<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><br><div><div>On Jun 7, 2012, at 6:09 PM, Kaelyn Uhrain wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div class="gmail_quote">On Thu, Jun 7, 2012 at 3:47 PM, Douglas Gregor <span dir="ltr"><<a href="mailto:dgregor@apple.com" target="_blank">dgregor@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="HOEnZb"><div class="h5"><br>
On May 18, 2012, at 4:42 PM, Kaelyn Uhrain wrote:<br>
<br>
> Author: rikka<br>
> Date: Fri May 18 18:42:49 2012<br>
> New Revision: 157085<br>
><br>
> URL: <a href="http://llvm.org/viewvc/llvm-project?rev=157085&view=rev" target="_blank">http://llvm.org/viewvc/llvm-project?rev=157085&view=rev</a><br>
> Log:<br>
> Suggest adding 'typename' when it would make the compiler<br>
> accept the template argument expression as a type.<br>
><br>
> Modified:<br>
> cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td<br>
> cfe/trunk/lib/Sema/SemaTemplate.cpp<br>
> cfe/trunk/test/SemaTemplate/typename-specifier.cpp<br>
><br>
> Modified: cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td<br>
> URL: <a href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=157085&r1=157084&r2=157085&view=diff" target="_blank">http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=157085&r1=157084&r2=157085&view=diff</a><br>
> ==============================================================================<br>
> --- cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td (original)<br>
> +++ cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td Fri May 18 18:42:49 2012<br>
> @@ -2361,6 +2361,8 @@<br>
> def note_member_of_template_here : Note<"member is declared here">;<br>
> def err_template_arg_must_be_type : Error<<br>
> "template argument for template type parameter must be a type">;<br>
> +def err_template_arg_must_be_type_suggest : Error<<br>
> + "template argument for template type parameter must be a type; did you forget 'typename'?">;<br>
> def err_template_arg_must_be_expr : Error<<br>
> "template argument for non-type template parameter must be an expression">;<br>
> def err_template_arg_nontype_ambig : Error<<br>
><br>
> Modified: cfe/trunk/lib/Sema/SemaTemplate.cpp<br>
> URL: <a href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaTemplate.cpp?rev=157085&r1=157084&r2=157085&view=diff" target="_blank">http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaTemplate.cpp?rev=157085&r1=157084&r2=157085&view=diff</a><br>
> ==============================================================================<br>
> --- cfe/trunk/lib/Sema/SemaTemplate.cpp (original)<br>
> +++ cfe/trunk/lib/Sema/SemaTemplate.cpp Fri May 18 18:42:49 2012<br>
> @@ -2438,6 +2438,45 @@<br>
><br>
> return true;<br>
> }<br>
> + case TemplateArgument::Expression: {<br>
> + // We have a template type parameter but the template argument is an<br>
> + // expression; see if maybe it is missing the "typename" keyword.<br>
> + CXXScopeSpec SS;<br>
> + DeclarationNameInfo NameInfo;<br>
> +<br>
> + if (DeclRefExpr *ArgExpr = dyn_cast<DeclRefExpr>(Arg.getAsExpr())) {<br>
> + SS.Adopt(ArgExpr->getQualifierLoc());<br>
> + NameInfo = ArgExpr->getNameInfo();<br>
<br>
</div></div>Shouldn't we check that there is a nested-name-specifier here?<br></blockquote><div><br></div><div>As far as I'm aware, there is no need. The code below is not dependent on the CXXScopeSpec being valid, and if there isn't a nested-name-specifier then the CXXScopeSpec will be invalid after adopting an empty NestedNameSpecifierLoc. Plus, from what I can tell, any checks for whether the expression needs to have a nested-name-specifier are handled elsewhere (IIRC they're triggered before this code is reached).</div></div></blockquote><div><br></div>Okay, this seems to work because we can't get NotFoundInCurrentInstantiation with unqualified lookup.</div><div><br><blockquote type="cite"><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0.8ex; border-left-width: 1px; border-left-color: rgb(204, 204, 204); border-left-style: solid; padding-left: 1ex; position: static; z-index: auto; "><div class="im">> + LookupResult Result(*this, NameInfo, LookupOrdinaryName);<br>
> + LookupParsedName(Result, CurScope, &SS);<br>
> +<br>
> + bool CouldBeType = Result.getResultKind() ==<br>
> + LookupResult::NotFoundInCurrentInstantiation;<br>
> +<br>
> + for (LookupResult::iterator I = Result.begin(), IEnd = Result.end();<br>
> + !CouldBeType && I != IEnd; ++I) {<br>
> + CouldBeType = isa<TypeDecl>(*I);<br>
> + }<br>
<br>
</div>I don't understand why we have this loop here. If the name lookup is able to find a type, we wouldn't be here in this code. It seems like the only interesting case is the NotFoundInCurrentInstantiation case.<br>
</blockquote><div><br></div><div>D'oh. It was left over from when I was trying to figure out how to look up whether an identifier might be a type and even how to get the nested-name-specifier in a form that I could use for the lookups. IIRC at least one of the avenues I'd tried needed the loop to handle lookups that returned multiple decls.</div></div></blockquote><div><br></div><div>Come to think of it, do you even need to do a lookup here? A DeclRefExpr won't refer to a type, so it's actually not an interesting case. DependentScopeDeclRefExpr and CXXDependentScopeMemberExpr (the latter only when it has a qualifier) are the interesting cases, and we wouldn't have built those nodes unless name lookup for the entity ended up with NotFoundInCurrentInstantiation. </div><br><blockquote type="cite"><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0.8ex; border-left-width: 1px; border-left-color: rgb(204, 204, 204); border-left-style: solid; padding-left: 1ex; position: static; z-index: auto; ">
<div class="im"><br>
> + if (CouldBeType) {<br>
> + SourceLocation Loc = AL.getSourceRange().getBegin();<br>
> + Diag(Loc, diag::err_template_arg_must_be_type_suggest);<br>
> + Diag(Param->getLocation(), diag::note_template_param_here);<br>
> + return true;<br>
> + }<br>
<br>
</div>Did you mean for this to be a Fix-It? If so, it needs to recover by changing the template argument itself. If you're not planning on doing that work, it's fine, but please leave a FIXME to say that we'd like to do that work.<br>
</blockquote><div><br></div><div>I've added a FIXME since a Fix-It was beyond what I had able to figure out while muddling through all of the template code...</div><div><br></div><div>The changes mentioned above along with a couple more tests have been committed as r158185.</div>
</div></blockquote><br>Okay, thanks for the fixes!</div><div><br></div><span class="Apple-tab-span" style="white-space:pre"> </span>- Doug</body></html>