Re: dangers of current NEED_CPU_H, CONFIG_SOFTMMU, CONFIG_USER_ONLY (was: [PATCH v11 7/7] cpu: introduce cpu_accel_instance_init)

[Peter Maydell]

On Thu, 17 Dec 2020 at 20:15, Peter Maydell <peter.maydell@linaro.org> wrote:
> (So in theory we could make CONFIG_USER_ONLY
> a poisoned identifier but that will require some work to
> adjust places where we currently use it in "safe" ways...)

Specifically, putting it in poison.h turns up these places
that would need to be made to do what they're doing in a
different way somehow:

../../hw/core/cpu.c:211:14: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/disas/disas.h:27:13: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/exec/address-spaces.h:24:9: error: attempt to use poisoned
"CONFIG_USER_ONLY"
include/exec/cpu-common.h:20:14: error: attempt to use poisoned
"CONFIG_USER_ONLY"
include/exec/cpu-common.h:6:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/exec/ioport.h:43:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/exec/memory.h:17:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/exec/ramblock.h:22:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/hw/core/cpu.h:1035:8: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/hw/core/cpu.h:518:14: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/hw/core/cpu.h:602:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/hw/hw.h:4:8: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/hw/semihosting/semihost.h:29:8: error: attempt to use poisoned
"CONFIG_USER_ONLY"
include/sysemu/accel.h:40:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/sysemu/cpus.h:65:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/sysemu/dma.h:34:9: error: attempt to use poisoned "CONFIG_USER_ONLY"
include/sysemu/xen.h:27:9: error: attempt to use poisoned "CONFIG_USER_ONLY"

That cpu.c one is definitely dubious given it's in a C file,
not a header.

thanks
-- PMM