Add fuzzing for fast_memcpy

[Copilot]

This unsafe memory operation crate had only indirect coverage via the sparse_mmap fuzzer. This adds direct fuzzing for its public APIs.

Changes

• fast_memcpy fuzzer: Tests memcpy with separate buffers and memmove with overlapping regions, validated against reference implementation

• Helper function: Extracted safe_copy_length() to ensure consistent bounds checking across operations

The fuzzer follows existing patterns from other fuzz targets in the repository (e.g., ucs2, sparse_mmap).

Example

The fast_memcpy fuzzer validates correctness:

// memcpy: separate buffers (undefined behavior if overlapping)
unsafe { fast_memcpy::memcpy(dest, src, len) };
assert_eq!(dest[..len], src[..len]);

// memmove: same buffer, validate against reference
unsafe { fast_memcpy::memmove(buf.add(dest_off), buf.add(src_off), len) };
assert_eq!(buf, reference_buf);  // reference computed with copy_within

Fixes #2348

Original prompt

---

This section details on the original issue you should resolve

<issue_title>fuzz trycopy and fast_memcpy</issue_title>
<issue_description>These crates already have significant usage by the rest of the codebase, and trycopy has some fuzz coverage provided by the sparse_mmap fuzzer, but we should fuzz these unsafe crates directly.</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes fuzz trycopy and fast_memcpy #2348

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[benhillis]

This might be above Copilot's paygrade, I'm just curious what it will come up with.

[smmalis37]

Probably want to try doing them separately so it doesn't get confused.

[benhillis]

@copilot - let's make this task more simple. Just implement fuzzing for fast_memcpy. Look at other examples of fuzzing done in the repo to make sure you are using a similar approach.

[Copilot]

@copilot - let's make this task more simple. Just implement fuzzing for fast_memcpy. Look at other examples of fuzzing done in the repo to make sure you are using a similar approach.

Simplified the PR to only include fast_memcpy fuzzing. Removed the trycopy fuzzer and kept the implementation consistent with other fuzzers in the repo (like ucs2 and sparse_mmap). The fuzzer tests both memcpy (with separate buffers) and memmove (with overlapping regions), validated against a reference implementation. Commit: 08b5f97