Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv accepted this revision.
faisalv added a reviewer: faisalv.
faisalv added a comment.
This revision is now accepted and ready to land.

Committed as r272480 
http://lists.llvm.org/pipermail/cfe-commits/Week-of-Mon-20160606/161728.html


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv added a comment.

*ping*


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Taewook Oh via cfe-commits]

twoh added a comment.

Hmm, yes it seems not so simple to move the call to another place. If I come up 
with a better idea I'll submit it as a separate patch, as I don't want to block 
this one. Thank you for your work, @faisalv!


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=,*this]")

[Faisal Vali via cfe-commits]

I'm also a huge advocate of simple composable interfaces that do the
'one' task that they are expected (i.e. named) to do - and do it well.
I thought about this some, but perhaps not exhaustively - and settled
on being ok with getCurThisType, always returning the correct type for
 'this' as typed/referred to in code (with the right 'cv' quals) -
regardless of where it is called from.  Nevertheless, I'm open to
suggestions if you have a better alternative candidate for where the
call to adjustCVQuals should go?

Thanks for thinking about this.

Faisal Vali



On Fri, May 20, 2016 at 1:33 AM, Taewook Oh  wrote:
> twoh added a comment.
>
> @faisalv, thank you for the update. I wonder if getCurrentThisType() is the 
> best place to call adjustCVQualifiersForCXXThisWithinLambda(). It seems that 
> getCurrentThisType() does more than its name suggests. Is there any other 
> place that the adjustment function can be called?
>
>
> http://reviews.llvm.org/D19783
>
>
>
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Taewook Oh via cfe-commits]

twoh added a comment.

@faisalv, thank you for the update. I wonder if getCurrentThisType() is the 
best place to call adjustCVQualifiersForCXXThisWithinLambda(). It seems that 
getCurrentThisType() does more than its name suggests. Is there any other place 
that the adjustment function can be called?


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv added inline comments.


Comment at: lib/Parse/ParseDeclCXX.cpp:3651
@@ -3652,1 +3650,3 @@
+  case AttributeList::AT_Unused:
+return !ScopeName && AttrName->getName().equals("maybe_unused");
   default:

This whitespace change shouldn't have been included - as an aside - we still 
strive unix style line endings right?


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv updated this revision to Diff 57733.
faisalv marked an inline comment as done.
faisalv added a comment.

This patch addresses all of Richard's comments - except one (on which I'm 
awaiting some additional clarity on, before I make any changes).


http://reviews.llvm.org/D19783

Files:
  include/clang/Sema/ScopeInfo.h
  lib/Parse/ParseDeclCXX.cpp
  lib/Sema/SemaDecl.cpp
  lib/Sema/SemaExprCXX.cpp
  test/SemaCXX/cxx1z-lambda-star-this.cpp

Index: test/SemaCXX/cxx1z-lambda-star-this.cpp
===
--- test/SemaCXX/cxx1z-lambda-star-this.cpp
+++ test/SemaCXX/cxx1z-lambda-star-this.cpp
@@ -3,6 +3,8 @@
 // RUN: %clang_cc1 -std=c++1z -verify -fsyntax-only -fblocks -fms-extensions %s -DMS_EXTENSIONS
 // RUN: %clang_cc1 -std=c++1z -verify -fsyntax-only -fblocks -fdelayed-template-parsing -fms-extensions %s -DMS_EXTENSIONS -DDELAYED_TEMPLATE_PARSING
 
