[PATCH] D42171: X86CallFrameOptimization: Bail on win64cc calls

[Zvi Rackover via Phabricator via llvm-commits]

zvi added a comment.

In https://reviews.llvm.org/D42171#978790, @rnk wrote:

> Isn't this MI buggy? We're adjusting SP down by 40 bytes and storing to SP+48, which could overwrite data. I think the assert is valid.


There are a couple of troubling issues seen here:

1. The caller is corrupting its own return address
2. byval is not honored - shouldn't the structs that are passed 'byval' be passed as copies?

This patch does not address these issues.


Repository:
  rL LLVM

https://reviews.llvm.org/D42171