Re: Default to -fstrict-volatile-bitfields for ARM EABI

[Jie Zhang <jie at codesourcery dot com>]

On 09/03/2010 12:48 AM, Joseph S. Myers wrote:
> On Fri, 3 Sep 2010, Jie Zhang wrote:
>
> > On 09/02/2010 11:55 PM, Joseph S. Myers wrote:
> > > On Thu, 2 Sep 2010, Jie Zhang wrote:
> > >
> > > > 20010518-2.c: In function 'main':
> > > > 20010518-2.c:34:21: warning: mis-aligned access used for structure member
> > > > [-fstrict-volatile-bitfields]
> > > > 20010518-2.c:34:21: note: When a volatile object spans multiple type-sized
> > > > locations, the compiler must choose between using a single mis-aligned
> > > > access
> > > > to preserve the volatility, or using multiple aligned accesses to avoid
> > > > runtime faults.  This code may fail at runtime if the hardware does not
> > > > allow
> > > > this access.
> > > >
> > > > So I think this test should be disabled for all targets which default to
> > > > -fstrict-volatile-bitfields. If this is what we should do, I will update
> > > > my
> > > > patch to disable it for all such targets.
> > >
> > > Why should it be disabled?  It should still execute OK; that's just a
> > > warning, and warnings aren't meant to cause c-torture tests to fail.
> > But GCC chooses to preserve the volatility in this case. It tries to use "ldr"
> > to access unaligned address. This test fails at runtime on arm-none-eabi
> > target at any optimization levels.
>
> That sounds like a bug in this option.  The test looks like perfectly
> valid GNU C to me and should not fail depending on the target.
Hmmm, there are no bitfields in the test case, but 
-fstrict-volatile-bitfields affected it. I will take a look tomorrow.


--
Jie Zhang
CodeSourcery