[Bug target/84064] ICE in ix86_expand_prologue related to -fstack-clash-protection and memcpy on i686

[law at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84064

--- Comment #2 from Jeffrey A. Law <law at redhat dot com> ---
I wasn't ever happy with the discrepancy between the computation of TO_ALLOCATE
in the layout code and ALLOCATE within ix86_expand_prologue.  It seemed ripe to
fall into this kind of problem.  sigh.


If ALLOCATE was that small we don't need to do any probing unless the current
function is marked as noreturn.

I wonder if deferring the assertion to the point where we actually need the
scratch or are going to call the msft probe function is possible.