dwblaikie wrote:
> Thanks @dwblaikie
>
> > Bit unfortunate to store template parameters in different ways (in the
> > extraData for the alias template, but in the templateParams for the
> > composite types) - but I guess it'd be more invasive to try to represent
> > alias templates as composite types?
>
> I hadn't actually tried using DICompsiteType. I didn't choose DIDerivedType
> for any particularly principled reasons: typedefs use derived types which
> made it very easy to find places that needed updating, and an alias "felt"
> more like a derived type than composite type. Having looked only briefly, I
> don't _think_ it'd be any more invasive to use composite types instead (not
> 100% sure without diving in). I'm happy to give that a go if that is your
> preference?
Eh, I guess if we've already got DIGlobalVariable having template params, we
aren't going to unify all 3 cases - not sure if unifying 2 out of three is
particularly valuable, especially if it ends up with a different conflict (that
alias templates look less like aliases & so need special casing in that
direction)
But perhaps at least you could add named accessors to DIDerivedType for the
template params, same as DIGlobalVariable/DICompositeType have?
(oh, and in case anyone hasn't mentioned it already - this would generally be
committed in smaller pieces upstream - adding the LLVM functionality first,
then adding clang patches that use that functionality)
> I've just found an input that causes an assertion failure in
> `CollectTemplateParams` with my implementation:
>
> ```
> template<typename Y, typename Z>
> struct X {
> Y m1;
> Z m2;
> };
>
> template<typename... Ts>
> using A = X<Ts...>;
>
> A<int, int> a;
> ```
>
> There's a mismatch between the template parameter list and template argument
> list sizes, created here:
>
> ```
> if (CGM.getCodeGenOpts().DebugTemplateAlias) {
> TemplateArgs Args = {TD->getTemplateParameters(),
> Ty->template_arguments()}; // <--- here
> ```
>
> I notice that we emit `DW_TAG_GNU_template_parameter_pack` for, e.g.,
> variadic template struct instantiations. I've not figured out how to fix this
> just yet, but thought I'd bring it up in case there's something obvious.
yeah, probably worth checking the APIs that we use for building the template
args for classes and variable templates, they probably have some handling for
this - would be good to share/reuse that code, rather than duplicating it, once
you find it
https://github.com/llvm/llvm-project/pull/87623
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
