Check the worker process for any available bytes

[Drvi]

before we stop redirecting its output. When printing the bytesavailable(proc) from the added code segment it occasionally returned non-zeros (in my experiments where the workers were instantly aborted in the init_worker_expr). I don't know how to test this properly, but there seems to be no harm in adding it.

[nickrobinson251]

in my experiments

Can you share how you were experimenting with this? Even if we don't want to add it as a test, it'd be good to know how to go about triggering the issue

[nickrobinson251]

Does ConcurrentUtils need the same fix?

[Drvi]

how you were experimenting with this?

I added display(string("bytes: $(bytesavailable(proc))")) between the sleep and bytesavailable rows from this PR and then I was running this version of the "worker crashes immediately" test. Then very rarely this would happen:

julia> @testset "worker crashes immediately" begin
           file = joinpath(TEST_FILES_DIR, "_happy_tests.jl")
           worker_init_expr = quote
               if iseven(Libc.getpid())
                   @eval ccall(:abort, Cvoid, ())
                   # sleep(10)
               end
           end
           captured = IOCapture.capture() do
               encased_testset() do
                   runtests(file; nworkers=4, worker_init_expr)
               end
           end
           @show captured
       end

"bytes: 0"
"bytes: 0"
"bytes: 0"
"bytes: 328"
"""bbybytytetess:es:  6:69 9""6

9"

Does ConcurrentUtils need the same fix?

I think so, yeah