> Before I send the fix, I did a quick test on linux as following:
>
> #include <stdio.h>
> #include <stdbool.h>
>
> int main()
> {
> printf("bool size = %d\n", sizeof(bool));
> }
>
>
> then:
> ./a.out
> bool size = 1
>
> and there is mptsas.c #include "qemu/osdep.h", osdep.h #include <stdbool.h>
>
> So, am I missing something?
You are missing that the patch is: 1) not portable, because there's no guarantee
that sizeof(bool)==1; 2) wrong, because using sizeof within VMSTATE_UNUSED would
suggest a dependency of the migration stream on the host---and such a dependency
should not be there at all.
In fact, the patch is also wrong because old versions of QEMU _do_ need
msi_in_use
so you need to either generate it or bump the migration version. I'm sending a
patch now.
Paolo
