I'm not sure if this is a good idea because the closure size of racer will be a lot higher, and lots of people probably aren't using nixpkgs' rust anyways, so the source won't be accurate. Can you do a closure size measurement?

lots of people probably aren't using nixpkgs' rust anyways, so the source won't be accurate.

I don't think this would be a problem -- /usr/local/src/rust/src is a fallback path and normally the proper path would be obtained with rustc --print sysroot which, I guess, works perfectly fine for those who installed rustc via rustup despite the hardcoded path in racer-mode

(defcustom racer-rust-src-path
  (or
   (getenv "RUST_SRC_PATH")
   (when (executable-find "rustc")
     (let* ((sysroot (s-trim-right
                      (shell-command-to-string
                       (format "%s --print sysroot" (executable-find "rustc")))))
            (src-path (f-join sysroot "lib/rustlib/src/rust/src")))
       (when (file-exists-p src-path)
         src-path)
       src-path))
   "/usr/local/src/rust/src")

Can you do a closure size measurement?

/nix/store/4fqbdg0d5jagndzkjhns6j6kppwzv08a-emacs-racer-20181023.1604     385828576
/nix/store/xqdaf1zm726h1031zb6zdv9sa9ns74q5-emacs-racer-20181023.1604     641303584

The former is the nix path-info -S for ./result without the patch and the latter -- with one.

Hmm I see. Alright I think this is fine then, even though it does increase closure size considerably. Currently probably most people won't take advantage of this, as everybody either uses rustup or has RUST_SRC_PATH set, but for everybody that installs it for the first time, this will be a nice change.