[cfe-commits] [PATCH] Allow typedef with unnecessary "typename" when ms-extensions are enabled.

[Richard Smith]

On Mon, May 7, 2012 at 4:23 AM, Will Wilson <will at indefiant.com> wrote:

> > You should consume the 'typename' token, and try to recover as if the
> > 'typename' keyword were not present, following the logic present later
> on in
> > that function. As a special case, if that later logic were to find that
> the
> > tokens after the 'typename' keyword can't be annotated as a type name,
> you
> > should produce a hard error (even with MS extensions enabled).
> >
> > (In particular, for 'typename identifier', it will be the identifier you
> > annotate, not the 'typename' keyword.)
>
> Sorry for the delay! I've attached an updated patch (with test cases)
> that seems to do the job. I've also had to update two other tests to
> allow for the different code paths (and thus different errors)
> followed due to the attempt at recovering from the error.
>
> All tests pass locally. It'd be great if some other people could
> verify it as well.
>

A few things: This is still only producing a warning if 'typename' is
followed by something which isn't a type name in MS mode. That needs to be
fixed to produce an error in that case. Also, it would be great to apply
typo correction in this case, and to avoid duplicating the code from the
second half of the function.

How about just recursively calling TryAnnotateTypeOrScopeToken after
consuming the invald 'typename' token, then emitting the error diagnostic
if it fails or we're not in MS mode, and the warning diagnostic otherwise?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20120507/cf07420e/attachment.html>