We could as psutil to the emsdk python if that would fix the issue?

I think the missing module has nothing to do with the --kill_exit but the --kill_start functionality of the script.
(I checked the script again, it may be used for many more things, there is a lot of globals involved 😬, my python knowledge is at this point a few hours old)

--kill_exit does not work because the script hangs on httpd.server_close().
This may have to do with a change of behaviour in python 3.7.
As stated here: https://docs.python.org/3/library/socketserver.html

Changed in version 3.7: socketserver.ForkingMixIn.server_close() and socketserver.ThreadingMixIn.server_close() now waits until all child processes and non-daemonic threads complete. Add a new socketserver.ForkingMixIn.block_on_close class attribute to opt-in for the pre-3.7 behaviour.

I dont know why a thread prevents the server_close() from finishing. (maybe a connection held by the browser?)

Sorry about mentioning two problems in one issue, I should have made a separate issue for the last sentence.
I have not tried yet if --kill_start works but I would not expect it too without psutil because the module is used to find instances of running browers.
After I have tested if it works I can also make a new issue for this if you prefer that, I cant try if it would work with the module because I dont know how to install a module to the emsdk python. (there is no pip?)

Hi, we observe the same issue on Windows: emrun will often hang after the server shuts down. We noticed that, as suggested above, setting the server's block_on_close to False sometimes helped,

https://docs.python.org/3/library/socketserver.html

ThreadingMixIn.server_close waits until all non-daemon threads complete, except if block_on_close attribute is False.

but sometimes the program would still block on sys.exit(...). We could then also replace that with os._exit(...) instead

https://stackoverflow.com/questions/19747371/python-exit-commands-why-so-many-and-when-should-each-be-used

os._exit() exits the program without calling cleanup handlers, flushing stdio buffers, etc. Thus, it is not a standard way to exit and should only be used in special cases.

And this combination then worked reliably but seems dirty. I think the reason sys.exit would hang if block_on_close were set to False is that then the server's daemon threads are not joined on close

https://github.com/python/cpython/blob/main/Lib/socketserver.py#L673C7-L673C21

So we found a (comparatively) cleaner solution to just set the server's socket itself to non-blocking before shutdown. This way the threads can still be joined and the ordinary and cleaner sys.exit(...) exit function can be used. Here is a diff that we propose to be adopted (which makes this behaviour a program option for emrun).

diff --git a/emrun.py b/emrun.py
index f7bc13df3..24c1a518a 100644
--- a/emrun.py
+++ b/emrun.py
@@ -704,6 +704,8 @@ class HTTPHandler(SimpleHTTPRequestHandler):
         if not emrun_options.serve_after_exit:
           page_exit_code = int(data[6:])
           logv('Web page has quit with a call to exit() with return code ' + str(page_exit_code) + '. Shutting down web server. Pass --serve-after-exit to keep serving even after the page terminates with exit().')
+          if emrun_options.force_exit:
+            self.server.socket.setblocking(False)
           self.server.shutdown()
           return
       else:
@@ -1563,6 +1565,9 @@ def parse_args(args):
   parser.add_argument('--dump-out-directory', default='dump_out', type=str,
                       help='If specified, overrides the directory for dump files using emrun_file_dump method.')

+  parser.add_argument('--force-exit', action='store_true',
+                      help='If true, sets server socket to nonblocking on shutdown to avoid sporadic deadlocks.')
+
   parser.add_argument('serve', nargs='?', default='')

   parser.add_argument('cmdlineparams', nargs='*')

Pull request: #22577