https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82079
--- Comment #2 from Martin Sebor <msebor at gcc dot gnu.org> --- I forgot that pointer offsets are treated internally as signed even if they are unsigned in the source code. That seems like an important detail for the new option to document. Although to be honest, even regardless of the signedness issue, I'm having a hard time coming up with a use case for this sort of instrumentation. When is it useful to only have pointer arithmetic overflow diagnosed without diagnosing all expressions that form out-of-bounds/invalid pointers from valid ones?
