================
@@ -1666,6 +1685,21 @@ namespace {
return inherited::TransformTemplateArgument(Input, Output, Uneval);
}
+ using TreeTransform::TransformTemplateSpecializationType;
+ QualType
+ TransformTemplateSpecializationType(TypeLocBuilder &TLB,
+ TemplateSpecializationTypeLoc TL) {
+ auto *T = TL.getTypePtr();
+ if (!getSema().ArgPackSubstIndex || !T->isSugared() ||
+ !isPackProducingBuiltinTemplateName(T->getTemplateName()))
+ return TreeTransform::TransformTemplateSpecializationType(TLB, TL);
+ // Look through sugar to get to the SubstBuiltinTemplatePackType that we
+ // need to substitute into.
+ QualType R = TransformType(T->desugar());
----------------
ilya-biryukov wrote:
> The issue I see in this case here is that you are just dropping the
> TemplateSpecializationType, why can't you just build a new one here to wrap
> over the transformed underlying type?
`TemplateSpecializationType` looks like the right sugar for the whole pack, but
after we pick a single element from it I don't think it models what we want
correctly. I.e. `dedup<int, double> == argpack{int, double}`, but **not equal**
to the individual `int` or `double` after we've picked them.
So I decided to stick with the lack of sugar rather than pick the incorrect one
(from that perspective).
Or do you feel `TemplateSpecializationType` is a good fit here?
https://github.com/llvm/llvm-project/pull/106730
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
