------- Comment #86 from ian at airs dot com 2007-05-18 17:24 ------- Re comment #80, comment #81, comment #82. My patch handles the placement new in comment #73 to indicate an alias between double and long. The mis-ordered code is actually aliasing int and long. That aliasing is due to a cast to (int*). There is no placement new in that cast. There is no reason for the compiler to avoid exchanging the store via int* and the store via long*. Why should it? If you omitted the placement new, this would be a standard aliasing violation between int and double. With the placement new, it's still an aliasing violation between int and long. The placement new doesn't excuse you from all aliasing violations: it just excuses you from aliasing violations involving the actual types in question.
If you want to make a principled argument that placement new excuses you from all aliasing violations, then again the result is that in a function which uses placement new we should disable TBAA for that function. That would be OK with me and easy to implement. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29286
