Re: [PATCH v2 for-4.20? 5/6] radix-tree: introduce RADIX_TREE{,_INIT}()

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 04/02/2025 8:36 am, Jan Beulich wrote:
> On 03.02.2025 17:58, Jan Beulich wrote:
>> On 03.02.2025 17:48, Andrew Cooper wrote:
>>> On 03/02/2025 4:26 pm, Jan Beulich wrote:
>>>> ... now that static initialization is possible. Use RADIX_TREE() for
>>>> pci_segments and ivrs_maps.
>>>>
>>>> Requested-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>>>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>>> I'd not considered having RADIX_TREE() but it's nicer than my attempt.
>>>
>>> However,
>>>
>>>> --- a/xen/include/xen/radix-tree.h
>>>> +++ b/xen/include/xen/radix-tree.h
>>>> @@ -72,6 +72,9 @@ struct radix_tree_root {
>>>>   *** radix-tree API starts here **
>>>>   */
>>>>  
>>>> +#define RADIX_TREE_INIT() {}
>>> ... this ought to be (struct radix_tree_root){} so it can't be used with
>>> other types, and radix_tree_init() wants to become:
>>>
>>> void radix_tree_init(struct radix_tree_root *root)
>>> {
>>>         *root = RADIX_TREE_INIT();
>>> }
>>>
>>> instead of the raw memset(), so the connection is retained.
>> Can do; in fact I did consider both, but omitted them for simplicity.
> Sadly I've now learned the hard way that the former can't be done. Gcc
> 4.3 complains "initializer element is not constant", which - aiui - is
> correct when considering plain C99 (and apparently the GNU99 extension
> to this was introduced only later).
>
> What I can do is make this compiler version dependent, adding type-
> safety on at least all more modern compilers. Thoughts?

I think you can guess what my view is about us still supporting GCC 4.3.

As this needs backporting, lets just go with {} for now.

~Andrew