[PATCH] D142704: [C++20][Modules] Handle template declarations in header units.

[Iain Sandoe via Phabricator via cfe-commits]

iains added a comment.

I think we need to find a way to proceed - because this causes a regression on the llvm-16 branch, and that should be resolved soon, if possible.
What is your suggestion for a way forward?



================
Comment at: clang/lib/Sema/SemaDecl.cpp:15265
       FD->getFormalLinkage() == Linkage::ExternalLinkage &&
-      !FD->isInvalidDecl() && BodyKind != FnBodyKind::Delete &&
+      !FD->isInvalidDecl() && !IsFnTemplate && BodyKind != FnBodyKind::Delete &&
       BodyKind != FnBodyKind::Default && !FD->isInlined()) {
----------------
iains wrote:
> ChuanqiXu wrote:
> > iains wrote:
> > > rsmith wrote:
> > > > Would it make sense to use `!isa<FunctionTemplateDecl>(D)` here instead of adding `IsFnTemplate`?
> > > > Would it make sense to use `!isa<FunctionTemplateDecl>(D)` here instead of adding `IsFnTemplate`?
> > > 
> > > I have changed this to use FD->isTemplated() to match the changes for VarDecls - where the template decl is not available.  Would it be better to use the isa<>() ?
> > > 
> > It looks not bad to me to use `isTemplated ()`. And it looks like `!FD->isTemplateInstantiation()` is not tested? And it looks a little bit odd since the instantiated template should be implicitly inline.
> > It looks not bad to me to use `isTemplated ()`. 
> 
> Hmm. now I am not sure what you prefer; in a previous review there was an objection to not using the provided function in the decl class.  
> 
> What would be your suggestion here?
> 
> (we cannot use isa<VarTemplateDecl> when testing the variables case because the caller pulls out the templated decl before we get to test it) - so it is more consistent (IMO) to use the same interface in both tests.
> 
> >And it looks like `!FD->isTemplateInstantiation()` is not tested?
> 
> Please could you expand a bit on what you mean here?
>  (the test is clearly required, in practice)
> 
> > And it looks a little bit odd since the instantiated template should be implicitly inline.
> 
> Did you see @dlaikie's comment on this topic above?
> 
> 

> Did you see @dlaikie's comment on this topic above?

sorry @dblaikie 's comment is in the PR (https://github.com/llvm/llvm-project/issues/60079#issuecomment-1406856501)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142704/new/

https://reviews.llvm.org/D142704