git: e92ab3adf4f9 - stable/13 - atomic: make atomic_store_ptr type-aware

[Mateusz Guzik]

I thought I responded to this bit, I did not address the (in my
opinion incorrect) claim that a fence can't be avoided rendering
consume primitive pointless. I have a somewhat reply for this.

but tl;dr yes, I'll add validation for pointer types

A bug was reported in the namecache which I want to take care of
before I do other stuff.

On 2/2/21, John Baldwin <jhb at freebsd.org> wrote:
> On 2/1/21 4:40 AM, Mateusz Guzik wrote:
>> The branch stable/13 has been updated by mjg:
>>
>> URL:
>> https://cgit.FreeBSD.org/src/commit/?id=e92ab3adf4f9848933439c9c2fca36df290c0884
>>
>> commit e92ab3adf4f9848933439c9c2fca36df290c0884
>> Author:     Mateusz Guzik <mjg at FreeBSD.org>
>> AuthorDate: 2021-01-25 19:39:14 +0000
>> Commit:     Mateusz Guzik <mjg at FreeBSD.org>
>> CommitDate: 2021-02-01 12:39:16 +0000
>>
>>      atomic: make atomic_store_ptr type-aware
>>
>>      (cherry picked from commit cc96f92a570e05636a20fdd15d4616b127bb9ecc)
>
> A couple of folks expressed concerns about this commit when it landed
> in main due to it being too flexible (it allows you to store pointers
> to non-pointer types, e.g.:
>
>    char c;
>    uintptr_t p;
>
>    atomic_store_ptr(&c, p);
>
> Silently becomes:
>
>    atomic_store_char(&c, (char)p);
>
> Are you planning to address this?
>
> --
> John Baldwin
>


-- 
Mateusz Guzik <mjguzik gmail.com>