> Nim:
> <https://github.com/jrfondren/chrestomathy/blob/master/pwcheck/nim/src/pwcheck.nim#L18>
> \-- crypt() is useful as-is
Let's see...
proc crypt(phrase, salt: cstring): cstring {.cdecl, importc, dynlib:
"libcrypt.so".}
Run
It's not useful without a `cast`, which we've already established is an unsafe
operator:
proc crypt(phrase, salt: cstring): cstring {.cdecl, importc, dynlib:
"libcrypt.so".}
echo "encrypted = " & crypt("x","y")
--> NimRunner.nim(2, 21) Error: type mismatch: got <string, cstring>
Run
It's not safe, either:
proc crypt(phrase, salt: cstring): cstring {.cdecl, importc, dynlib:
"libcrypt.so".}
echo crypt(nil, nil)
--> SIGSEGV: Illegal storage access. (Attempt to read from nil?)
Run
I've wrapped a few hundred functions by now, and I'd say most of them are
absolutely _not_ safe, mostly because they (a) blow up when passed `nil`, or
(b) require manual freeing of objects, which means use-after-free bugs are
possible.
Nim's C FFI is very nice and convenient, but it does not magically make C APIs
safe. And it doesn't magically make them idiomatic Nim, either; a really
well-designed C API can come close, but most are far from that level.
Magical thinking isn't any better than a cargo cult.
