https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117360
--- Comment #3 from Jeffrey A. Law <law at gcc dot gnu.org> --- What's interesting is I did a bootstrap with ubsan a while back to chase down this stuff. Could be something recently introduced or a path we didn't trigger before. Regardless I'll case it down. Yes, we need to change some of those constants to HOST_WIDE_INT_UC and such. I think Andrew pointed that out a little while ago. Finally, yes, checking the shift constants for validity is done elsewhere in ext-dce to catch problems early. Those checks probably aren't as pervasive/consistent as they need to be.
