std depends on cfg-if 1.0.0 and 0.1.10

[chbaker0]

#101861 added cfg-if v1.0.0 as a std dependency.

std still depends on cfg-if v0.1.10:

$ cargo tree -i -p cfg-if@0.1.10
cfg-if v0.1.10
├── getrandom v0.1.14
│   ├── rand v0.7.3
│   │   [dev-dependencies]
│   │   └── std v0.0.0 (/usr/local/google/home/collinbaker/chromium/src/third_party/rust_src/src/library/std)
│   └── rand_core v0.5.1
│       ├── rand v0.7.3 (*)
│       └── rand_chacha v0.2.2
│           └── rand v0.7.3 (*)
├── panic_abort v0.0.0 (/usr/local/google/home/collinbaker/chromium/src/third_party/rust_src/src/library/panic_abort)
│   └── std v0.0.0 (/usr/local/google/home/collinbaker/chromium/src/third_party/rust_src/src/library/std)
└── unwind v0.0.0 (/usr/local/google/home/collinbaker/chromium/src/third_party/rust_src/src/library/unwind)
    └── std v0.0.0 (/usr/local/google/home/collinbaker/chromium/src/third_party/rust_src/src/library/std)

Depending on only one version would be nice. While it's a small library, it's still unnecessary bloat.

[chbaker0]

@rustbot label T-libs

[thomcc]

Note that this bloat is only exposed to users of -Zbuild-std so it doesn't matter much. I'd take a patch removing it from panic_abort and unwind though.

getrandom should only be used as a dev-dep, and also we can't easily update it for various unfortunate reasons, so I'm not sure how much I'd recommend pushing on that.

[chbaker0]

Note that this bloat is only exposed to users of -Zbuild-std so it doesn't matter much. I'd take a patch removing it from panic_abort and unwind though.

not exactly true, it's shipped in the normal distribution:

$ ls .rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/x86_64-unknown-linux-gnu/lib
libaddr2line-74211b12b443811c.rlib          libmemchr-e89379089032cd0e.rlib             librustc-nightly_rt.tsan.a
libadler-ce60aba693c46806.rlib              libminiz_oxide-6d1ec8f1337db26d.rlib        librustc_std_workspace_alloc-b113a1d1429de7b6.rlib
liballoc-c6c03e024a2f1e46.rlib              libobject-4a45f4075ae80a22.rlib             librustc_std_workspace_core-522518611024dce5.rlib
libcfg_if-a51843dfc5fc8b4b.rlib             libpanic_abort-07057e0406e64bc5.rlib        librustc_std_workspace_std-013cf041d0880b37.rlib
libcfg_if-fcf994c37af81dc5.rlib             libpanic_unwind-15accacd1251d941.rlib       libstd-7b2106000b625742.rlib
libcompiler_builtins-b7c79d85cf21a511.rlib  libproc_macro-80378fa4c5002a0d.rlib         libstd-7b2106000b625742.so
libcore-05898138a596088a.rlib               libprofiler_builtins-f931f59836842a86.rlib  libstd_detect-bb8d1566a1046906.rlib
libgetopts-55c2848d703ec364.rlib            librustc_demangle-db4d651637ed1365.rlib     libtest-90e69e1db733a8fb.rlib
libgimli-1f391b051eb3b957.rlib              librustc-nightly_rt.asan.a                  libtest-90e69e1db733a8fb.so
libhashbrown-1791beb5b36e409b.rlib          librustc-nightly_rt.lsan.a                  libunicode_width-803f21b2ca81a772.rlib
liblibc-3e961d059b9bcde7.rlib               librustc-nightly_rt.msan.a                  libunwind-757bd6fa410f1121.rlib

Though it is pretty minor.

getrandom should only be used as a dev-dep, and also we can't easily update it for various unfortunate reasons, so I'm not sure how much I'd recommend pushing on that.

Locally I was able to update it from 0.1.14 -> 0.1.16, which depends on cfg-if 1.0.0. Will this break in ways I don't understand?

[thomcc]

not exactly true, it's shipped in the normal distribution

Good point.

Locally I was able to update it from 0.1.14 -> 0.1.16, which depends on cfg-if 1.0.0. Will this break in ways I don't understand?

That's probably be fine. I think the issue is with getrandom 0.2 rather than the 0.1.x series -- whichever one requires wasm_bindgen for the wasm impl and compile_errors when it fails to resolve an impl (since std cannot depend on wasm-bindgen).

Feel free to r? @thomcc on a PR moving the stdlib to cfg-if v1.x.y.

[chbaker0]

Cool! I put it up for discussion: #103367