[PATCH] make isScopedEnumerationType look through typedefs

[Stephan Bergmann]

On 12/19/2014 06:05 AM, Richard Smith wrote:
> LGTM, do you need someone to commit this for you?

yes, please; I don't have commit rights

> On Thu, Dec 18, 2014 at 12:28 AM, Stephan Bergmann <sbergman at redhat.com
> <mailto:sbergman at redhat.com>> wrote:
>
>     On 12/17/2014 08:40 PM, Richard Smith wrote:
>
>         The right way to write this is:
>
>         -  if (const EnumType *ET = dyn_cast<EnumType>(T))
>         +  if (const EnumType *ET = T->getAs<EnumType>())
>
>
>     updated patch attached
>
>         On Tue, Dec 16, 2014 at 5:51 AM, Rafael Espíndola
>         <rafael.espindola at gmail.com <mailto:rafael.espindola at gmail.com>
>         <mailto:rafael.espindola at __gmail.com
>         <mailto:rafael.espindola at gmail.com>>> wrote:
>
>              LGTM.
>
>              Clang is not my expertise, but the patch looks clearly
>         correct, fixes
>              a bug and has a testcase.
>
>
>              On 16 December 2014 at 04:49, Stephan Bergmann
>         <sbergman at redhat.com <mailto:sbergman at redhat.com>
>              <mailto:sbergman at redhat.com <mailto:sbergman at redhat.com>>>
>         wrote:
>               > ping
>               >
>               >
>               > On 11/25/2014 03:04 PM, Rafael Espíndola wrote:
>               >>
>               >> ccing Richard.
>               >> On 25 November 2014 at 03:38, Stephan Bergmann
>              <sbergman at redhat.com <mailto:sbergman at redhat.com>
>         <mailto:sbergman at redhat.com <mailto:sbergman at redhat.com>>>
>               >> wrote:
>               >>>
>               >>> ping
>               >>>
>               >>>
>               >>> On 11/10/2014 06:18 PM, Stephan Bergmann wrote:
>               >>>>
>               >>>>
>               >>>> On 10/23/2014 10:41 AM, David Majnemer wrote:
>               >>>>>
>               >>>>>
>               >>>>> Per the developer policy [1], please add a test case to
>              accompany your
>               >>>>> change.
>               >>>>>
>               >>>>> [1]
>         http://llvm.org/docs/__DeveloperPolicy.html#test-__cases
>         <http://llvm.org/docs/DeveloperPolicy.html#test-cases>
>               >>>>>
>               >>>>> Looks good otherwise.
>               >>>>
>               >>>>
>               >>>>
>               >>>> Attached updated patch including test case.
>               >>>>
>               >>>>> On Thu, Oct 23, 2014 at 12:43 AM, Stephan Bergmann
>              <sbergman at redhat.com <mailto:sbergman at redhat.com>
>         <mailto:sbergman at redhat.com <mailto:sbergman at redhat.com>>
>               >>>>> <mailto:sbergman at redhat.com
>         <mailto:sbergman at redhat.com> <mailto:sbergman at redhat.com
>         <mailto:sbergman at redhat.com>>>> wrote:
>               >>>>>
>               >>>>>      [ping]
>               >>>>>
>               >>>>>
>               >>>>>      On 10/14/2014 04:12 PM, Stephan Bergmann wrote:
>               >>>>>
>               >>>>>              $ clang++ --version
>               >>>>>              clang version 3.6.0 (trunk 219190)
>               >>>>>              Target: x86_64-unknown-linux-gnu
>               >>>>>              Thread model: posix
>               >>>>>
>               >>>>>              $ cat test.cc
>               >>>>>              enum class E { e };
>               >>>>>              typedef E E2;
>               >>>>>              E2 f1() { return E::e; }
>               >>>>>              E f2() { return E::e; }
>               >>>>>              bool g1() { return !f1(); }
>               >>>>>              bool g2() { return !f2(); }
>               >>>>>
>               >>>>>              $ clang++ -std=c++11 -c test.cc
>               >>>>>              test.cc:6:20: error: invalid argument
>         type 'E'
>              to unary
>               >>>>>              expression
>               >>>>>              bool g2() { return !f2(); }
>               >>>>>                                  ^~~~~
>               >>>>>              1 error generated.
>               >>>>>
>               >>>>>
>               >>>>>          rightly complains about the invalid contextual
>              conversion to
>               >>>>>          bool in g2
>               >>>>>          but erroneously not also about the one in g1.
>               >>>>>
>               >>>>>              Index: lib/Sema/SemaExpr.cpp
>               >>>>>
>               >>>>>
>
>         ==============================____============================__==__=======
>               >>>>>              --- lib/Sema/SemaExpr.cpp    (revision
>         219190)
>               >>>>>              +++ lib/Sema/SemaExpr.cpp    (working copy)
>               >>>>>              @@ -7414,7 +7414,7 @@
>               >>>>>                }
>               >>>>>
>               >>>>>                static bool
>         isScopedEnumerationType(____QualType
>              T) {
>               >>>>>              -  if (const EnumType *ET =
>         dyn_cast<EnumType>(T))
>               >>>>>              +  if (const EnumType *ET =
>               >>>>>
>         dyn_cast<EnumType>(T.____getCanonicalType()))
>               >>>>>
>               >>>>>                    return ET->getDecl()->isScoped();
>               >>>>>                  return false;
>               >>>>>                }
>               >>>>>
>               >>>>>
>               >>>>>          would fix that for me.
>
>