-fschedule-insns problem with arm-linux-uclibc g++

[Manuel Klimek]

Hi,

I found something strange I don't understand, but I
don't know if it's really a bug.

If I compile the following simple file x.cc:
class A
{
public:
  A();
  ~A();
  int a();
};
class B
{
public:
  static int b();
};

int B::b()
{
    A a;
    return a.a();
}

I get the following assembler output for function B::b:
with
 arm-linux-g++ -o x -g -O1 -fno-schedule-insns -fPIC -c x.cc
int B::b()
   0:   e1a0c00d        mov     ip, sp
   4:   e92ddff0        stmdb   sp!, {r4, r5, r6, r7, r8, r9, sl, fp,
ip, lr, pc
}
   8:   ed2d420c        sfm     f4, 4, [sp, #-48]!
   c:   e24cb004        sub     fp, ip, #4      ; 0x4
  10:   e24dd044        sub     sp, sp, #68     ; 0x44
  14:   e59f30bc        ldr     r3, [pc, #188]  ; d8 <.text+0xd8>
  18:   e7933003        ldr     r3, [r3, r3]
  1c:   e50b3078        str     r3, [fp, #-120]
  20:   e59f30b4        ldr     r3, [pc, #180]  ; dc <.text+0xdc>
  24:   e0833003        add     r3, r3, r3
  28:   e50b3074        str     r3, [fp, #-116]
  2c:   e24b3058        sub     r3, fp, #88     ; 0x58
  30:   e50b3070        str     r3, [fp, #-112]
  34:   e59f30a4        ldr     r3, [pc, #164]  ; e0 <.text+0xe0>
  38:   e0833003        add     r3, r3, r3
  3c:   e50b306c        str     r3, [fp, #-108]
  40:   e50bd068        str     sp, [fp, #-104]
  44:   e24b0090        sub     r0, fp, #144    ; 0x90
  48:   ebfffffe        bl      0 <_Unwind_SjLj_Register>


with
 arm-linux-g++ -o x -g -O1 -fschedule-insns -fPIC -c x.cc:
int B::b()
   0:   e1a0c00d        mov     ip, sp
   4:   e92ddff0        stmdb   sp!, {r4, r5, r6, r7, r8, r9, sl, fp,
ip, lr, pc
}
   8:   ed2d420c        sfm     f4, 4, [sp, #-48]!
   c:   e24cb004        sub     fp, ip, #4      ; 0x4
  10:   e24dd044        sub     sp, sp, #68     ; 0x44
  14:   e59f30bc        ldr     r3, [pc, #188]  ; d8 <.text+0xd8>
  18:   e7921003        ldr     r1, [r2, r3]
  1c:   e59f20b8        ldr     r2, [pc, #184]  ; dc <.text+0xdc>
  20:   e0822002        add     r2, r2, r2
  24:   e59f30b4        ldr     r3, [pc, #180]  ; e0 <.text+0xe0>
  28:   e0823003        add     r3, r2, r3
  2c:   e50b1078        str     r1, [fp, #-120]
  30:   e50b3074        str     r3, [fp, #-116]
  34:   e50b206c        str     r2, [fp, #-108]
  38:   e24b3058        sub     r3, fp, #88     ; 0x58
  3c:   e50b3070        str     r3, [fp, #-112]
  40:   e50bd068        str     sp, [fp, #-104]
  44:   e24b0090        sub     r0, fp, #144    ; 0x90
  48:   ebfffffe        bl      0 <_Unwind_SjLj_Register>

Now I get a segfault in
  18:   e7921003        ldr     r1, [r2, r3]

when I compare this to the -fno-schedule-insns case I think that
this line should read:
  18:   e7921003        ldr     r1, [r3, r3]

which would yield the same result as with -fno-schedule-insns.

Since I'm no scheduling expert, I don't know if this is a bug
or just some arm weirdness.

some info:
arm-linux-uclibc-g++ -v
Using built-in specs.
Target: arm-linux-uclibc
Configured with:
/home/klimek/stuff/toolchains/st4000-gcc4.2/build/toolchain/buildroot2/toolchain_build_arm/gcc-4.2-20061007/configure
--prefix=/home/klimek/stuff/toolchains/st4000-gcc4.2/toolchain
--build=i386-pc-linux-gnu --host=i386-pc-linux-gnu
--target=arm-linux-uclibc --enable-languages=c,c++ --enable-shared
--disable-__cxa_atexit --enable-target-optspace --with-gnu-ld
--disable-nls --enable-threads --disable-multilib --disable-libstdcxx-pch
: (reconfigured)
/home/klimek/stuff/toolchains/st4000-gcc4.2/build/toolchain/buildroot2/toolchain_build_arm/gcc-4.2-20061007/configure
--prefix=/home/klimek/stuff/toolchains/st4000-gcc4.2/toolchain
--build=i386-pc-linux-gnu --host=i386-pc-linux-gnu
--target=arm-linux-uclibc --enable-languages=c,c++ --enable-shared
--disable-__cxa_atexit --enable-target-optspace --with-gnu-ld
--disable-nls --enable-threads --disable-multilib --disable-libstdcxx-pch
Thread model: posix
gcc version 4.2.0 20061007 (experimental)

arm-linux-uclibc-g++ -dumpmachine
arm-linux-uclibc

Cheers,
Manuel