[Bug rtl-optimization/93996] [10 Regression] ICE in lookup_page_table_entry

[law at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93996

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at redhat dot com

--- Comment #8 from Jeffrey A. Law <law at redhat dot com> ---
That code was part of the fix for pr41883 and I doubt the change is right
as-is.

The questions I'd have:

1. The code clearly assumes that after the EPILOGUE_BEG we have a real insn. 
Why is that assumption not true here.

2. Assuming the reason for #1 is valid, where should the REG_SAVE_NOTE be
placed to keep the dwarf2 emitter happy.  Blindly going to the next real insn
may not be correct.  Consider if the EPILOGUE_BEG note is the same as TAIL.  In
the case where we "just" cross into another block, is placing the note on an
insn in another block correct?