Remove 4146 from libunwind

[AaronRobinsonMSFT]

Contributes to #66154

This converts the invalid arithmetic negation on an unsigned value to use the identity of bitwise negation plus one.

Upstream PR: libunwind/libunwind#333

/cc @GrabYourPitchforks @jkotas @janvorli @am11

[ghost]

Tagging subscribers to this area: @dotnet/runtime-infrastructure
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Contributes to #66154

This converts the invalid arithmetic negation on an unsigned value to use the identity of bitwise negation plus one.

Upstream PR: libunwind/libunwind#333

/cc @GrabYourPitchforks @jkotas @janvorli @am11

Author:	AaronRobinsonMSFT
Assignees:	-
Labels:	area-Infrastructure
Milestone:	7.0.0

[GrabYourPitchforks]

Just out of curiosity, do you prefer ~a + 1 or (size_t)(-(ssize_t)a)? The former is more compact, but I wonder if the latter is easier to understand. But if you have an existing preferred pattern then carry on. :)

[AaronRobinsonMSFT]

Just out of curiosity, do you prefer ~a + 1 or (size_t)(-(ssize_t)a)?

@GrabYourPitchforks That is a fair point. I try to avoid casting when possible as it feels like an escape hatch for doing something I shouldn't be doing. If you think the latter would be preferable, I am fine changing it to that.

[jkotas]

~a + 1

Nit: This may be less efficient than the original code (depends on whether the compiler knows how to optimize back to -a).

I believe that the most common pattern for these situations is x & ~(sizeof(Y) - 1). We use it e.g. here:

runtime/src/coreclr/inc/clrtypes.h

Line 322 in d94ca57

return (value+alignment-1)&~(alignment-1);

It really depends on what the libunwind maintainer prefers...

[AaronRobinsonMSFT]

@jkotas @GrabYourPitchforks One of the maintainers of libunwind responded and indicated they don't have a strong preference. This would seem to imply we can do what we think is best. See the reply and options at libunwind/libunwind#333 (comment).