Bug: @timeit would created a infinitely nested TimerOutput within a @threads block

[jingc769]

 using TimerOutputs, Base.Threads
 function parallel_task()
           to = TimerOutput()

           # define the loop to be run in parallel
           @threads for i in 1:10
               # measure the execution time of each loop iteration using @timeit
               @timeit to "within threads" begin
                   # perform some time-consuming task here
                   sleep(rand())
               end
           end

           # print the timing results
           to
       end

Would produce result like

 ─────────────────────────────────────────────────────────────────────────────────────
                                             Time                    Allocations
                                    ───────────────────────   ────────────────────────
          Tot / % measured:              1.64s /  16.4%           2.77MiB /   0.3%

 Section                    ncalls     time    %tot     avg     alloc    %tot      avg
 ─────────────────────────────────────────────────────────────────────────────────────
 within threads                  1    270ms  100.0%   270ms   9.86KiB  100.0%  9.86KiB
   within threads                1    335ms  124.1%   335ms    439KiB  4447.7%   439KiB
     within threads              2    1.58s  586.1%   791ms   1.53MiB  15867.6%   782KiB
       within threads            1    273ms  100.9%   273ms   13.0KiB  131.5%  13.0KiB
         within threads          4    2.31s  857.0%   579ms   2.77MiB  28808.8%   710KiB
           within threads        1    545ms  201.9%   545ms    862KiB  8740.7%   862KiB
 ─────────────────────────────────────────────────────────────────────────────────────

I guess the issue is because multi-threads messed up with "previous_timer" and "timer_stack".

[KristofferC]

Yeah, this package doesn't work well in a concurrent context right now.

[KristofferC]

I'll close this in favour of #72.