Re: [PATCH 19/39] hw/pci: replace assert(false) with g_assert_not_reached()

[Pierrick Bouvier]

On 9/10/24 22:50, Philippe Mathieu-Daudé wrote:
> Hi Pierrick,
>
> On 11/9/24 00:15, Pierrick Bouvier wrote:
> > Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
> > ---
> >    hw/pci/pci-stub.c | 4 ++--
> >    1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/hw/pci/pci-stub.c b/hw/pci/pci-stub.c
> > index f0508682d2b..c6950e21bd4 100644
> > --- a/hw/pci/pci-stub.c
> > +++ b/hw/pci/pci-stub.c
> > @@ -46,13 +46,13 @@ void hmp_pcie_aer_inject_error(Monitor *mon, const QDict 
> > *qdict)
> >    /* kvm-all wants this */
> >    MSIMessage pci_get_msi_message(PCIDevice *dev, int vector)
> >    {
> > -    g_assert(false);
> > +    g_assert_not_reached();
> >        return (MSIMessage){};
>
> The tail of this series remove the unreachable 'break' lines.
> Why 'return' lines aren't problematic? Is that a GCC TSan bug?

It's related to how control flow analysis works, but I don't have a 
deeper answer. I reported the issue with 'break' for gcc and the same 
bug was created several years ago, so it was just marked as duplicate.

I'll clean the extra return with have though, as part of v2.

> >    }
> >    
> >    uint16_t pci_requester_id(PCIDevice *dev)
> >    {
> > -    g_assert(false);
> > +    g_assert_not_reached();
> >        return 0;
> >    }
> >