my test hangs locally on my computer with the changes I've done in bpo-32642 but doesn't hang on TravisCI, anyone able to help with checking what's wrong here (sounds like another edge case with my env but I could be wrong)

the trace for the command https://pastebin.com/q4FKnPZH

There are some resource leaks:

* When error occurred in setUp() function, server thread is not stopped.  It leaked threads and sockets for the server.
* When error occurred in server thread's run() method, asyncore.close_all() is not called.

These leaks makes dangling threads.  It caused hang in test_poplib shutdown process.

Additionally, the error is not printed in these cases.  It makes harder to fix real problem.

This is likely a duplicate of BPO #32706 and #32753

@Christian

Precisely, my PR doesn't fix #32706 and #32753.
It fixes only hang caused by them and print stack trace for them.

Just in case this would be useful to others, git bisect led me to this commit as the source of the problem:

61d478c71c5341cdc54e6bfb4ace4252852fd972 is the first bad commit
commit 61d478c71c5341cdc54e6bfb4ace4252852fd972
Author: Christian Heimes <christian@python.org>
Date:   Sat Jan 27 15:51:38 2018 +0100

    bpo-31399: Let OpenSSL verify hostname and IP address (#3462)
    
    bpo-31399: Let OpenSSL verify hostname and IP
    
    The ssl module now uses OpenSSL's X509_VERIFY_PARAM_set1_host() and
    X509_VERIFY_PARAM_set1_ip() API to verify hostname and IP addresses.
    
    * Remove match_hostname calls
    * Check for libssl with set1_host, libssl must provide X509_VERIFY_PARAM_set1_host()
    * Add documentation for OpenSSL 1.0.2 requirement
    * Don't support OpenSSL special mode with a leading dot, e.g. ".example.org" matches "www.example.org". It's not standard conform.
    * Add hostname_checks_common_name
    
    Signed-off-by: Christian Heimes <christian@python.org>

:040000 040000 09f4c8a18941f926c1f79e2f07dfd7731edf74c2 798ae3fab880e3365c490d56935901a4c18fd3bd M	Doc
:040000 040000 f35a1b4e2144050dfc4c7dbb02b7d92e3de1c2d0 7af4e4a8e82eafc6e7f2f4392c951213777f12c4 M	Lib
:040000 040000 e287c27f4db61beae905c21b96666194bb930c13 59e6fc63d296f7fe3187ff0987a04c9999e2d1d0 M	Misc
:040000 040000 95250f4c1009b5f10ec392c186908698d78f9cd3 160ebeb2c63d1ca305c7e37a0b555900b5084d20 M	Modules
:040000 040000 8adc7f664a8eaa667f4540f537304d7f9de672d4 41ea621f5433077f6dd776f4c2c51f7a8e4cab0f M	PC
:100644 100644 a6f4488cc99941abaa8b6fb7c53079b7cc292f9f ba0a7624cfcd0129e73bc218510f398001514df8 M	setup.py

Thanks for you confirmation. I already knew which commit lead to the issue. :)

The problem didn't "pop" up in my local tests or on Travis CI. I can only reproduce it locally, when I run the test in a tight loop and put extra load on the other CPU cores.

@christian.heimes I figured you probably didn't need it, but I did that work before I discovered this ticket. And as a nice bonus, I got to learn git bisect. :-)