newly added test_size::test_proc_macro2_wrapper_size_without_locations test fails on i686

[decathorpe]

I noticed that the proc-macro2 crate's test suite started to fail on Fedora Linux - from the build logs:

     Running `/builddir/build/BUILD/proc-macro2-1.0.84/target/rpm/deps/test_size-05b53b4047993840 --skip src/lib.rs`
running 5 tests
test test_proc_macro2_fallback_size_with_locations ... ignored
test test_proc_macro2_fallback_size_without_locations ... ignored
test test_proc_macro2_wrapper_size_with_locations ... ignored
test test_proc_macro_size ... ok
test test_proc_macro2_wrapper_size_without_locations ... FAILED
failures:
---- test_proc_macro2_wrapper_size_without_locations stdout ----
thread 'test_proc_macro2_wrapper_size_without_locations' panicked at tests/test_size.rs:52:5:
assertion `left == right` failed
  left: 20
 right: 24
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
failures:
    test_proc_macro2_wrapper_size_without_locations
test result: FAILED. 1 passed; 1 failed; 3 ignored; 0 measured; 0 filtered out; finished in 0.00s
error: test failed, to rerun pass `--test test_size`

The output points to this assertion:
https://github.com/dtolnay/proc-macro2/blob/1.0.84/tests/test_size.rs#L52

Looks like the size of proc_macro2::Group is 4 bytes smaller on i686 than on other architectures.
Possibly this affects all 32-bit architectures, but i686 is the only one that we still partially support in Fedora Linux.

As far as I can tell, the size difference is harmless (probably you just added it to make sure the size doesn't accidentally grow?), so I have ignored this test in our builds for now, but I still wanted to file an issue in case you want to add an additional not(target_pointer_width = "64") clause to the test's #[ignore] cases.

[dtolnay]

You inferred correctly, we don't need the types to be a particular size, I just need to be cognizant of when they change size because it impacts syn's performance. And for that purpose, I personally only care about 64-bit.

Thanks for reporting the issue. I have added #[cfg_attr(not(target_pointer_width = "64"), ignore)] in proc-macro2 1.0.85.

$ cargo miri test --test test_size --target i686-unknown-linux-gnu
    Finished `test` profile [unoptimized + debuginfo] target(s) in 0.04s
     Running tests/test_size.rs (target/miri/i686-unknown-linux-gnu/debug/deps/test_size-b5488b1435a7fdbe)

running 5 tests
test test_proc_macro2_fallback_size_with_locations ... ignored
test test_proc_macro2_fallback_size_without_locations ... ignored
test test_proc_macro2_wrapper_size_with_locations ... ignored
test test_proc_macro2_wrapper_size_without_locations ... ignored
test test_proc_macro_size ... ignored

test result: ok. 0 passed; 0 failed; 5 ignored; 0 measured; 0 filtered out; finished in 0.35s

[decathorpe]

Great, thank you!