Re: [RFC PATCH] efi: const correct EFI functions

[Jan Beulich <jbeulich@xxxxxxxx>]

On 14.09.2020 16:46, Trammell Hudson wrote:
> On Monday, September 14, 2020 10:30 AM, Jan Beulich <jbeulich@xxxxxxxx> wrote:
>> On 14.09.2020 16:25, Trammell Hudson wrote:
>>> By defining IN as const, the EFI handler functions become almost
>>> const-correct and allow most of the rest of the EFI boot code to
>>> use constant strings.
>>
>> How does this work with combined "IN OUT"? I'm afraid there is a
>> reason why things aren't done the way you suggest.
> 
> WTF FFS UEFI. They really do continue to find new ways to disappoint me.
> 
> So I see three options:
> 
> Option 1 is to retroactively modify the C spec to allow us to have
> a "nonconst" that will override any prior "const" modifiers
> (last one wins, like Duck Season/Rabbit Season).

LoL

> Option 2 would be to modify the imported header to change
> the 30 uses of "IN OUT" to "INOUT" and define that to be an
> empty string.

This is something that one might hope even the gnu-efi folks
could be persuaded to do.

> Option 3 would be to write wrappers for the few functions that are
> used in the EFI boot path that cast-away the constness of their
> arguments (while also silently cursing the UEFI forum for not
> writing const-correct code).

This would be kind of a last resort fallback (except for the
cursing, which of course we can do at any time).

Jan