+template constexpr bool is_same = false;
+template constexpr bool is_same = true;
 
 namespace test_star_this {
 namespace ns1 {
@@ -69,4 +71,137 @@
   b.foo(); //expected-note{{in instantiation}}
 } // end main  
 } // end ns4
+namespace ns5 {
+
+struct X {
+  double d = 3.14;
+  X(const volatile X&);
+  void foo() {
+  
+  }
+  
+  void foo() const { //expected-note{{const}}
+
+auto L = [*this] () mutable { 
+  static_assert(is_same);
+  ++d;
+  auto M = [this] { 
+static_assert(is_same);  
+++d;
+auto N = [] {
+  static_assert(is_same); 
+};
+  };
+};
+
+auto L1 = [*this] { 
+  static_assert(is_same);
+  auto M = [this] () mutable { 
+static_assert(is_same);  
+auto N = [] {
+  static_assert(is_same); 
+};
+  };
+  auto M2 = [*this] () mutable { 
+static_assert(is_same);  
+auto N = [] {
+  static_assert(is_same); 
+};
+  };
+};
+auto L2 = [this] () mutable {
+  static_assert(is_same);  
+  ++d; //expected-error{{cannot assign}}
+};
+auto GL = [*this] (auto a) mutable {
+  static_assert(is_same);
+  ++d;
+  auto M = [this] (auto b) { 
+static_assert(is_same);  
+++d;
+auto N = [] (auto c) {
+  static_assert(is_same); 
+};
+N(3.14);
+  };
+  M("abc");
+};
+GL(3.14);
+ 
+  }
+  void foo() volatile const {
+auto L = [this] () {
+  static_assert(is_same);
+  auto M = [*this] () mutable { 
+static_assert(is_same);
+auto N = [this] {
+  static_assert(is_same);
+  auto M = [] {
+static_assert(is_same);
+  };
+};
+auto N2 = [*this] {
+  static_assert(is_same);
+};
+  };
+  auto M2 = [*this] () {
+static_assert(is_same); 
+auto N = [this] {
+  static_assert(is_same);
+};
+  };
+};
+  }
+  
+};
+
+} //end ns5
+namespace ns6 {
+struct X {
+  double d;
+  auto foo() const {
+auto L = [*this] () mutable {
+  auto M = [=] (auto a) {
+auto N = [this] {
+  ++d;
+  static_assert(is_same);
+  auto O = [*this] {
+static_assert(is_same);
+  };
+};
+N();
+static_assert(is_same);
+  };
+  return M;
+};
+return L;
+  }
+}; 
+
+int main() {
+  auto L = X{}.foo();
+  auto M = L();
+  M(3.14);
+}
+} // end ns6
+namespace ns7 {
+
+struct X {
+  double d;
+  X();
+  X(const X&); 
+  X(X&) = delete;
+  auto foo() const {
+//OK - the object used to initialize our capture is a const object and so prefers the non-deleted ctor.
+const auto & = [*this] { };
+  }
+  
+}; 
+int main() {
+  X x;
+  x.foo();
+}
+} // end ns7
+
 } //end ns test_star_this
+
Index: lib/Sema/SemaExprCXX.cpp
===
--- lib/Sema/SemaExprCXX.cpp
+++ lib/Sema/SemaExprCXX.cpp
@@ -872,6 +872,92 @@
   return false;
 }
 
+static QualType adjustCVQualifiersForCXXThisWithinLambda(
+ArrayRef FunctionScopes, QualType ThisTy,
+DeclContext *CurSemaContext, ASTContext ) {
+
+  QualType ClassType = ThisTy->getPointeeType();
+  LambdaScopeInfo *CurLSI = nullptr;
+  DeclContext *CurDC = CurSemaContext;
+
+  // Iterate through the stack of lambdas starting from the innermost lambda to
+  // the outermost lambda, checking if '*this' is ever captured by copy - since
+  // that could change 

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv marked 5 inline comments as done.
faisalv added a comment.

