Daniel Dickman <[email protected]> wrote: > I’m not totally sure. I think we’d need to check if the callers are direct or > indirect.
No, that's not how this works. > If they are all direct then I think this diff may introduce new landing pads > that weren’t there before and it broadens the BTI target set unnecessarily. That risk is less severe than marking the entire binary incapable of using BTI. > I suggest that should be checked first. What do you think? Disagree. Sometimes a compiler can make an accurate decision. But manually generated entrypoints which has failed once, need bti.
