https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70690
--- Comment #3 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> I think the new GCC is correct here and the old one accidentally just got it
> "right".
> We have:
> CentralFreeListPadded Static::central_cache_[kNumClasses];
>
>
> Which is of type:
> class CentralFreeListPadded : public CentralFreeListPaddedTo<
> sizeof(CentralFreeList) % 64>
>
> CentralFreeListPaddedTo is defined as either:
> class CentralFreeListPaddedTo : public CentralFreeList
>
> or
> class CentralFreeListPaddedTo<0> : public CentralFreeList
>
> CentralFreeList has the following constructor:
> CentralFreeList() : lock_(base::LINKER_INITIALIZED) { }
>
>
> So it just happened that LINKER_INITIALIZED was also zero:
> namespace base {
> enum LinkerInitialized { LINKER_INITIALIZED };
> }
>
>
> So this is a bug in tcmalloc as far as I can tell.
>
> Before the call to the constructor was not done but after it is done.
Could you come up with a small testcase?
