https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92486
--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> --- While I don't share this preference about assignments, SRA only ignores padding when doing "total scalarization" aka the poor man's aggregate copy propagation because in that mode it would have to invent small accesses to copy padding and I'm sure some people would be unhappy about those.
