[cfe-dev] Thread Safety for C structs RFC.

[Ethan Jackson]

All of this sounds good to me.  I'll write up a patch and send it out
this week.  Thanks for being so responsive.

Ethan

On Sun, Jul 14, 2013 at 11:51 PM, David Blaikie <dblaikie at gmail.com> wrote:
> On Sat, Jul 13, 2013 at 8:29 PM, Ethan Jackson <ethan at nicira.com> wrote:
>> I'm in the process of evaluating using clang for the Open vSwitch project with
>> hopes of taking advantage of its thread safety annotations.  Unfortunately, on
>> tot these annotations only work in C++ because it's only possible to declare C++
>> classes and structs as "lockable".  This simple patch appears to fix the issue
>> (in my testing), but it's so trivial that I'm wondering if I've overlooked
>> something.  I don't know clang well, so any advice in this area would be
>> appreciated.
>>
>> Thanks,
>> Ethan Jackson
>>
>> ---
>>  lib/Sema/SemaDeclAttr.cpp |    3 +--
>>  1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/lib/Sema/SemaDeclAttr.cpp b/lib/Sema/SemaDeclAttr.cpp
>> index 4c18a33..dc1c75d 100644
>> --- a/lib/Sema/SemaDeclAttr.cpp
>> +++ b/lib/Sema/SemaDeclAttr.cpp
>> @@ -584,8 +584,7 @@ static bool checkLockableAttrCommon(Sema &S, Decl *D,
>>    if (!checkAttributeNumArgs(S, Attr, 0))
>>      return false;
>>
>> -  // FIXME: Lockable structs for C code.
>> -  if (!isa<CXXRecordDecl>(D)) {
>> +  if (!isa<CXXRecordDecl>(D) && !isa<RecordDecl>(D)) {
>
> Before this is committed, though - you can just replace this with "if
> (!isa<RecordDecl>(D))" I hope (please test/try this, of course), since
> all CXXRecordDecls are RecordDecls. (& as Delesley mentioned - a test
> case will be appropriate/necessary)
>
>>      S.Diag(Attr.getLoc(), diag::warn_thread_attribute_wrong_decl_type)
>>        << Attr.getName() << ThreadExpectedClassOrStruct;
>>      return false;
>> --
>> 1.7.9.5
>>
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev