JustinStitt wrote:
My original idea was to get the SIO sanitizer working with `-fwrapv`, the issue
[here](https://github.com/KSPP/linux/issues/26) even suggests it as a viable
option. However, after seeing literal checks like:
```cpp
case LangOptions::SOB_Undefined:
if (!CGF.SanOpts.has(SanitizerKind::SignedIntegerOverflow))
```
... I think the best option is to instrument a wrap sanitizer (which this PR
does). A wrap sanitizer best captures the language semantics at hand while
maintaining existing functionality of the SIO sanitizer. I think Kees can speak
first hand about how picky some folks are about the language being used for
this arithmetic overflow/wraparound stuff (he linked some gcc threads above my
comment but I've also seen some spicy LKML discussions about how OVERFLOW
doesn't exist in the Linux Kernel and as such WRAPAROUND instrumentation is
needed).
I think this PR bridges the gap between folks like Kees (who just want all this
suspicious kernel arithmetic to go away) and folks like Linus (who is really
particular about language).
https://github.com/llvm/llvm-project/pull/80089
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
