[Bug tree-optimization/116079] [15 Regression] ice during GIMPLE pass lim

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116079

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
So the main issue is that we are not updating the store motion pass meta-data
after applying store-motion to an outer loop and then continue to analyze
inner loops.  In particular the somewhat recent addition to determine if
we are dealing with a single store BB.  But now as we not only rewrite stmts
but remove them this becomes an issue - but we can get away with this by
clearing the VDEF on such stmt.