[libc] Suspected strict aliasing violation in string_length_wide_read

[Voultapher]

The following code in libc:

llvm-project/libc/src/string/string_utils.h

Line 65 in b41240b

for (; reinterpret_cast<uintptr_t>(char_ptr) % sizeof(Word) != 0;

invokes what I understand to be strict aliasing UB. I think block reads would have to go through memcpy/__builtin_memcpy which can then be optimized by the compiler as appropriate.

Tagging @michaelrj-google.

[michaelrj-google]

Thanks for filing this, I can look into it after the holidays. If someone else want to look into it before I get to it feel free.

[nickdesaulniers]

Thanks for the report. Was this something that was actually flagged by UB? Or spotted manually? I assume you meant memcmp (not memcpy)?

[Voultapher]

@nickdesaulniers spotted manually. No I mean memcpy not that line itself but the content would have to be copied into a stack buffer via memcpy to my understanding that's required to avoid strict aliasing violations.

[Voultapher]

For a possible fix this could serve as inspiration 850ca5a.