Adding this here, because I know I'll never be able to find this again in my webmail....
> I didn't hear about this possibility before
> Regardless , Is that something I can do during the runtime or we have
> to recompile epengine with that option ?
> If you have the binary compiled with that option I can run some tests
> on it today to confirm standard malloc itself is sufficient
I believe all you need to do is to export two variables at runtime:
> On Feb 17, 2012, at 11:32 PM, Trond Norbye <email@example.com> wrote:
>> Did we ever try running the test with the standard malloc and using the FORCE_NEW flag? I've asked this question a number of times now, but I haven't gotten an answer yet. Replacing the standard memory allocator with tcmalloc didn't solve the problem itself, so it seemed that it root of the problem was the memory allocator inside the C++ layer.
>> From the experiments we've had during the integration it didn't feel that the windows support was that mature and well tested. It feels a bit "risky business" to me to give a version to a customer where we change such a central component on platform with so little testing. I would personally sleep way better at night using Microsofts memory allocator if it solves the problem.
>> just my 0.5 nok