threads and contexts and allocations and frees

Nathaniel McCallum npmccallum at redhat.com
Tue Apr 23 08:47:53 EDT 2013


On Wed, 2013-04-17 at 14:38 -0700, Russ Allbery wrote:
> Greg Hudson <ghudson at MIT.EDU> writes:
> > On 04/17/2013 04:49 PM, Russ Allbery wrote:
> 
> >> IIRC, this sort of thing can cause problems on Windows due to the way
> >> that the allocators work (although possibly doesn't at the moment
> >> depending on how the code is structured).
> 
> > On Windows, you want malloc() and free() calls to be made from the same
> > compiled code unit, because the allocator memory layout depends on
> > compilation flags (as I understand it).  But that's not what "this sort
> > of thing" is.  I don't think Chris Hecker is proposing to mix
> > library-malloc with application-free or vice versa.
> 
> I had vaguely remembered that Windows might use separate per-thread malloc
> arenas in some situations, but some brief searching and research shows
> that I probably made that up.  Apologies for the noise.

They have per-binary malloc arenas. But that isn't a concern here.

Nathaniel




More information about the krbdev mailing list