Describe efficiently creating a Vec of zeroes

[Shnatsel]

No description provided.

[nnethercote]

I wonder if this needs to be said. How else would you intialize a Vec of zeroes?

[Shnatsel]

I've seen a developer use .resize(length, 0), realize it's too slow, change it to mem::uninitialized, and end up with a security vulnerability. I wrote an entire blog post about this.

[Shnatsel]

The standard library documentation did add a notice about it after my blog post - open the Vec docs and Ctrl+F "This may be more efficient than performing allocation and initialization in separate steps, especially when initializing a vector of zeros:"

[bobrik]

The fix for the bug you reported:

• ruuda/claxon@fec2467

It replaces unsafe code. Using std::mem::uninitialized (deprecated) and new std::mem::MaybeUninit are also unsafe.

[nnethercote]

This is a good suggestion. I'm picky about the text of the perf book so I ended up rewriting the paragraph from scratch, in 3e165fb, so there is no need to merge this PR. Thank you for the contribution!

BTW, from your blog post:

There is a tool that can detect reads from uninitialized memory, called Memory Sanitizer, but it currently doesn’t work with Rust standard library. So unless you completely avoid using Rust standard library, there is no tool that let you detect reads from uninitialized memory in Rust.

Valgrind's Memcheck tool detects unsafe uses of uninitialized memory. Also, miri might be able to as well, though I'm not sure about that.

[Shnatsel]

Yes, Miri does detect use of uninit memory as well. At the time of writing the blog post it didn't exist. Also, neither Valgrind nor Miri can be combined with fuzzing.

There are even more tools for detecting use of uninit, which are listed in the README: https://github.com/Shnatsel/libdiffuzz#see-also

Fortunately Memory Sanitizer now works well with Rust, and can be combined with fuzzing, so that's what I use these days. It makes libdiffuzz pretty much obsolete.

[Shnatsel]

Since the wording is being changed anyway, should I create issues instead of pull requests in the future?

[nnethercote]

Probably makes sense, yes.