On win 64, I have some code that consistently gives a segmentation fault with the current trunk fpc under -O4, and debugging suggests bad code generation. On Linux it works fine. I only intermittently compile with fpc on windows; stepping back through the revisions indicated that 43384 was the change that caused it to start generating bad code. (I test trunk regularly on Linux, and haven't seen this problem.) It is not obvious how to generate a simple test case from my code - any suggestions as to how to resolve this?

Colin