Open Darksonn opened 4 years ago
From the back trace this looks like it may be due to budget stuff?
Locally on my windows machine I can't reproduce this with individual runs.
The panic in the stack trace is from the panic_in_task
test, and should be unrelated since each test runs in a separate runtime. As far as I have found, backtraces from intended panics interleave with the "ok"/"fail" messages even when all tests are successful.
I can replicate the threaded_scheduler_1_thread::runtime_in_thread_local
deadlock all the time with #2649. Upon investigation, I think the issue is a deadlock in the destructor of a thread local storage variable, but I couldn't figure out exactly what.
I used to have threaded_scheduler_1_thread::shutdown_timeout
block from time to time too, but not with #2649. I think it was related with the shutdown_timeout
bug that I fixed.
CI sometimes fails the Windows tests with a deadlock. The tests that deadlock are:
threaded_scheduler_1_thread::runtime_in_thread_local
threaded_scheduler_1_thread::shutdown_timeout
They also print a panic message, which I assume is the cause of a deadlock.
Click to expand backtrace
``` thread 'tokio-runtime-worker' panicked at 'explicit panic', tokio\tests\rt_common.rs:661:17 stack backtrace: test threaded_scheduler_1_thread::panic_in_block_on ... ok 0: core::fmt::write at /rustc/4fb7144ed159f94491249e86d5bbd033b5d60550\/src\libcore\fmt\mod.rs:1063 1: std::io::Write::write_fmt