On Oct 29, 2012, at 17:48 , Anna Zaks <[email protected]> wrote:
>
> On Oct 29, 2012, at 4:08 PM, Jordan Rose wrote:
>
>>
>> On Oct 29, 2012, at 15:51 , Anna Zaks <[email protected]> wrote:
>>
>>> Author: zaks
>>> Date: Mon Oct 29 17:51:44 2012
>>> New Revision: 166975
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=166975&view=rev
>>> Log:
>>> [analyzer] Add checker helpers to CheckerContext.
>>>
>>> - Adding Immutable Map to GDM and getIdentifierInfo helper method.
>>>
>>> Modified:
>>> cfe/trunk/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h
>>> cfe/trunk/lib/StaticAnalyzer/Checkers/StreamChecker.cpp
>>>
>>> Modified:
>>> cfe/trunk/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h?rev=166975&r1=166974&r2=166975&view=diff
>>> ==============================================================================
>>> ---
>>> cfe/trunk/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h
>>> (original)
>>> +++
>>> cfe/trunk/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h
>>> Mon Oct 29 17:51:44 2012
>>> @@ -16,6 +16,21 @@
>>> #define LLVM_CLANG_SA_CORE_PATHSENSITIVE_CHECKERCONTEXT
>>>
>>> #include "clang/StaticAnalyzer/Core/PathSensitive/ExprEngine.h"
>>> +#include "llvm/ADT/ImmutableMap.h"
>>> +
>>> +// Declare an immutable map suitable for placement into program states's
>>> GDM.
>>
>> Doxygen comment? Also, it'd be nice to mention that this must not be used
>> inside any namespaces.
> will commit later with the macro rename
Also, I don't think CheckerContext is the right file for this, since it's
perfectly usable by Core. (DynamicDispatchBifurcationMap?) ProgramStateTrait.h?
>>> +#define REGISTER_MAP_WITH_GDM(Map, Key, Value) \
>>> + typedef llvm::ImmutableMap<Key, Value> Map; \
>>
>> This is not correct; two maps with the same keys and values will share a GDM
>> entry this way. This could easily happen with maps to bool, for example. No
>> matter what the template argument for ProgramStateTrait must be a unique
>> type. (…which should then probably be in an anonymous namespace.)
>>
> That right! Thanks for catching.
>>
>>> + namespace clang { \
>>> + namespace ento { \
>>> + template <> \
>>> + struct ProgramStateTrait<Map> \
>>> + : public ProgramStatePartialTrait<Map> { \
>>> + static void *GDMIndex() { static int Index; return &Index; } \
>>
>> I think ideally we'd rewrite the accessor in ProgramState to be templated
>> here, like how CheckerManager generates unique tags for checkers. That way
>> even custom implementers won't have to include the GDMIndex boilerplate.
>>
>>> + }; \
>>> + } \
>>> + }
>>> +
>>
>>
>>
>
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
