Intermittent failure in region-0.0.5 test suite in Rust 1.15

[brson]

https://github.com/darfink/region-rs at 1ae7a6a157b398cb1109422a55da970b696c55c4

This happens perhaps 50% of runs on Rust 1.15 and never on Rust 1.14

brian@ip-10-145-43-250:~/dev/region-rs⟫ cargo +beta test
    Finished debug [unoptimized + debuginfo] target(s) in 0.0 secs
     Running target/debug/deps/region-dd0594b891b7a4e4

running 21 tests
test os::linux::tests::parse_flags ... ok
test lock::tests::lock_page_release ... ok
test lock::tests::lock_page ... ok
test tests::protect_alloc ... ok
test os::tests::page_rounding ... ok
test os::tests::page_size_value ... ok
test tests::protect_code ... ok
test tests::protect_null ... ok
test os::linux::tests::parse_region ... ok
test tests::protect_overlap ... ok
test tests::query_alloc ... ok
test tests::query_area_overlap ... ok
test tests::query_null ... ok
test tests::query_area_alloc ... ok
test view::tests::view_check_size ... ok
test tests::query_area_zero ... ok
test tests::query_code ... ok
test view::tests::view_exec_prot ... ok
test view::tests::view_get_prot ... ok
test view::tests::view_prot_prev ... ok
test view::tests::view_prot_initial ... FAILED

failures:

---- view::tests::view_prot_initial stdout ----
        thread 'view::tests::view_prot_initial' panicked at 'called `Result::unwrap()` on an `Err` value: Error(Free, State { next_error: None, backtrace: None })', /buildslave/rust-buildbot/slave/beta-dist-rustc-linux/build/src/libcore/result.rs:837
note: Run with `RUST_BACKTRACE=1` for a backtrace.


failures:
    view::tests::view_prot_initial

test result: FAILED. 20 passed; 1 failed; 0 ignored; 0 measured

error: test failed

The exact test that fails differs from run to run.

cc @darfink

[alexcrichton]

@brson hm I seem to be seeing the opposite behavior locally, where the test suite fails sporadically on stable but always passes on beta?

[brson]

@alexcrichton on further inspection I can repro on stable and beta, though it seems less frequent on stable still than on beta. So it's probably not a bug on our side.