[Bug rtl-optimization/70460] [4.9/5/6 Regression] Miscompilation of glibc on i686-linux starting with r234101

[bernds at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70460

--- Comment #4 from Bernd Schmidt <bernds at gcc dot gnu.org> ---
Can we change it to just use SET_SRC (single_set (...))?

I don't think the new pass was intended to handle anything new, only to
disentangle the logic in ira. Alan mentioned gfortran.dg/pr46755.f as the only
preexisting testcase where the label replacement triggers a
delete_unreachable_blocks call.