OK - agree (and addressed in a forthcoming patch) all your comments - except 
for the one I could use some clarity on - please see below



Comment at: lib/Sema/SemaExprCXX.cpp:910
@@ +909,3 @@
+  assert(IsFirstIteration);
+  assert(CurLSI->CallOperator->getParent()->getParent() == CurDC);
+  CurDC = CurLSI->CallOperator;

rsmith wrote:
> Please add a comment explaining this, I have no idea what special case you're 
> checking for here.
That special case should have been removed before submission - it crept in as I 
was making incremental changes to the initial patch, and getCurrentThisType was 
still being called from RebuildLambdaScopeInfo. (Alternatively, I might have 
left it in to give Richard a sense of purpose ;) 


Comment at: lib/Sema/SemaExprCXX.cpp:952
@@ +951,3 @@
+while (Closure &&
+   IsThisCaptured(Closure, IsByCopyCapture, IsConstCapture)) {
+  if (IsByCopyCapture) {

rsmith wrote:
> Should you really stop here if this is not captured? There could still be a 
> surrounding lambda with a mutable *this capture.
Do you have the following example in mind?  Or is it something different?

void foo() const {
  auto L1 = [*this] { 
  static_assert(is_same);

  auto M2 = [*this] () mutable { 
static_assert(is_same);  
auto N = [] {
  static_assert(is_same); 
};
  };
};
}


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

There are tests from the test file in my patch that don't pass(*) if
you just apply Oh's fix .  That's not surprising since Oh's patch only
meant to fix the crash, but not the 'cv' qualification issue of
'*this' that we didn't have to deal with prior to by-value captures of
'*this' and my initial implementation of feature patch
incorrectly/incompletely handled.  Also my patch attempts to simplify
the code by removing the storage of 'ThisType' in the Capture, since
it never gets queried, so no reason to store it (which was also what
was triggering the assertion initially).


(*) error: 'error' diagnostics seen but not expected:
  File F:\clang-trunk\llvm\tools\clang\test\semacxx\cxx1z-lambda-star-this.cpp
Line 88: static_assert failed
  File F:\clang-trunk\llvm\tools\clang\test\semacxx\cxx1z-lambda-star-this.cpp
Line 89: cannot assign to non-static data
 member within const member function 'foo'
  File F:\clang-trunk\llvm\tools\clang\test\semacxx\cxx1z-lambda-star-this.cpp
Line 91: static_assert failed
  File F:\clang-trunk\llvm\tools\clang\test\semacxx\cxx1z-lambda-star-this.cpp
Line 92: cannot assign to non-static data
 member within const member function 'foo'

Faisal Vali



On Wed, May 18, 2016 at 12:34 PM, Richard Smith  wrote:
> I'd also like to know whether there are cases that this patch addresses but
> Taewook Oh's patch does not, as the other patch involves a lot less
> complexity.
>
> On 18 May 2016 10:15 a.m., "Richard Smith via cfe-commits"
>  wrote:
>>
>> rsmith added inline comments.
>>
>> 
>> Comment at: lib/Sema/SemaExprCXX.cpp:898
>> @@ +897,3 @@
>> +  // end of the TU) we need to be able to examine its enclosing lambdas
>> and so
>> +  // we use the DeclContext to get a hold of the ClosureClass and query
>> it for
>> +  // capture information.  The reason we don't just resort to always
>> using the
>> 
>> No camel case for ClosureClass.
>>
>> 
>> Comment at: lib/Sema/SemaExprCXX.cpp:910
>> @@ +909,3 @@
>> +  assert(IsFirstIteration);
>> +  assert(CurLSI->CallOperator->getParent()->getParent() == CurDC);
>> +  CurDC = CurLSI->CallOperator;
>> 
>> Please add a comment explaining this, I have no idea what special case
>> you're checking for here.
>>
>> 
>> Comment at: lib/Sema/SemaExprCXX.cpp:952
>> @@ +951,3 @@
>> +while (Closure &&
>> +   IsThisCaptured(Closure, IsByCopyCapture, IsConstCapture)) {
>> +  if (IsByCopyCapture) {
>> 
>> Should you really stop here if this is not captured? There could still be
>> a surrounding lambda with a mutable *this capture.
>>
>> 
>> Comment at: lib/Sema/SemaExprCXX.cpp:1062
>> @@ -964,1 +1061,3 @@
>> +  Expr *This =
>> +  new (Context) CXXThisExpr(Loc, AdjustedThisTy, /*isImplicit*/
>> true);
>>if (ByCopy) {
>> 
>> faisalv wrote:
>> > Hmm - I wonder if instead of using AdjustedThisTy, I should always use
>> > 'ThisTy' since it is being used in the initializer expression, and should
>> > probably inherit its cv qualifiers from an enclosing lambda? correct?
>> Yes, don't use AdjustedThisTy here. You can test for this bug by giving
>> the class both a (const T&) and a (T&)=delete copy ctor.
>>
>>
>> http://reviews.llvm.org/D19783
>>
>>
>>
>> ___
>> cfe-commits mailing list
>> cfe-commits@lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=,*this]")

[Faisal Vali via cfe-commits]

OK - thanks -  will take a closer look at this hopefuly this evening
or tomorrow - and respond to Richard's other comments too.
Faisal Vali



On Wed, May 18, 2016 at 1:46 PM, Taewook Oh  wrote:
> twoh added a comment.
>
> My patch passes check-clang and the test cases in this patch as well.
>
> BTW, newly added test cases in the patch seem to be passed even without the 
> patch. Isn't the bug appears when template instantiation generates nested 
> lambdas (because template instantiation updates the 'context')?
>
>
> http://reviews.llvm.org/D19783
>
>
>
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Taewook Oh via cfe-commits]

twoh added a comment.

My patch passes check-clang and the test cases in this patch as well.

BTW, newly added test cases in the patch seem to be passed even without the 
patch. Isn't the bug appears when template instantiation generates nested 
lambdas (because template instantiation updates the 'context')?


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Richard Smith via cfe-commits]

rsmith added a subscriber: rsmith.
rsmith added a comment.

I'd also like to know whether there are cases that this patch addresses but
Taewook Oh's patch does not, as the other patch involves a lot less
complexity.


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Richard Smith via cfe-commits]

I'd also like to know whether there are cases that this patch addresses but
Taewook Oh's patch does not, as the other patch involves a lot less
complexity.
On 18 May 2016 10:15 a.m., "Richard Smith via cfe-commits" <
cfe-commits@lists.llvm.org> wrote:

> rsmith added inline comments.
>
> 
> Comment at: lib/Sema/SemaExprCXX.cpp:898
> @@ +897,3 @@
> +  // end of the TU) we need to be able to examine its enclosing lambdas
> and so
> +  // we use the DeclContext to get a hold of the ClosureClass and query
> it for
> +  // capture information.  The reason we don't just resort to always
> using the
> 
> No camel case for ClosureClass.
>
> 
> Comment at: lib/Sema/SemaExprCXX.cpp:910
> @@ +909,3 @@
> +  assert(IsFirstIteration);
> +  assert(CurLSI->CallOperator->getParent()->getParent() == CurDC);
> +  CurDC = CurLSI->CallOperator;
> 
> Please add a comment explaining this, I have no idea what special case
> you're checking for here.
>
> 
> Comment at: lib/Sema/SemaExprCXX.cpp:952
> @@ +951,3 @@
> +while (Closure &&
> +   IsThisCaptured(Closure, IsByCopyCapture, IsConstCapture)) {
> +  if (IsByCopyCapture) {
> 
> Should you really stop here if this is not captured? There could still be
> a surrounding lambda with a mutable *this capture.
>
> 
> Comment at: lib/Sema/SemaExprCXX.cpp:1062
> @@ -964,1 +1061,3 @@
> +  Expr *This =
> +  new (Context) CXXThisExpr(Loc, AdjustedThisTy, /*isImplicit*/ true);
>if (ByCopy) {
> 
> faisalv wrote:
> > Hmm - I wonder if instead of using AdjustedThisTy, I should always use
> 'ThisTy' since it is being used in the initializer expression, and should
> probably inherit its cv qualifiers from an enclosing lambda? correct?
> Yes, don't use AdjustedThisTy here. You can test for this bug by giving
> the class both a (const T&) and a (T&)=delete copy ctor.
>
>
> http://reviews.llvm.org/D19783
>
>
>
> ___
> cfe-commits mailing list
> cfe-commits@lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Richard Smith via cfe-commits]

rsmith added inline comments.


Comment at: lib/Sema/SemaExprCXX.cpp:898
@@ +897,3 @@
+  // end of the TU) we need to be able to examine its enclosing lambdas and so
+  // we use the DeclContext to get a hold of the ClosureClass and query it for
+  // capture information.  The reason we don't just resort to always using the

No camel case for ClosureClass.


Comment at: lib/Sema/SemaExprCXX.cpp:910
@@ +909,3 @@
+  assert(IsFirstIteration);
+  assert(CurLSI->CallOperator->getParent()->getParent() == CurDC);
+  CurDC = CurLSI->CallOperator;

Please add a comment explaining this, I have no idea what special case you're 
checking for here.


Comment at: lib/Sema/SemaExprCXX.cpp:952
@@ +951,3 @@
+while (Closure &&
+   IsThisCaptured(Closure, IsByCopyCapture, IsConstCapture)) {
+  if (IsByCopyCapture) {

Should you really stop here if this is not captured? There could still be a 
surrounding lambda with a mutable *this capture.


Comment at: lib/Sema/SemaExprCXX.cpp:1062
@@ -964,1 +1061,3 @@
+  Expr *This =
+  new (Context) CXXThisExpr(Loc, AdjustedThisTy, /*isImplicit*/ true);
   if (ByCopy) {

faisalv wrote:
> Hmm - I wonder if instead of using AdjustedThisTy, I should always use 
> 'ThisTy' since it is being used in the initializer expression, and should 
> probably inherit its cv qualifiers from an enclosing lambda? correct?
Yes, don't use AdjustedThisTy here. You can test for this bug by giving the 
class both a (const T&) and a (T&)=delete copy ctor.


http://reviews.llvm.org/D19783



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

Re: [PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv added a comment.

Is it too soon for a *ping* ;)

p.s. just added some comments.



Comment at: lib/Sema/SemaExprCXX.cpp:904
@@ +903,3 @@
+  for (int I = FunctionScopes.size();
+   I-- && dyn_cast(FunctionScopes[I]);
+   IsFirstIteration = false,

That dyn_cast should be an 'isa'


Comment at: lib/Sema/SemaExprCXX.cpp:1062
@@ -964,1 +1061,3 @@
+  Expr *This =
+  new (Context) CXXThisExpr(Loc, AdjustedThisTy, /*isImplicit*/ true);
   if (ByCopy) {

Hmm - I wonder if instead of using AdjustedThisTy, I should always use 'ThisTy' 
since it is being used in the initializer expression, and should probably 
inherit its cv qualifiers from an enclosing lambda? correct?


http://reviews.llvm.org/D19783



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

[PATCH] D19783: Fix cv-qualification of '*this' captures (and nasty bug PR27507 introduced by commit 263921 "Implement Lambda Capture of *this by Value as [=, *this]")

[Faisal Vali via cfe-commits]

faisalv created this revision.
faisalv added reviewers: rsmith, hubert.reinterpretcast.
faisalv added subscribers: cfe-commits, gnzlbg.

The bug report by Gonzalo (https://llvm.org/bugs/show_bug.cgi?id=27507 -- which 
results in clang crashing when generic lambdas that capture 'this' are 
instantiated in contexts where the functionscopeinfo stack is not in a reliable 
state - yet getCurrentThisType expects it to be) - unearthed some additional 
bugs in regards to maintaining proper cv qualification through 'this' when 
performing by value captures of '*this'.

This patch attempts to correct those bugs and makes the following changes:
  - when capturing 'this', we do not need to remember the type of 'this' within 
the LambdaScopeInfo's Capture - it is never really used for a this capture - so 
remove it.
  - teach getCurrentThisType to walk the stack of lambdas (even in scenarios 
where we run out of LambdaScopeInfo's such as when instantiating call 
operators) looking
for by copy captures of '*this' and resetting the type of 'this' based on 
the constness of that capturing lambda's call operator.

As an example, consider the member function 'foo' below and follow along with 
the static_asserts:

void foo() const { 

auto L = [*this] () mutable { 
  static_assert(is_same);
  ++d;
  auto M = [this] { 
static_assert(is_same);  
++d;
auto N = [] {
  static_assert(is_same); 
};
  };
};

auto L1 = [*this] { 
  static_assert(is_same);
  auto M = [this] () mutable { 
static_assert(is_same);  
auto N = [] {
  static_assert(is_same); 
};
  };
};
auto L2 = [this] () mutable {
  static_assert(is_same);  
};
auto GL = [*this] (auto a) mutable {
  static_assert(is_same);
  ++d;
  auto M = [this] (auto b) { 
static_assert(is_same);  
++d;
auto N = [] (auto c) {
  static_assert(is_same); 
};
N(3.14);
  };
  M("abc");
};
GL(3.14);
  }

Please see the test file for additional tests.

Thanks!

http://reviews.llvm.org/D19783

Files:
  include/clang/Sema/ScopeInfo.h
  lib/Sema/SemaDecl.cpp
  lib/Sema/SemaExprCXX.cpp
  test/SemaCXX/cxx1z-lambda-star-this.cpp

Index: test/SemaCXX/cxx1z-lambda-star-this.cpp
===
--- test/SemaCXX/cxx1z-lambda-star-this.cpp
+++ test/SemaCXX/cxx1z-lambda-star-this.cpp
@@ -3,6 +3,8 @@
 // RUN: %clang_cc1 -std=c++1z -verify -fsyntax-only -fblocks -fms-extensions %s -DMS_EXTENSIONS
 // RUN: %clang_cc1 -std=c++1z -verify -fsyntax-only -fblocks -fdelayed-template-parsing -fms-extensions %s -DMS_EXTENSIONS -DDELAYED_TEMPLATE_PARSING
 
+template constexpr bool is_same = false;
+template constexpr bool is_same = true;
 
 namespace test_star_this {
 namespace ns1 {
@@ -69,4 +71,112 @@
   b.foo(); //expected-note{{in instantiation}}
 } // end main  
 } // end ns4
+namespace ns5 {
+
+struct X {
+  double d = 3.14;
+  X(const volatile X&);
+  void foo() {
+  
+  }
+  
+  void foo() const { //expected-note{{const}}
+
+auto L = [*this] () mutable { 
+  static_assert(is_same);
+  ++d;
+  auto M = [this] { 
+static_assert(is_same);  
+++d;
+auto N = [] {
+  static_assert(is_same); 
+};
+  };
+};
+
+auto L1 = [*this] { 
+  static_assert(is_same);
+  auto M = [this] () mutable { 
+static_assert(is_same);  
+auto N = [] {
+  static_assert(is_same); 
+};
+  };
+};
+auto L2 = [this] () mutable {
+  static_assert(is_same);  
+  ++d; //expected-error{{cannot assign}}
+};
+auto GL = [*this] (auto a) mutable {
+  static_assert(is_same);
+  ++d;
+  auto M = [this] (auto b) { 
+static_assert(is_same);  
+++d;
+auto N = [] (auto c) {
+  static_assert(is_same); 
+};
+N(3.14);
+  };
+  M("abc");
+};
+GL(3.14);
+ 
+  }
+  void foo() volatile const {
+auto L = [this] () {
+  static_assert(is_same);
+  auto M = [*this] () mutable { 
+static_assert(is_same);
+auto N = [this] {
+  static_assert(is_same);
+  auto M = [] {
+static_assert(is_same);
+  };
+};
+auto N2 = [*this] {
+