On Thu, Feb 27, 2025 at 12:42:30AM +0000, Zhijian Li (Fujitsu) wrote:
>
>
> On 26/02/2025 23:51, Peter Xu wrote:
> > On Wed, Feb 26, 2025 at 02:30:42PM +0800, Li Zhijian wrote:
> >> control_save_page() is for RDMA only, unfold it to make the code more
> >> clear.
> >> In addition:
> >> - Similar to other branches style in ram_save_target_page(), involve RDMA
> >> only if the condition 'migrate_rdma()' is true.
> >> - Further simplify the code by removing the RAM_SAVE_CONTROL_NOT_SUPP.
> >>
> >> Signed-off-by: Li Zhijian <lizhij...@fujitsu.com>
> >
> > [...]
> >
> >> @@ -56,7 +55,7 @@ static inline
> >> int rdma_control_save_page(QEMUFile *f, ram_addr_t block_offset,
> >> ram_addr_t offset, size_t size)
> >> {
> >> - return RAM_SAVE_CONTROL_NOT_SUPP;
> >> + g_assert_not_reached();
> >> }
> >
> > Not sure if some compiler will be unhappy on the retval not provided, but
> > anyway we'll see..
>
> There is no problem in fedora 40(gcc 14.2.1) and ubuntu2204(gcc 11.4.0) with
> --disable-rdma.
>
> I also noticed we have a few existing same usage:
>
> 1708 bool ram_write_tracking_compatible(void)
> 1709 {
> 1710 g_assert_not_reached();
> 1711 }
> 1712
> 1713 int ram_write_tracking_start(void)
> 1714 {
> 1715 g_assert_not_reached();
> 1716 }
> 1717
> 1718 void ram_write_tracking_stop(void)
> 1719 {
> 1720 g_assert_not_reached();
> 1721 }
Right.
The other question is what about G_DISABLE_ASSERT, then I found this:
osdep.h:
#ifdef G_DISABLE_ASSERT
#error building with G_DISABLE_ASSERT is not supported
#endif
So yeah, we should be good.
>
>
> I also asked the AI/Deepseek-R1, pasted a piece of his answer
>
> ```
> 3. Why No Warning for Missing return? 🚨
> Typical case: A non-void function missing a return triggers -Wreturn-type
> warnings (enabled by -Wall).
> This case: The noreturn annotation ensures no execution path exists beyond
> g_assert_not_reached(). Because the compiler recognizes this, no warning is
> necessary.
>
> Conclusion
> GCC trusts the noreturn annotation of g_assert_not_reached(), recognizing
> that the function’s control flow ends there. Thus, no warning is emitted. For
> code safety, ensure assertions are active or add fallback code if needed.
> ```
>
> >
> > Reviewed-by: Peter Xu <pet...@redhat.com>
> >
> >> #endif
> >> #endif
> >> --
> >> 2.44.0
> >>
> >
--
Peter Xu
