13:17 <gonidelis[m]> hkaiser meet today?

13:20 <hkaiser> gonidelis[m]: uhh, we have a holiday today, I completely forgot about our meeting :/

13:20 <hkaiser> I'm sorry

13:20 <gonidelis[m]> ahhhh no worries

13:21 <gonidelis[m]> I thought 4th of July was tomorrow ;p

13:21 <hkaiser> can we skip today's meeting? do you need help?

13:21 <hkaiser> yah, but they moved it to today as tomorrow is a saturday

13:21 <gonidelis[m]> hkaiser: don't worry... there is nothing that I can't just ask here on IRC

13:21 <hkaiser> thanks

13:25 <hkaiser> gonidelis[m]: you should be able to still meet with Aurian, though

13:25 <hkaiser> the zoom link should work in any case

13:25 <rori> 👍️

13:26 <gonidelis[m]> ok great

19:14 <jbjnr> hkaiser: yt?

19:19 <hkaiser> jbjnr: hey

20:14 <jbjnr> hkaiser: I'm looking into https://github.com/STEllAR-GROUP/hpx/issues/4800 and I can see what has gone wrong, but not where or why. When a task suspends, it keeps the state as pending and goes back onto the queue, but when it resumes, it has it's coroutine thread_id reset by mistake (race). Where in the code does this happen - I am not sure where the right place is. It is either 1) the thread suspends, but gets it's

20:14 <jbjnr> coroutine terminated by mistake - or it terminates, but the state stays as pending - either way, when it comes off the queue and is resumed - bang, we get the invalid coroutine assertion. Where is the coroutine reset in the code?

20:22 <hkaiser> jbjnr: let me think

20:25 <jbjnr> I'm getting close to the source of the problem.

20:25 <jbjnr> I'm using my printf debug stuff to show thread ids and suchlike at differnet places and I can see that after the scheduling loop executes the task, the set_thread state afterwards is possibly the source of problems

21:04 <hkaiser> jbjnr: the thread_id should get reset only if a thread terminates

21:04 <hkaiser> I don't know why it is reset just on resumtion