Comment 3 for bug 1440186

I haven't been able to replicate the crash. It's a race condition, so I'm likely not getting the timing "right". I'm not too worried by this.

I couldn't find any authoritative statement about the alignment guarantees of "new char[]", so I'm attaching a modified patch. It replaces new[] and reinterpret_cast<>() with C-style allocation/casting. It also removes the "magic".

I'm also wondering about bounding the length of this free list, or for that matter replacing it with std::vector<char*>. However, these changes are too big for a 3.14.x patch.