail dot com]

------- Comment #3 from hjl dot tools at gmail dot com  2010-05-10 16:16 -------
Gcc 4.6 generates expected code.


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|INVALID                     |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060

[Bug target/44060] miscompilation with -Os on x86

[pluto at agmk dot net]

------- Comment #4 from pluto at agmk dot net  2010-05-10 16:26 -------
(In reply to comment #3)
> Gcc 4.6 generates expected code.
> 

PR43987 dupllication?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060

[Bug target/44060] miscompilation with -Os on x86

[bernhardloos at googlemail dot com]

------- Comment #5 from bernhardloos at googlemail dot com  2010-05-10 16:48 -------
It is a duplicate, my problem happens in exactly the same place.
Sorry I missed the other bug.

*** This bug has been marked as a duplicate of 43987 ***


-- 

bernhardloos at googlemail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |DUPLICATE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44060