Le 08/06/2020 à 18:06, Philippe Mathieu-Daudé a écrit :
> When compiling with GCC 10 (Fedora 32) using CFLAGS=-O2 we get:
>
> CC or1k-softmmu/hw/openrisc/openrisc_sim.o
> hw/openrisc/openrisc_sim.c: In function ‘openrisc_sim_init’:
> hw/openrisc/openrisc_sim.c:87:42: error: ‘cpu_irqs[0]’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> 87 | sysbus_connect_irq(s, i, cpu_irqs[i][irq_pin]);
> | ~~~~~~~~^~~
>
> While humans can tell smp_cpus will always be in the [1, 2] range,
> (openrisc_sim_machine_init sets mc->max_cpus = 2), the compiler
> can't.
>
> Add an assertion to give the compiler a hint there's no use of
> uninitialized data.
>
> Buglink: https://bugs.launchpad.net/qemu/+bug/1874073
> Reported-by: Martin Liška <email address hidden>
> Suggested-by: Peter Maydell <email address hidden>
> Reviewed-by: Thomas Huth <email address hidden>
> Tested-by: Eric Blake <email address hidden>
> Reviewed-by: Eric Blake <email address hidden>
> Signed-off-by: Philippe Mathieu-Daudé <email address hidden>
> ---
> v2: Fixed typo in subject (eblake)
> Supersedes: <email address hidden>
> ---
> hw/openrisc/openrisc_sim.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/hw/openrisc/openrisc_sim.c b/hw/openrisc/openrisc_sim.c
> index d08ce61811..02f5259e5e 100644
> --- a/hw/openrisc/openrisc_sim.c
> +++ b/hw/openrisc/openrisc_sim.c
> @@ -134,6 +134,7 @@ static void openrisc_sim_init(MachineState *machine)
> int n;
> unsigned int smp_cpus = machine->smp.cpus;
>
> + assert(smp_cpus >= 1 && smp_cpus <= 2);
> for (n = 0; n < smp_cpus; n++) {
> cpu = OPENRISC_CPU(cpu_create(machine->cpu_type));
> if (cpu == NULL) {
>
Le 08/06/2020 à 18:06, Philippe Mathieu-Daudé a écrit : hw/openrisc/ openrisc_ sim.o openrisc_ sim.c: In function ‘openrisc_ sim_init’ : openrisc_ sim.c:87: 42: error: ‘cpu_irqs[0]’ may be used uninitialized in this function [-Werror= maybe-uninitial ized] connect_ irq(s, i, cpu_irqs[ i][irq_ pin]); sim_machine_ init sets mc->max_cpus = 2), the compiler /bugs.launchpad .net/qemu/ +bug/1874073 openrisc_ sim.c | 1 + openrisc_ sim.c b/hw/openrisc/ openrisc_ sim.c .02f5259e5e 100644 openrisc_ sim.c openrisc_ sim.c sim_init( MachineState *machine) CPU(cpu_ create( machine- >cpu_type) );
> When compiling with GCC 10 (Fedora 32) using CFLAGS=-O2 we get:
>
> CC or1k-softmmu/
> hw/openrisc/
> hw/openrisc/
> 87 | sysbus_
> | ~~~~~~~~^~~
>
> While humans can tell smp_cpus will always be in the [1, 2] range,
> (openrisc_
> can't.
>
> Add an assertion to give the compiler a hint there's no use of
> uninitialized data.
>
> Buglink: https:/
> Reported-by: Martin Liška <email address hidden>
> Suggested-by: Peter Maydell <email address hidden>
> Reviewed-by: Thomas Huth <email address hidden>
> Tested-by: Eric Blake <email address hidden>
> Reviewed-by: Eric Blake <email address hidden>
> Signed-off-by: Philippe Mathieu-Daudé <email address hidden>
> ---
> v2: Fixed typo in subject (eblake)
> Supersedes: <email address hidden>
> ---
> hw/openrisc/
> 1 file changed, 1 insertion(+)
>
> diff --git a/hw/openrisc/
> index d08ce61811.
> --- a/hw/openrisc/
> +++ b/hw/openrisc/
> @@ -134,6 +134,7 @@ static void openrisc_
> int n;
> unsigned int smp_cpus = machine->smp.cpus;
>
> + assert(smp_cpus >= 1 && smp_cpus <= 2);
> for (n = 0; n < smp_cpus; n++) {
> cpu = OPENRISC_
> if (cpu == NULL) {
>
Applied to my trivial-patches branch.
Thanks,
Laurent