r214333 - Add a state variable to the loop hint attribute.

[Aaron Ballman]

On Fri, Aug 1, 2014 at 8:46 AM, Kostya Serebryany <kcc at google.com> wrote:
>
>
>
> On Fri, Aug 1, 2014 at 4:42 PM, Aaron Ballman <aaron at aaronballman.com>
> wrote:
>>
>> On Fri, Aug 1, 2014 at 8:38 AM, Kostya Serebryany <kcc at google.com> wrote:
>> >>
>> >> I don't have access to either a valgrind or a msan build, but I took a
>> >
>> > For msan you do need a separate build, but for valgrind your just run
>> > valgrind on a regular build.
>>
>> Valgrind doesn't work on Windows; I'm uncertain whether msan does or not?
>
> Nope. On Windows  your are on your own against uninits (unless you want to
> try DrMemory)
>>
>>
>> >>
>> >> stab in the dark with r214502. I'll see if the msan bot thinks I've
>> >
>> > Nope, the report is still there.
>>
>> Drat! I'll do a bit more investigating...
>
>
> valgrind with --track-origins=yes gives this:
> ==3308==  Uninitialised value was created by a heap allocation
> ==3308==    at 0x402F7C4: malloc
> (valgrind/coregrind/m_replacemalloc/vg_replace_malloc.c:270)
> ==3308==    by 0x25CEAEA: clang::Preprocessor::AllocateMacroInfo()
> ==3308==    by 0x25CEC3C:
> clang::Preprocessor::AllocateMacroInfo(clang::SourceLocation)
> ==3308==    by 0x25D312A:
> clang::Preprocessor::HandleDefineDirective(clang::Token&, bool)
> ==3308==    by 0x25D8E58:
> clang::Preprocessor::HandleDirective(clang::Token&)
> ...

Yes, I found another code path where this wasn't getting initialized.
Perhaps 214504 fixes this?

(Thank you for your help with tracking this down!)

~Aaron