Re: [PATCH] memory: Use QEMU_ALWAYS_INLINE to silence compile-time warning

[Zenghui Yu]

Hi Richard,

On 2020/2/5 18:31, Richard Henderson wrote:
> On 2/5/20 10:01 AM, Richard Henderson wrote:
> > On 2/5/20 8:17 AM, Zenghui Yu wrote:
> > > Our robot reported the following compile-time warning while compiling
> > > Qemu with -fno-inline cflags:
> >
> > Why are you doing this?

I'm not sure why it was added in the building test. As you said,
there's just a arbitrary combination of compile flags.

> > > In function 'load_memop',
> > >      inlined from 'load_helper' at /qemu/accel/tcg/cputlb.c:1578:20,
> > >      inlined from 'full_ldub_mmu' at /qemu/accel/tcg/cputlb.c:1624:12:
> > > /qemu/accel/tcg/cputlb.c:1502:9: error: call to 'qemu_build_not_reached' 
> > > declared with attribute error: code path is reachable
> > >           qemu_build_not_reached();
> > >           ^~~~~~~~~~~~~~~~~~~~~~~~
> > >      [...]
> >
> > Of course, the assert is compiled out when optimization is off, which is the
> > only time we expect inlining to be off.
> >
> > The patch isn't wrong, exactly, but I question whether we want to support
> > completely arbitrary combinations of compile flags.

It doesn't hurt to do the right thing, and you already have the
good approach :)

> To follow up: if you *really* need to support -fno-inline, then perhaps the
> correct patch is to change
>
> - #ifdef __OPTIMIZE__
> + #if defined(__OPTIMIZE__) && !defined(__NO_INLINE__)
>    extern void QEMU_NORETURN QEMU_ERROR("code path is reachable")
>        qemu_build_not_reached(void);
>    #else
>    #define qemu_build_not_reached()  g_assert_not_reached()
>    #endif
>
> within include/qemu/compiler.h.

Thanks for this. I've tested it with -fno-inlie and it indeed works.
I will send it as v2.


Thanks,
Zenghui