[clang] [lldb] [Clang] Introduce OverflowBehaviorType for fine-grained overflow control (PR #148914)

[Kees Cook via cfe-commits]

kees wrote:

> You would only get a trap when you get a narrowing conversion, maybe that should only be the case for _implicit_ conversions? e.g `(int)some_obt_qualified_64bit_type` - I'm not sure I like that as to me it feels like the idiomatic casting behaviour would silently truncate, presumably unexpectedly, e.g the default developer response to an implicit narrowing warning is just to add `(dest_type)expr` which would silently drop high bits even though that would seem unexpected?

Right this gets to my concern about casts: I don't want a way to return to ambiguous values.

> It seems not unreasonable to maybe have something like `__obt_ignore` qualifier in casts to make it easier to perform a narrowing cast that drops instrumentation/overflow checks as a usability improvement that makes it explicit that you do not care about overflow.

I want to have overflow behavior be explicit, so I think if we wanted to "strip" an `__ob_trap`, it should get an `__ob_wrap` cast: `(dest_type __ob_wrap)expr` ?

> @kees @JustinStitt Are you folk going to be at the US devmeeting? Maybe it will be easier to go over in person? If not it might be easier to set up a telcon or something?

Yeah, @JustinStitt will be there for sure; I can probably fly down or we can do a video chat. Email me at kees at kernel.org and let's take this offline to coordinate?



https://github.com/llvm/llvm-project/pull/148914