On 01/31/2018 03:41 AM, Uros Bizjak wrote: >> >> The one concern I have is do we need to tell the CFI machinery that >> %eax's value was restored to its entry value? > > Can you or someone that knows CFI stuff please investigate this a bit? > I'm not expert in this area, and I don't feel comfortable to approve > the patch that has some known loose edges in the area I don't know > that well. Fair enough. I'm not terribly familiar with the CFI bits either, but I was able to work this out.
I've looked at gcc-6, gcc-7 and the trunk. None of them record anything WRT %eax at the restore point. If the CFA is the stack pointer both record that the CFA offset changed, but nothing about %eax. So this patch doesn't change change the accuracy of the CFI info. One might be able to argue that we should record the restores, but that seems like an independent issue to me. I'll post an updated patch momentarily. jeff
