Adam Williamson

It doesn't look like 3899092 (from 3888467) is a perl thread to me: ``` Thread 2 (Thread 0xfffe9e20f180 (LWP 3899092) "/usr/bin/isotov"): #0 0x0000ffffb691f7e4 in syscall () from /lib64/libc.so.6 #1 0x0000ffffb2133384...

Just going by the PIDs it looks "younger" than all the other threads from that process. The parent is 3888467 , its children start at 3888500 and go to 3888578...

Here's the full backtrace of that thread: ``` Thread 2 (Thread 0xfffe9e20f180 (LWP 3899092) "/usr/bin/isotov"): #0 syscall () at ../sysdeps/unix/sysv/linux/aarch64/syscall.S:38 No locals. #1 0x0000ffffb2133384 in tbb::detail::r1::futex_wait (comparand=2, futex=0xffff07706a24) at /usr/src/debug/tbb-2021.11.0-5.fc40.aarch64/src/tbb/semaphore.h:100...

sorry, I got pulled away onto other stuff, but trying to get back to this. we'd have to patch os-autoinst to launch isotovideo that way and dump the output somewhere,...

poking about at this...on a host with 64 CPUs (according to `/proc/cpuinfo` anyhow), counting threads with the `grep ^SigBlk /proc/(PID)/task/*/status` trick, the count seems kinda interesting. In a process where...

So I hacked up os-autoinst to dump out the relevant numbers, and that confirms my expectation: ``` XXX cv::getNumThreads: 64 XXX cv::getNumberofCPUs: 64 XXX thread count: 63 ``` That "thread...

Hah! So...I think a silly guess I had about this was right. I think the first iteration of the `parallel_for_` loop runs on *the parent thread*. It doesn't create a...

uff, maybe not. if I change that code to try and spawn one more thread, the tests fail because isotovideo doesn't start up (presumably it's deadlocked trying to start the...

oh, I forgot to mention, I have been trying the strace thing too but haven't yet seemingly managed to hit the bug that way. I edited `[email protected]` to run through...

sigh, I tried another crazy guess but it didn't work. there definitely seems to be a shy bug element to this: I really just *cannot* get it to trigger when...