Fix some UBSan errors

[fvrmatteo]

Opened by mistake, I'm currently doing tests on my fork. Although it might be useful to confirm if these UBSan errors are legitimate.

[marjevan]

Hi,
If this is a mistake, please close the PR.
Thank you!

[fvrmatteo]

Hi @marjevan!

The PR was opened by mistake, but I now confirmed that the UBSan errors are legitimate (due to type punning and strict aliasing being violated) and that this PR fixes them.

I see that the issue was acknowledged in 2021, but it doesn't seem to have been addressed since then:

# undefined behavior sanitizer
#
# 2021-05-13: xed has several locations where it dereferences
#  unaligned pointers for 16/32/64 displacements & immediates. The
#  code generated by clang -O3 is worse for these cases when using
#  legal C. Thinking about how to handle that.

The usual approach to handle this case is to use std::memcpy and let the compiler recognise the idiom and lower it into a direct memory access with proper alias information associated to the load instruction (more details here: https://gist.github.com/shafik/848ae25ee209f698763cffee272a58f8).

Please let me know if we want to proceed with fixing these type punning issues (e.g. if someone wants to take a look at the issue over the whole codebase, since it was officially acknowledged in the past) or if we want to keep ignoring it (since I also see xed_mbuild.py is purposefully disabling the UBSan sanitiser).

[marjevan]

Thank you for reaching out! We're constantly working to improve the XED library and its code quality, and we greatly value contributions from our user community.
If you're interested in contributing, please refer to our contribution guide at the following link.

[fvrmatteo]

Hello @marjevan, I read the contribution guide and signed-off my commit.