| *** mhen_ is now known as mhen | 01:23 | |
| opendevreview | Dmitry Tantsur proposed openstack/futurist master: Remove sleeps from DynamicThreadPoolExecutor tests https://review.opendev.org/c/openstack/futurist/+/955758 | 09:13 |
|---|---|---|
| dtantsur | TheJulia: some problems I had with Queue and threads were actually between chair and keyboard ^^ | 09:17 |
| dtantsur | which makes the behavior you're observing even more puzzling | 09:17 |
| TheJulia | dtantsur: I think i figured it out, tl;dr its the rejection method because we end up queing in our code | 14:58 |
| TheJulia | because we had a delete and a add collide | 14:59 |
| TheJulia | so I'm thinking a slightly smarter rejection check for our use which understand "hey, we could be colliding, lets do the needful and then move on | 14:59 |
| dtantsur | Possibly, my brain refuses to think about queues and threading any more for now :D | 15:10 |
| TheJulia | don't blame you | 15:14 |
| TheJulia | bahahahaha, yeah, I found it | 15:17 |
| TheJulia | its colliding additions where the lock is held and the queue gets consulted. dtantsur your change is good, we just need to use a smarter check for work rejection on the ironic side :) | 15:49 |
| dtantsur | I see, it will be interesting to see the final code | 15:53 |
| TheJulia | yeah, the fix is just a smarter rejection check with a big note of what is happening | 15:57 |
| TheJulia | and why it exists :) | 15:57 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!