Re: r268594 - Fix implementation of C++'s restrictions on using-declarations referring to enumerators:
FYI, this change broke bionic, which has this exact pattern:
https://android.googlesource.com/platform/bionic/+/master/tools/relocation_packer/src/debug.h#84
struct A {
enum E {
X = 3
};
};
typedef A::E T;
using T::X;
I expect users are going to have lots of issues with this, and it probably
warrants a release note.
On Wed, May 4, 2016 at 7:13 PM, Richard Smith via cfe-commits <
cfe-commits@lists.llvm.org> wrote:
> Author: rsmith
> Date: Wed May 4 21:13:49 2016
> New Revision: 268594
>
> URL: http://llvm.org/viewvc/llvm-project?rev=268594&view=rev
> Log:
> Fix implementation of C++'s restrictions on using-declarations referring
> to enumerators:
>
> * an unscoped enumerator whose enumeration is a class member is itself a
> class
>member, so can only be the subject of a class-scope using-declaration.
>
> * a scoped enumerator cannot be the subject of a class-scope
> using-declaration.
>
> Added:
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3.cpp
> - copied, changed from r268583,
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3-cxx0x.cpp
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx11.cpp
> - copied, changed from r268583,
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx0x.cpp
> Removed:
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3-cxx0x.cpp
> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx0x.cpp
> Modified:
> cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
> cfe/trunk/lib/Sema/SemaDeclCXX.cpp
> cfe/trunk/test/CXX/drs/dr4xx.cpp
> cfe/trunk/test/SemaCXX/enum-scoped.cpp
>
> Modified: cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
> URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=268594&r1=268593&r2=268594&view=diff
>
> ==
> --- cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td (original)
> +++ cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td Wed May 4
> 21:13:49 2016
> @@ -396,7 +396,9 @@ def note_using_decl_class_member_workaro
>"use %select{an alias declaration|a typedef declaration|a reference}0 "
>"instead">;
> def err_using_decl_can_not_refer_to_namespace : Error<
> - "using declaration cannot refer to namespace">;
> + "using declaration cannot refer to a namespace">;
> +def err_using_decl_can_not_refer_to_scoped_enum : Error<
> + "using declaration cannot refer to a scoped enumerator">;
> def err_using_decl_constructor : Error<
>"using declaration cannot refer to a constructor">;
> def warn_cxx98_compat_using_decl_constructor : Warning<
>
> Modified: cfe/trunk/lib/Sema/SemaDeclCXX.cpp
> URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaDeclCXX.cpp?rev=268594&r1=268593&r2=268594&view=diff
>
> ==
> --- cfe/trunk/lib/Sema/SemaDeclCXX.cpp (original)
> +++ cfe/trunk/lib/Sema/SemaDeclCXX.cpp Wed May 4 21:13:49 2016
> @@ -7738,7 +7738,7 @@ bool Sema::CheckUsingShadowDecl(UsingDec
>// function will silently decide not to build a shadow decl, which
>// will pre-empt further diagnostics.
>//
> - // We don't need to do this in C++0x because we do the check once on
> + // We don't need to do this in C++11 because we do the check once on
>// the qualifier.
>//
>// FIXME: diagnose the following if we care enough:
> @@ -8227,7 +8227,7 @@ NamedDecl *Sema::BuildUsingDeclaration(S
> }
>}
>
> - // C++0x N2914 [namespace.udecl]p6:
> + // C++14 [namespace.udecl]p6:
>// A using-declaration shall not name a namespace.
>if (R.getAsSingle()) {
> Diag(IdentLoc, diag::err_using_decl_can_not_refer_to_namespace)
> @@ -8235,6 +8235,16 @@ NamedDecl *Sema::BuildUsingDeclaration(S
> return BuildInvalid();
>}
>
> + // C++14 [namespace.udecl]p7:
> + // A using-declaration shall not name a scoped enumerator.
> + if (auto *ED = R.getAsSingle()) {
> +if (cast(ED->getDeclContext())->isScoped()) {
> + Diag(IdentLoc, diag::err_using_decl_can_not_refer_to_scoped_enum)
> +<< SS.getRange();
> + return BuildInvalid();
> +}
> + }
> +
>UsingDecl *UD = BuildValid();
>
>// The normal rules do not apply to inheriting constructor declarations.
> @@ -8359,8 +8369,10 @@ bool Sema::CheckUsingDeclQualifier(Sourc
>
> // If we weren't able to compute a valid scope, it must be a
> // dependent class scope.
> -if (!NamedContext || NamedContext->isRecord()) {
> - auto *RD = dyn_cast_or_null(NamedContext);
> +if (!NamedContext || NamedContext->getRedeclContext()->isRecord()) {
> + auto *RD = NamedContext
> + ?
> cast(NamedContext->getRedeclContext())
> + : nullptr;
>if (RD && RequireCompleteDeclContext(const_cast(SS),
> RD))
> RD = nullptr;
>
> @@ -8444,7 +8456,7
Re: r268594 - Fix implementation of C++'s restrictions on using-declarations referring to enumerators:
On Thu, May 5, 2016 at 10:40 AM, Reid Kleckner wrote:
> FYI, this change broke bionic, which has this exact pattern:
>
> https://android.googlesource.com/platform/bionic/+/master/tools/relocation_packer/src/debug.h#84
>
> struct A {
> enum E {
> X = 3
> };
> };
> typedef A::E T;
> using T::X;
>
> I expect users are going to have lots of issues with this, and it probably
> warrants a release note.
>
Given that GCC rejects this code (and has done for as long as it's accepted
the Enum::Member syntax), I don't expect the problems to be too widespread.
