On 03/02/2025 17.38, Richard Henderson wrote:
> On 2/3/25 02:08, Thomas Huth wrote:
>> On 03/02/2025 04.18, Richard Henderson wrote:
>>> Use CONFIG_TCG as a project-wide flag to indicate that TCG is enabled
>>> for *some* target. Use CONFIG_TCG_TARGET to indicate that TCG is
>>> enabled for a specific target.
>>>
>>> Within a specific compilation unit, we can remap CONFIG_TCG based on
>>> CONFIG_TCG_TARGET. This allows us to avoid changes to the bulk of
>>> the code base.
>>>
>>> Within meson.build, while CONFIG_TCG may be set in config_host_data,
>>> it may not be set within config_target. Thus all references to
>>> CONFIG_TCG in source_set 'when:' need not be updated.
>>>
>>> For the moment, CONFIG_TCG and CONFIG_TCG_TARGET are identical.
>>>
>>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>>> ---
>>> include/qemu/osdep.h | 7 +++++++
>>> meson.build | 11 +++++++----
>>> 2 files changed, 14 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
>>> index 112ebdff21..1f6f73a148 100644
>>> --- a/include/qemu/osdep.h
>>> +++ b/include/qemu/osdep.h
>>> @@ -34,9 +34,16 @@
>>> #include "config-host.h"
>>> #ifdef COMPILING_PER_TARGET
>>> #include CONFIG_TARGET
>>> +# ifdef CONFIG_TCG_TARGET
>>> +# undef CONFIG_TCG_TARGET
>>> +# else
>>> +# undef CONFIG_TCG
>>> +# endif
>>> #else
>>> #include "exec/poison.h"
>>> #endif
>>> +#pragma GCC poison CONFIG_TCG_TARGET
>>
>> Shouldn't that rather go before the "#endif" instead?
>>
>> Also, would it be possible to rather adjust scripts/make-config-poison.sh
>> instead of poisoning this switch manually?
>
> No, I want to unconditionally poison it so that no other uses are ever
> introduced.
Ok, but then I think you should put a comment in front of it, explaining the
rationale, since it is not so obvious (at least it was not obvious for me
when I read the patch).
Thanks,
Thomas