We've noticed that on our Win32 (XP) Continuous Integration box, the mongrel_service occasionally fails to kill the
ruby.exe process when it shuts down.
The initial error we see is: "Error in my_thread_global_end(): 1 threads didn't exit"
The service itself does stop, but leaves the ruby.exe process running.
Once this has happened, subsequent net start/stop requests can throw the following error (although not always): "System
error 1067 has occurred" and the serviced may report a 'did not start' error, however the application is still
being served from the old ruby.exe process and so tests run (although the application remains at the previously loaded
version).
There are some posts in the PHP forums suggesting that this is a MySql? driver issue, (we are using MySql? 5.0.45). But
I thought it might be worth posting here in case there is a timing problem in the way in which the mongrel_service shuts
down the mongrel.
Forgive me if this is not well phrased, I'm personally not very familiar with Ruby or Mongrel.
