GCC 4.6.x miscompiling arm-linux?

[David Jander]

On Tue, 11 Sep 2012 12:29:05 +0200
Michael Olbrich <mol at pengutronix.de> wrote:

> On Tue, Sep 11, 2012 at 11:31:39AM +0200, David Jander wrote:
> > On Tue, 11 Sep 2012 10:48:53 +0200
> > Sascha Hauer <s.hauer at pengutronix.de> wrote:
> > 
> > > On Tue, Sep 11, 2012 at 09:27:53AM +0200, David Jander wrote:
> > > > 
> > > > Hi Matt,
> > > > 
> > > > >         @ Function supports interworking.
> > > > >         @ args = 0, pretend = 0, frame = 0
> > > > >         @ frame_needed = 0, uses_anonymous_args = 0
> > > > >         @ link register save eliminated.
> > > > >         cmp     r0, #9
> > > > >         mov     r3, #0
> > > > >         str     r3, [r1, #0]
> > > > >         mov     r0, #0
> > > > >         strgt   r3, [r1, #4]
> > > > >         bx      lr
> > > > >         .size   flexcan_chip_start, .-flexcan_chip_start
> > > > >         .ident  "GCC: (GNU) 4.3.3"
> > > > >         .section        .note.GNU-stack,"",%progbits
> > > > > 
> > > > > I think this looks correct. Perhaps you could try the angstrom arm5te
> > > > > toolchain and see if it's a toolchain issue?
> > > > 
> > > > Yes, this looks a lot better, and is exactly what I get when I compile this
> > > > code with CodeSourcery GCC-4.4.1
> > > > 
> > > > I have tries building gcc-4.6.3 also with OSELAS/PTXdist, and it gives the
> > > > same (wrong) result as with gcc-4.6.2
> > > 
> > > AFAIK our OSELAS/PTXdist Toolchains use Linaro patches, this would
> > > explain why both have the same result.
> 
> No, OSELAS Toolchains only uses the Linaro GCC for Cortex-A8/-A9
> 
> > It would indeed, if I had just used OSELAS as-is, but I didn't ;-)
> > I made a copy of
> > arm-v5te-linux-gnueabi_gcc-4.6.2_glibc-2.14.1_binutils-2.21.1a_kernel-2.6.39-sanitized.ptxconfig
> > and replaced the version string 4.6.2 with 4.6.3. AFAIK, this way PTXdist will
> > download pristine gcc-4.6.3 sources and not apply any patches, since there is
> > no directory named patches/gcc-4.6.3, right? Can I assume that I have an
> > unpatched version of gcc-4.6.3 then?
> 
> Correct. This way you'll get gcc-4.6.3 without any patches.
> 
> > Btw, Here's what I get from it:
> > 
> > flexcan_chip_start:
> >         @ args = 0, pretend = 0, frame = 0
> >         @ frame_needed = 0, uses_anonymous_args = 0
> >         @ link register save eliminated.
> >         mov     r3, #0
> >         cmp     r0, #9
> >         str     r3, [r1, #0]
> >         ldrle   r3, [r1, #4]
> >         mov     r0, #0
> >         str     r3, [r1, #4]
> >         bx      lr
> >         .size   flexcan_chip_start, .-flexcan_chip_start
> >         .ident  "GCC: (OSELAS.Toolchain-2011.11.1) 4.6.3"
> >         .section        .note.GNU-stack,"",%progbits
> 
> How can I reproduce this output? I'd like to test this with different
> toolchain versions.

I basically do this:

$ /opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/bin/arm-v5te-linux-gnueabi-gcc-4.6.3 -Os -S -o - -x c -

struct flexcan_regs {
        unsigned int mcr;
        unsigned int rxfgmask;
};

#define flexcan_read(a) (*(volatile unsigned int *)(a))
#define flexcan_write(v,a)      (*(volatile unsigned int *)(a) = (v))

int flexcan_chip_start(int ver, struct flexcan_regs *regs)
{
        flexcan_write(0, &regs->mcr);

        if (ver >= 10)
                flexcan_write(0, &regs->rxfgmask);

        return 0;
}

> > Wrong code.
> > See the version string? It is clearly not an existing OSELAS version.
> 
> Hmmm, so far it seems all tested gcc-4.6.x versions produce wrong code,
> correct?

Looks like that, yes.

Best regards,

-- 
David Jander
Protonic Holland.