Excerpts from Rainer Orth's message of Januar 5, 2025 12:42 pm:
> Hi Iain,
>
>> At some point during GCC 15 development, the compiler-generated code for
>> D ModuleInfo support now gets transformed into the following on x86:
>>
>> _12 = (uint) &__stop_minfo;
>> _13 = (uint) &__start_minfo;
>> _14 = (uint) &gdc.dso_slot;
>> _15 = {1, _14, _13, _12};
>> gdc.dso_initialized = 0;
>> MEM <unsigned char[8]> [(void *)&D.1974 + 16B] = {};
>> MEM <vector(4) uint> [(void *)&D.1974] = _15;
>> _d_dso_registry (&D.1974);
>>
>> This causes the scan-tree-dump-not test to fail. As ModuleInfo is not
>> what's being checked for in PR110406, remove its generation altogether.
>>
>> Bootstrapped and regression tested on x86_64-linux-gnu/-m32, committed
>> to mainline.
>
> this is PR tree-optimization/117707.
>
> Rainer
>
Thanks Rainer!
I'm currently on catch-up, and unfortunately my unread emails are in
their hundreds, and I couldn't find a related PR when first looking.
I'll close the PR now, referencing the commit.
Iain.
