Unclear docs on settings threads on workers

[jishnub]

The interactive use section seems to suggest that the following would work:
Using

using ParallelTestRunner
using MyPackage

testsuite = Dict("threads"=>:(@show Threads.nthreads()))
args = parse_args(ARGS)
runtests(MyPackage, args; testsuite)

and calling

julia> Pkg.test("MyPackage", julia_args=`--threads=4`)

one should be able to enable multiple threads on the workers. However, this doesn't seem to be the case.

Running 1 tests in parallel. If this is too many, specify the `--jobs=N` argument to the tests, or set the `JULIA_CPU_THREADS` environment variable.
               │          │ ──────────────── CPU ──────────────── │
Test  (Worker) │ Time (s) │ GC (s) │ GC % │ Alloc (MB) │ RSS (MB) │
threads    (1) │     0.17 │   0.00 │  0.0 │       5.20 │   409.45 │

Output generated during execution of 'threads':
[ Threads.nthreads() = 1

Test Summary: | Total   Time
  Overall     |     0  11.1s
    SUCCESS
     Testing MyPackage tests passed 

What does work is creating a custom worker using addworker(; exeflags = ["--threads=4"]), but ideally I don't want to create a worker per test. Is there a way to pass the number of threads to the default workers? It seems

ParallelTestRunner.jl/src/ParallelTestRunner.jl

Line 476 in 15115ec

push!(env, "JULIA_NUM_THREADS" => "1")

sets the number of threads to 1, and

ParallelTestRunner.jl/src/ParallelTestRunner.jl

Line 1009 in 15115ec

wrkr = p = addworker(; init_worker_code, io_ctx.color)

doesn't pass exeflags. I would appreciate pointers on this.

[vchuravy]

We could add a flag --worker-threads. Right now we assume single threaded execution and launch a set of parallel runners to not oversubscribe the workers. If we add a custom argument we could then scale down on parallelism, but I am unsure if that is useful.