Re: r272247 - [Sema] Don't crash when a field w/ a mem-initializer clashes with a record name

[Kim Gräsman via cfe-commits]

On Thu, Jun 9, 2016 at 4:51 PM, David Majnemer  wrote:
> It would mean that the instantiation of the class template gained a field
> which should be impossible.

OK, so assert(Lookup.size() > 0) always holds? Makes sense, thanks.

- Kim
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: r272247 - [Sema] Don't crash when a field w/ a mem-initializer clashes with a record name

[David Majnemer via cfe-commits]

It would mean that the instantiation of the class template gained a field
which should be impossible.

On Thursday, June 9, 2016, Kim Gräsman  wrote:

> On Thu, Jun 9, 2016 at 7:26 AM, David Majnemer via cfe-commits
> > wrote:
> > Author: majnemer
> > Date: Thu Jun  9 00:26:56 2016
> > New Revision: 272247
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=272247=rev
> > Log:
> > [Sema] Don't crash when a field w/ a mem-initializer clashes with a
> record name
> >
> > It is possible for a field and a class to have the same name.  In such
> > cases, performing lookup for the field might return a result set with
> > more than one entry.  An overzealous assertion fired, causing us to
> > crash instead of using the non-class lookup result.
> >
> > This fixes PR28060.
> >
> > Modified:
> > cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
> > cfe/trunk/test/SemaCXX/member-init.cpp
> >
> > Modified: cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
> > URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp?rev=272247=272246=272247=diff
> >
> ==
> > --- cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp (original)
> > +++ cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp Thu Jun  9 00:26:56
> 2016
> > @@ -2637,8 +2637,7 @@ Sema::InstantiateClassMembers(SourceLoca
> >  Instantiation->getTemplateInstantiationPattern();
> >  DeclContext::lookup_result Lookup =
> >  ClassPattern->lookup(Field->getDeclName());
> > -assert(Lookup.size() == 1);
> > -FieldDecl *Pattern = cast(Lookup[0]);
> > +FieldDecl *Pattern = cast(Lookup.front());
> >  InstantiateInClassInitializer(PointOfInstantiation, Field,
> Pattern,
> >TemplateArgs);
> >}
>
> Now what if there is no match? Or is that guaranteed (given the prior
> assert)?
>
> - Kim
>
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: r272247 - [Sema] Don't crash when a field w/ a mem-initializer clashes with a record name

[Kim Gräsman via cfe-commits]

On Thu, Jun 9, 2016 at 7:26 AM, David Majnemer via cfe-commits
 wrote:
> Author: majnemer
> Date: Thu Jun  9 00:26:56 2016
> New Revision: 272247
>
> URL: http://llvm.org/viewvc/llvm-project?rev=272247=rev
> Log:
> [Sema] Don't crash when a field w/ a mem-initializer clashes with a record 
> name
>
> It is possible for a field and a class to have the same name.  In such
> cases, performing lookup for the field might return a result set with
> more than one entry.  An overzealous assertion fired, causing us to
> crash instead of using the non-class lookup result.
>
> This fixes PR28060.
>
> Modified:
> cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
> cfe/trunk/test/SemaCXX/member-init.cpp
>
> Modified: cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
> URL: 
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp?rev=272247=272246=272247=diff
> ==
> --- cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp (original)
> +++ cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp Thu Jun  9 00:26:56 2016
> @@ -2637,8 +2637,7 @@ Sema::InstantiateClassMembers(SourceLoca
>  Instantiation->getTemplateInstantiationPattern();
>  DeclContext::lookup_result Lookup =
>  ClassPattern->lookup(Field->getDeclName());
> -assert(Lookup.size() == 1);
> -FieldDecl *Pattern = cast(Lookup[0]);
> +FieldDecl *Pattern = cast(Lookup.front());
>  InstantiateInClassInitializer(PointOfInstantiation, Field, Pattern,
>TemplateArgs);
>}

Now what if there is no match? Or is that guaranteed (given the prior assert)?

- Kim
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

r272247 - [Sema] Don't crash when a field w/ a mem-initializer clashes with a record name

[David Majnemer via cfe-commits]

Author: majnemer
Date: Thu Jun  9 00:26:56 2016
New Revision: 272247

URL: http://llvm.org/viewvc/llvm-project?rev=272247=rev
Log:
[Sema] Don't crash when a field w/ a mem-initializer clashes with a record name

It is possible for a field and a class to have the same name.  In such
cases, performing lookup for the field might return a result set with
more than one entry.  An overzealous assertion fired, causing us to
crash instead of using the non-class lookup result.

This fixes PR28060.

Modified:
cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
cfe/trunk/test/SemaCXX/member-init.cpp

Modified: cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp?rev=272247=272246=272247=diff
==
--- cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp (original)
+++ cfe/trunk/lib/Sema/SemaTemplateInstantiate.cpp Thu Jun  9 00:26:56 2016
@@ -2637,8 +2637,7 @@ Sema::InstantiateClassMembers(SourceLoca
 Instantiation->getTemplateInstantiationPattern();
 DeclContext::lookup_result Lookup =
 ClassPattern->lookup(Field->getDeclName());
-assert(Lookup.size() == 1);
-FieldDecl *Pattern = cast(Lookup[0]);
+FieldDecl *Pattern = cast(Lookup.front());
 InstantiateInClassInitializer(PointOfInstantiation, Field, Pattern,
   TemplateArgs);
   }

Modified: cfe/trunk/test/SemaCXX/member-init.cpp
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/SemaCXX/member-init.cpp?rev=272247=272246=272247=diff
==
--- cfe/trunk/test/SemaCXX/member-init.cpp (original)
+++ cfe/trunk/test/SemaCXX/member-init.cpp Thu Jun  9 00:26:56 2016
@@ -192,3 +192,13 @@ struct S {
   int x[3] = {[N] = 3};
 };
 }
+
+namespace PR28060 {
+template 
+void foo(T v) {
+  struct s {
+T *s = 0;
+  };
+}
+template void foo(int);
+}


___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits