[PATCH v2 5/6] Add a flag to BuildDeclarationNameExpr to not reject invalid decls.

[Kaelyn Takata]

On Tue, Nov 4, 2014 at 4:36 PM, David Blaikie <dblaikie at gmail.com> wrote:

> It's helpful to either use an enum (which always seems a bit silly) or a
> comment (more realistic, especially since you only have to comment that
> single caller in attemptRecovery) to describe what the boolean value is for
> ("/*AcceptInvalidDecl*/ true").
>

I've added comments for both boolean values in the call to
SemaRef.BuildDeclarationNameExpr from attemptRecovery.

>
> What's this flag needed/used for?
>

The flag is needed to get a valid DeclRefExpr during error recovery even
when the Decl isn't marked as valid. For example, with the full patch set,
not having the flag causes the last error of
test/SemaCXX/conversion-function.cpp (at line 419) to fail to be
typo-corrected. With the flag or without this patchset:

~/llvm/tools/clang/test/SemaCXX/conversion-function.cpp:419:18: error: no
member named 'e' in 'PR18234::A'; did you mean simply 'e'?
  bool k1 = e == A::e; // expected-error {{no member named 'e'}}
                 ^~~~
                 e
~/llvm/tools/clang/test/SemaCXX/conversion-function.cpp:418:8: note: 'e'
declared here
  A::E e = a; // expected-note {{here}}
       ^
4 warnings and 26 errors generated.

With this patchset but without this flag:

~/llvm/tools/clang/test/SemaCXX/conversion-function.cpp:419:21: error: no
member named 'e' in 'PR18234::A'
  bool k1 = e == A::e; // expected-error {{no member named 'e'}}
                 ~~~^
4 warnings and 26 errors generated.


> On Wed, Oct 29, 2014 at 12:49 PM, Kaelyn Takata <rikka at google.com> wrote:
>
>> ---
>>  include/clang/Sema/Sema.h |  6 ++++--
>>  lib/Sema/SemaExpr.cpp     | 15 ++++++++-------
>>  lib/Sema/SemaExprCXX.cpp  |  2 +-
>>  3 files changed, 13 insertions(+), 10 deletions(-)
>>
>>
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20141105/ea7fa48c/attachment.html>