Closed denisri closed 4 years ago
I remember touching some code that handles signals in Anatomist recently: that was https://github.com/brainvisa/anatomist-free/commit/0cb86b866589c5cf4f5f3ab8c474198e1a45d841
Could it be that my change is causing a regression?
I also made the same change to the same duplicated bit of code here: https://github.com/brainvisa/axon/commit/3faa4ab1d614dda1b9835225ad1e413b413cf1fd#diff-05dcf59915ea0d781ea79c2f062e223a
This code was a copy of tornado.ioloop with a small modif for IPython 1. On our machines (and mine) we are now using later versions of ipython, thus this code is not used. Moreover it is not run until we open an ipython shell from anatomist or axon, so it should not impact this problem. Thanks for the hint anyway.
I suspect rather https://github.com/brainvisa/anatomist-free/commit/d31c580d9bd279e9eba9b275a237010e4eedc97a since I added a signal handler there (which doesn't work fully anyway)
Confirmed. This is the cause of the hangup.
I have experienced several hangups in Axon, commands processes (context.system()) actually finished but still existing as dfunct. After several trials I could see that the commands hang after an anatomist viewer is opened. So there must be something in anatomist startup which breaks child processes end / signals.