Lints for target_* cfg

[est31]

Inspired by seeing #85738 fixing a typo by replacing target_os = "opensbd" with target_os = "openbsd".

I wonder whether cases like this could be discovered by lints. There are multiple target_* values that rustc sets, and it would be interesting to think about validating them.

Rust has support for custom target json files, and ideally you would want to have older rustc versions not warn about code supporting targets added by future rustc versoins, making a warn-by-default lint that complains about any non-builtin target_* value a bit tricky.

However, one could think about an allow-by-default non_builtin_target_cfg lint, as well as a warn-by-default target_cfg_typo lint that checks for targets in a close edit distance to a builtin target. If you enable the non_builtin_target_cfg lint, maybe the target_cfg_typo could be silenced.

As prior art I was only able to find a clippy lint.

[est31]

@rustbot label A-lint

[est31]

Running a cursory check to find cfg typos across the rust codebase, I couldn't find any:

$ rg -oI 'target_[a-z]* = "[a-z]*"' | sed 's/.* = "\([a-z]*\)".*/\1/' | sort | uniq -c | sort -rn
   3669 arm
    298 linux
    181 android
    164 macos
    161 emscripten
    120 little
    108 freebsd
    104 windows
    100 vxworks
     81 netbsd
     80 ios
     77 openbsd
     74 dragonfly
     61 fuchsia
     58 powerpc
     55 solaris
     53 redox
     53 msvc
     47 sgx
     47 gnu
     42 mips
     42 illumos
     32 haiku
     28 big
     23 musl
     23 mclass
     21 uwp
     17 fortanix
     14 hermit
     14 atomics
     13 wasi
     12 none
     12 asmjs
      9 unix
      8 x
      7 sse
      6 vsx
      6 neon
      6 hexagon
      6 cloudabi
      6 altivec
      5 newlib
      4 sparc
      4 pmull
      4 nvptx
      4 msa
      4 freertos
      4 dsp
      4 avx
      3 uclibc
      3 libnx
      3 
      2 xsaves
      2 xsaveopt
      2 xsavec
      2 xsave
      2 watchos
      2 unknown
      2 tvos
      2 tsc
      2 tbm
      2 sve
      2 sha
      2 rtm
      2 rdseed
      2 rdrand
      2 rdm
      2 rcpc
      2 psp
      2 popcnt
      2 pclmulqdq
      2 mmx
      2 lzcnt
      2 lse
      2 fxsr
      2 fp
      2 fma
      2 eabihf
      2 dotprod
      2 crypto
      2 crc
      2 apple
      2 aes
      2 adx
      2 abm
      1 spirv
      1 avr

Maybe it's not that valuable to add this if this doesn't occur in the wild very often?

[est31]

cc #82450

[est31]

Yeah I think this can be closed in favour of #82450