On 8/20/21 5:43 PM, Ani Sinha wrote: > On Thu, 19 Aug 2021, Philippe Mathieu-Daudé wrote: >> On 8/12/21 9:14 AM, Ani Sinha wrote: > >>> + return; >> >> I suppose if you replace all 'return' by 'g_assert_not_reached()' >> both issues reproducers crash? >> >> Your patch is not incorrect, and indeed fixes the issues, but >> I feel we are going backward now allowing call which should >> never be there in the first place. >> > > Linux kernel does something like this all over the place. They simply > replace functions with NOOPS when they are not allowed for a > configuration. They do this relying on preprocessor macros ofcourse! > > It will be hard to do anythiung better without rearchitecting the modules.
Which is why this situation is unsolved since various years <:) > That would have significant impact particularly on x86. I don't believe so, the rework has to be done at the machine creation, no change during runtime.