Improve tests and benchmarks for slice::sort and slice::sort_unstable

[ghost]

This PR just hardens the tests and improves benchmarks.
More specifically:

1. Benchmarks don't generate vectors in Bencher::iter loops, but simply clone pregenerated vectors.
2. Benchmark *_strings doesn't allocate Strings in Bencher::iter loops, but merely clones a Vec<&str>.
3. Benchmarks use seeded XorShiftRng to be more consistent.
4. Additional tests for slice::sort are added, which test sorting on slices with several ascending/descending runs. The implementation identifies such runs so it's a good idea to test that scenario a bit.
5. More checks are added to run-pass/vector-sort-panic-safe.rs. Sort algorithms copy elements around a lot (merge sort uses an auxilliary buffer and pdqsort copies the pivot onto the stack before partitioning, then writes it back into the slice). If elements that are being sorted are internally mutable and comparison function mutates them, it is important to make sure that sort algorithms always use the latest "versions" of elements. New checks verify that this is true for both slice::sort and slice::sort_unstable.

As a side note, all of those improvements were made as part of the parallel sorts PR in Rayon (nikomatsakis/rayon#379) and now I'm backporting them into libcore/libstd.

r? @alexcrichton

[alexcrichton]

@bors: r+

Thanks!

[bors]

📌 Commit 12205f1 has been approved by alexcrichton

[bors]

⌛ Testing commit 12205f1 with merge 32091e1b39664ecd81d82276c1ce000e8a5c75b2...

[bors]

💔 Test failed - status-travis

[alexcrichton]

Looks like a legitimate failure?

[01:06:00] [01:06:00] failures:
[01:06:00] [01:06:00] ---- [run-pass] run-pass/vector-sort-panic-safe.rs stdout ----
[01:06:00] [01:06:00] error: test run failed!
[01:06:00] status: exit code: 101
[01:06:00] command: /checkout/obj/build/x86_64-unknown-linux-gnu/stage0-tools/x86_64-unknown-linux-gnu/release/remote-test-client run /checkout/obj/build/x86_64-unknown-linux-gnu/test/run-pass/vector-sort-panic-safe.stage2-arm-unknown-linux-gnueabihf
[01:06:00] stdout:
[01:06:00] ------------------------------------------
[01:06:00] uploaded "/checkout/obj/build/x86_64-unknown-linux-gnu/test/run-pass/vector-sort-panic-safe.stage2-arm-unknown-linux-gnueabihf", waiting for result
[01:06:00] [01:06:00] ------------------------------------------
[01:06:00] stderr:
[01:06:00] ------------------------------------------
[01:06:00] thread 'main' panicked at 'attempt to calculate the remainder with a divisor of zero', /checkout/src/test/run-pass/vector-sort-panic-safe.rs:151
[01:06:00] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[01:06:00] [01:06:00] ------------------------------------------
[01:06:00] [01:06:00] thread '[run-pass] run-pass/vector-sort-panic-safe.rs' panicked at 'explicit panic', /checkout/src/tools/compiletest/src/runtest.rs:2480
[01:06:00] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[01:06:00] [01:06:00] [01:06:00] failures:
[01:06:00] [run-pass] run-pass/vector-sort-panic-safe.rs
[01:06:00] [01:06:00] test result: �[31mFAILED�(B�[m. 2696 passed; 1 failed; 9 ignored; 0 measured; 0 filtered out

[ghost]

The failure is apparently due to a codegen bug on ARM platforms.

I believe simply moving let mut rng = thread_rng() (here) outside the innermost loop would work around the build failure. Moving that line of code is a good idea anyway, so... can we do that or should we wait until the codegen issue is resolved?

[alexcrichton]

@stjepang yeah sounds good to me!

[ghost]

All right, let's see if it works this time...

[alexcrichton]

@bors: r+

[bors]

📌 Commit 723833f has been approved by alexcrichton

[ghost]

The new failures are spurious, it seems... three targets were not tested at all (they're marked with ! rather than X). Perhaps retry?

[kennytm]

@stjepang Don't worry about the travis-ci/pr failure, it is fine as long as the IMAGE=x86_64-gnu-llvm-3.7 ALLOW_PR=1 RUST_BACKTRACE=1 one passes.

[Mark-Simulacrum]

@bors rollup

[Mark-Simulacrum]

@bors r- This failed the rollup, with I think the same failure. Possibly will be fixed after the LLVM update, not sure.

[01:05:40] ---- [run-pass] run-pass/vector-sort-panic-safe.rs stdout ----
[01:05:40]
[01:05:40] error: test run failed!
[01:05:40] status: exit code: 101
[01:05:40] command: /checkout/obj/build/x86_64-unknown-linux-gnu/stage0-tools/x86_64-unknown-linux-gnu/release/remote-test-client run /checkout/obj/build/x86_64-unknown-linux-gnu/test/run-pass/vector-sort-panic-safe.stage2-arm-unknown-linux-gnueabihftdout:
[01:05:40] ------------------------------------------
[01:05:40] uploaded "/checkout/obj/build/x86_64-unknown-linux-gnu/test/run-pass/vector-sort-panic-safe.stage2-arm-unknown-linux-gnueabihf", waiting for result
[01:05:40]
[01:05:40] ------------------------------------------
[01:05:40] stderr:
[01:05:40] ------------------------------------------
[01:05:40] thread 'main' panicked at 'attempt to calculate the remainder with a divisor of zero', /checkout/src/test/run-pass/vector-sort-panic-safe.rs:152
[01:05:40] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[01:05:40]
[01:05:40] ------------------------------------------
[01:05:40]
[01:05:40] thread '[run-pass] run-pass/vector-sort-panic-safe.rs' panicked at 'explicit panic', /checkout/src/tools/compiletest/src/runtest.rs:2479
[01:05:40] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[01:05:40]
[01:05:40]
[01:05:40] failures:
[01:05:40] [run-pass] run-pass/vector-sort-panic-safe.rs
[01:05:40]
[01:05:40] test result: FAILED. 2697 passed; 1 failed; 9 ignored; 0 measured; 0 filtered out
[01:05:40]
[01:05:40] thread 'main' panicked at 'Some tests failed', /checkout/src/tools/compiletest/src/main.rs:325

[ghost]

It will be fixed after the update, yes.

[ghost]

Ok, LLVM has been updated. Should I rebase now?

[alexcrichton]

@bors: r+

Nah should still apply cleanly!

[bors]

📌 Commit 723833f has been approved by alexcrichton

[bors]

⌛ Testing commit 723833f with merge 05b5797...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: alexcrichton
Pushing 05b5797 to master...