gcc-3.3.1-hammer breaks mm/memory.c

gcc-3.3.1-hammer breaks mm/memory.c

[J.A. Magallon]

Hi all...

After some binary search and a 'couple' kernel builds, I narrowed the
problem to mm/memory.c. With that file built at -O1:

CFLAGS_memory.o = -O1

The kernel boots, starts /sbin/init and looks like working normally
(2.4.22-pre5).

Anybody can see what is miscompiled with an assembler listing ?
Any #pragma to switch optmizations on the half of a file or for
a function ?

TIA

-- 
J.A. Magallon <jamagallon@able.es>      \                 Software is like sex:
werewolf.able.es                         \           It's better when it's free
Mandrake Linux release 9.2 (Cooker) for i586
Linux 2.4.22-pre5 (gcc 3.3.1 (Mandrake Linux 9.2 3.3.1-0.2mdk))

Re: gcc-3.3.1-hammer breaks mm/memory.c

[J.A. Magallon]

On 07.16, J.A. Magallon wrote:
> Hi all...
> 
> After some binary search and a 'couple' kernel builds, I narrowed the
> problem to mm/memory.c. With that file built at -O1:
> 
> CFLAGS_memory.o = -O1
> 
> The kernel boots, starts /sbin/init and looks like working normally
> (2.4.22-pre5).
> 
> Anybody can see what is miscompiled with an assembler listing ?
> Any #pragma to switch optmizations on the half of a file or for
> a function ?
> 

I finally got this: -O2 fails but

CFLAGS_memory.o = -Os -falign-functions -falign-jumps -falign-loops -falign-labels -fprefetch-loop-arrays -freorder-blocks

works.

Those are specific options disabled by -Os, but -Os does more things...
(assembler output is very different)

Any SuSE site (hammer comes from SuSE, isn't it ?) to submit a decent
bug report ?

TIA

-- 
J.A. Magallon <jamagallon@able.es>      \                 Software is like sex:
werewolf.able.es                         \           It's better when it's free
Mandrake Linux release 9.2 (Cooker) for i586
Linux 2.4.22-pre5-jam1m (gcc 3.3.1 (Mandrake Linux 9.2 3.3.1-0.2mdk))

Re: gcc-3.3.1-hammer breaks mm/memory.c

[Andrea Arcangeli]

On Wed, Jul 16, 2003 at 02:05:32AM +0200, J.A. Magallon wrote:
> 
> On 07.16, J.A. Magallon wrote:
> > Hi all...
> > 
> > After some binary search and a 'couple' kernel builds, I narrowed the
> > problem to mm/memory.c. With that file built at -O1:
> > 
> > CFLAGS_memory.o = -O1
> > 
> > The kernel boots, starts /sbin/init and looks like working normally
> > (2.4.22-pre5).
> > 
> > Anybody can see what is miscompiled with an assembler listing ?
> > Any #pragma to switch optmizations on the half of a file or for
> > a function ?
> > 
> 
> I finally got this: -O2 fails but
> 
> CFLAGS_memory.o = -Os -falign-functions -falign-jumps -falign-loops -falign-labels -fprefetch-loop-arrays -freorder-blocks
> 
> works.
> 
> Those are specific options disabled by -Os, but -Os does more things...
> (assembler output is very different)
> 
> Any SuSE site (hammer comes from SuSE, isn't it ?) to submit a decent
> bug report ?

bugs@x86-64.org should be appropriate.

Andrea