[Bug rtl-optimization/57915] [4.8/4.9 Regression] ICE in set_address_disp, at rtlanal.c:5537

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57915

--- Comment #9 from Andrey Belevantsev <abel at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #7)
> I'd say if (plus:SI (symbol_ref:SI ...) (const_int ...)) not surrounded by
> (const:SI ) is considered invalid IL, then trying to workaround cprop bug in
> some other pass is still workaround rather than fix.

Sure, I agree that simplifying when inside cprop is the proper fix, though I
wasn't sure which of the general interfaces to fix.  The question of why fwprop
does not simplify this particular insn, i.e. whether the limitation of fwprop
working within loops makes sense, still remains IMHO, but this is a separate
issue.