Broken `partialsort` test unexpectedly passed in CI job

[jishnub]

This is probably a random number issue, as the broken test passes if all the array elements are chosen to be identical. In https://buildkite.com/julialang/julia-master/builds/31349#018c859d-6e20-43d7-bf95-804b688d1ac5,

The global RNG seed was 0x6822a35e4a9f43d3b58013ecc6981379.
Error in testset sorting:
Error During Test at C:\buildkite-agent\builds\win2k22-amdci6-4\julialang\julia-master\julia-f7de26d7ee\share\julia\test\sorting.jl:725
 Unexpected Pass
 Expression: partialsort(1:n, 172, lt = ((i, j)->begin
                    #= C:\buildkite-agent\builds\win2k22-amdci6-4\julialang\julia-master\julia-f7de26d7ee\share\julia\test\sorting.jl:725 =#
                    v[i] <= v[j]
                end)) == perm[172]
 Got correct result, please change to @test if no longer broken.
ERROR: LoadError: Test run finished with errors
in expression starting at C:\buildkite-agent\builds\win2k22-amdci6-4\julialang\julia-master\julia-f7de26d7ee\share\julia\test\runtests.jl:97
[...]
ERROR: A test has failed. Please submit a bug report (https://github.com/JuliaLang/julia/issues)
including error messages above and the output of versioninfo():
Julia Version 1.11.0-DEV.1130
Commit f7de26d7ee (2023-12-20 05:04 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Windows (i686-w64-mingw32)
  CPU: 8 × AMD EPYC Processor (with IBPB)
  WORD_SIZE: 32
  LLVM: libLLVM-15.0.7 (ORCJIT, znver1)
  Threads: 1 on 8 virtual cores
Environment:
  JULIA_BINARYDIST_FILENAME =
  JULIA_CMD_FOR_TESTS = julia-f7de26d7ee/bin/julia.exe .buildkite/utilities/timeout.jl julia-f7de26d7ee/bin/julia.exe
  JULIA_CPU_TARGET = pentium4
  JULIA_CPU_THREADS = 8
  JULIA_IMAGE_THREADS = 8
  JULIA_INSTALL_DIR = julia-f7de26d7ee
  JULIA_NUM_THREADS = 1
  JULIA_SHELL = C:/Program Files/Git/usr/bin/bash
  JULIA_TEST_IS_BASE_CI = true
  JULIA_TEST_MAXRSS_MB = 1536
  JULIA_TEST_VERBOSE_LOGS_DIR = C:\Users\julia\AppData\Local\Temp\jl_a8jqbI
  JULIA_VERSION = 1.11.0-DEV
Stacktrace:
 [1] error(s::String)
   @ Base .\error.jl:35
 [2] runtests(tests::String; ncores::Int32, exit_on_error::Bool, revise::Bool, seed::Nothing)
   @ Base .\util.jl:704
 [3] top-level scope
   @ none:1

The test is at

julia/test/sorting.jl

Lines 706 to 725 in d53fc50

	n = 1000
	v = rand(1:5, n);
	s = sort(v);
	# Nevertheless, it still works...
	for alg in safe_algs
	@test sort(v, alg=alg, lt = <=) == s
	end
	@test partialsort(v, 172, lt = <=) == s[172]
	@test partialsort(v, 315:415, lt = <=) == s[315:415]
	# ...and it is consistently reverse stable. All these algorithms swap v[i] and v[j]
	# where i < j if and only if lt(o, v[j], v[i]). This invariant holds even for
	# this invalid lt order.
	perm = reverse(sortperm(v, rev=true))
	for alg in safe_algs
	@test sort(1:n, alg=alg, lt = (i,j) -> v[i]<=v[j]) == perm
	end
	# Broken by the introduction of BracketedSort in #52006 which is unstable
	@test_broken partialsort(1:n, 172, lt = (i,j) -> v[i]<=v[j]) == perm[172]

[KristofferC]

#52494 (comment)