Hello,
On Fri, Oct 14 2022, Martin Liška wrote:
> On 10/14/22 08:12, Richard Biener wrote:
>> On Fri, Oct 14, 2022 at 12:54 AM Eric Botcazou via Gcc-patches
>> <gcc-patches@gcc.gnu.org> wrote:
>>>
>>>> Not a fan as it could potentially hide a real issue, but I don't really
>>>> have a better solution.
>>>
>>> Thanks.
>>>
>>>> I pondered suggesting "access" affect type identity, but the cases where
>>>> that's really important are probably better handled by the "fn spec"
>>>> attribute, leaving "access" strictly impacting diagnostics.
>>>
>>> I can expand a bit here, because I tried to change the "access" attribute
>>> that
>>> way and this badly breaks the C compiler, for example:
>>>
>>> int foo (int n, char m[1][n]);
>>>
>>> int foo (int n, char m[1][n]) {}
>>>
>>> no longer compiles with an error about different function types.
>>
>> Note in discussion with IPA folks we agreed that IPA cloning that modifies
>> arguments either has to remove access attributes, adjust them or refrain
>> from cloning.
>>
>> Martin - has anything been done to this respect?
>
> I think it's more for Martin Jambor who's the IPA specialist when it comes
> to parameter manipulation.
>
They are being dropped since 2af63f0f53a
Adjusting them accordingly is an item buried quite deep in my TODO list.
>>
>> I suppose there's also a way to figure if a clone has arguments
>> changed in any way?
Look whether clone_info::get (node) exists and its param_adjustments is
non-NULL.
In theory the param_adjustments could contain description of the very
same signature the original function has but in practice it does not
currently happen and is unlikely to happen ever.
Martin
