[PATCH] N3651 -- Variable Templates -- A first (incremental) step.

[Richard Smith]

Hi Larisse,

This looks like a great start!

Adding two pointers to every VarDecl to support this seems like quite
a high cost to pay. The DefinitionData field looks unnecessary (we can
reconstruct the definition by following the redeclaration chain
already), and if you drop that you should be able to remove the extra
parameter to VarDecl::Create too. Can you find somewhere else to hang
the TemplateOrInstantiation data, so we don't need to pay for it if we
don't use it? (Maybe derive from the ExtInfo in DeclaratorDecl and put
the data there, or add a class derived from VarDecl for the cases
where we need this data, or put it in a DenseMap on the ASTContext.)

+      return IsVarTemplate? NameClassification::TypeTemplate(Template)
+              : NameClassification::VarTemplate(Template);

This looks backwards.

+      if (!getLangOpts().CPlusPlus1y) {
+        if (TemplateParams->size() > 0) {
+          // There is no such thing as a variable template.
+          Diag(D.getIdentifierLoc(), diag::err_template_variable)
+            << II
+            << SourceRange(TemplateParams->getTemplateLoc(),
+                           TemplateParams->getRAngleLoc());

Can you downgrade this from an Error to an ExtWarn? There doesn't seem
to be a good reason not to support variable templates as an extension
in all C++ language modes.

This patch needs some cleaning up to prepare for check in (removing
commented-out code etc), and a lot more tests; I suggest going over
the patch and trying to add some tests for every change, where that's
possible. Other than that, this seems close to being ready to commit,
as a first step towards variable templates.

Thank you!

On Tue, Jun 18, 2013 at 5:50 PM, Larisse Voufo <lvoufo at google.com> wrote:
> There was a bug in this earlier version of the patch. The correction is
> attached.
>
>
> On Wed, Jun 12, 2013 at 6:52 PM, Larisse Voufo <lvoufo at google.com> wrote:
>>
>> Hello --
>>
>> I have  started implementing variable templates, and you can find attached
>> the current state.
>> While it may cover more cases, the only guarantee it has is that top level
>> variable templates can be declared and used (as illustrated in the
>> SemaCXX/cxx1y-variable-templates.cpp test file). That is, it does not yet
>> cover any of the following:
>>
>> * variable template explicit (partial) specialization and instantiations,
>> * static member variable templates,
>> * constexpr variables, and
>> * default variable template arguments.
>>
>> Things will get much cleaner as I complete the implementation.
>>
>> Thanks,
>> -- Larisse.
>>
>>
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>