If they are, we could trivially accept this as an extension; it seems like
a pointless restriction.
> On Wed, May 4, 2016 at 7:13 PM, Richard Smith via cfe-commits <
> cfe-commits@lists.llvm.org> wrote:
>
>> Author: rsmith
>> Date: Wed May 4 21:13:49 2016
>> New Revision: 268594
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=268594&view=rev
>> Log:
>> Fix implementation of C++'s restrictions on using-declarations referring
>> to enumerators:
>>
>> * an unscoped enumerator whose enumeration is a class member is itself a
>> class
>>member, so can only be the subject of a class-scope using-declaration.
>>
>> * a scoped enumerator cannot be the subject of a class-scope
>> using-declaration.
>>
>> Added:
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3.cpp
>> - copied, changed from r268583,
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3-cxx0x.cpp
>>
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx11.cpp
>> - copied, changed from r268583,
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx0x.cpp
>> Removed:
>>
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p3-cxx0x.cpp
>>
>> cfe/trunk/test/CXX/dcl.dcl/basic.namespace/namespace.udecl/p6-cxx0x.cpp
>> Modified:
>> cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
>> cfe/trunk/lib/Sema/SemaDeclCXX.cpp
>> cfe/trunk/test/CXX/drs/dr4xx.cpp
>> cfe/trunk/test/SemaCXX/enum-scoped.cpp
>>
>> Modified: cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td
>> URL:
>> http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?rev=268594&r1=268593&r2=268594&view=diff
>>
>> ==
>> --- cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td (original)
>> +++ cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td Wed May 4
>> 21:13:49 2016
>> @@ -396,7 +396,9 @@ def note_using_decl_class_member_workaro
>>"use %select{an alias declaration|a typedef declaration|a reference}0 "
>>"instead">;
>> def err_using_decl_can_not_refer_to_namespace : Error<
>> - "using declaration cannot refer to namespace">;
>> + "using declaration cannot refer to a namespace">;
>> +def err_using_decl_can_not_refer_to_scoped_enum : Error<
>> + "using declaration cannot refer to a scoped enumerator">;
>> def err_using_decl_constructor : Error<
>>"using declaration cannot refer to a constructor">;
>> def warn_cxx98_compat_using_decl_constructor : Warning<
>>
>> Modified: cfe/trunk/lib/Sema/SemaDeclCXX.cpp
>> URL:
>> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaDeclCXX.cpp?rev=268594&r1=268593&r2=268594&view=diff
>>
>> ==
>> --- cfe/trunk/lib/Sema/SemaDeclCXX.cpp (original)
>> +++ cfe/trunk/lib/Sema/SemaDeclCXX.cpp Wed May 4 21:13:49 2016
>> @@ -7738,7 +7738,7 @@ bool Sema::CheckUsingShadowDecl(UsingDec
>>// function will silently decide not to build a shadow decl, which
>>// will pre-empt further diagnostics.
>>//
>> - // We don't need to do this in C++0x because we do the check once on
>> + // We don't need to do this in C++11 because we do the check once on
>>// the qualifier.
>>//
>>// FIXME: diagnose the following if we care enough:
>> @@ -8227,7 +8227,7 @@ NamedDecl *Sema::BuildUsingDeclaration(S
>> }
>>}
>>
>> - // C++0x N2914 [namespace.udecl]p6:
>> + // C++14 [namespace.udecl]p6:
>>// A using-declaration shall not name a namespace.
>>if (R.getAsSingle()) {
>> Diag(IdentLoc, diag::err_using_decl_can_not_refer_to_namespace)
>> @@ -8235,6 +8235,16 @@ NamedDecl *Sema::BuildUsingDeclaration(S
>> return BuildInvalid();
>>}
>>
>> + // C++14 [namespace.udecl]p7:
>> + // A using-declaration shall not name a scoped enumerator.
>> + if (auto *ED = R.getAsSingle()) {
>> +if (cast(ED->getDeclContext())->isScoped()) {
>> + Diag(IdentLoc, diag::err_using_decl_can_not_refer_to_scoped_enum)
>> +<< SS.getRange();
>> + return BuildInvalid();
>> +}
>> + }
>> +
>>UsingDecl *UD = BuildValid();
>>
>>// The normal rules do not apply to inheriting constructor
>> declarations.
>> @@ -8359,8 +8369,10 @@ bool Sema::CheckUsingDeclQualifier(Sourc
>>
>> // If we weren't able to compute a valid scope, it must be a
>> // dependent class sc
Re: r268594 - Fix implementation of C++'s restrictions on using-declarations referring to enumerators:
On Thu, May 5, 2016 at 11:15 AM, Richard Smith wrote: > Given that GCC rejects this code (and has done for as long as it's > accepted the Enum::Member syntax), I don't expect the problems to be too > widespread. If they are, we could trivially accept this as an extension; it > seems like a pointless restriction. > I think there will be widespread issues on Mac, where many projects haven't built with GCC in years. ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
Re: r268594 - Fix implementation of C++'s restrictions on using-declarations referring to enumerators:
On Thu, May 5, 2016 at 11:21 AM, Reid Kleckner wrote: > On Thu, May 5, 2016 at 11:15 AM, Richard Smith > wrote: > >> Given that GCC rejects this code (and has done for as long as it's >> accepted the Enum::Member syntax), I don't expect the problems to be too >> widespread. If they are, we could trivially accept this as an extension; it >> seems like a pointless restriction. >> > > I think there will be widespread issues on Mac, where many projects > haven't built with GCC in years. > I've added some release notes in r268663. If people complain that this is breaking a significant amount of code, we can think about allowing this as an extension. ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
