Restore detection of missing terminado package #5465
Conversation
kevin-bates
changed the title
|
Support for the command line disablement still isn't right (sigh). Will remove [WIP] once resolved. |
notebook/notebookapp.py
Outdated
| @@ -300,7 +306,7 @@ def init_settings(self, jupyter_app, kernel_manager, contents_manager, | |||
| allow_password_change=jupyter_app.allow_password_change, | |||
| server_root_dir=root_dir, | |||
| jinja2_env=env, | |||
| terminals_available=False, # Set later if terminals are available | |||
| terminals_available=terminals_available, | |||
There was a problem hiding this comment.
There are three related things, which I think might be getting a bit confused here:
a. The terminals_enabled config option, which is True by default.
b. Whether terminado can be imported
c. Whether we are exposing terminals (i.e. enabled and importable)
Currently (before this PR), we use terminals_enabled for a and terminals_available for c (i.e. available to the user). In this PR,terminals_available refers to b, and terminals_enabled to c.
There was a problem hiding this comment.
Agreed - see latest commit.
kevin-bates
changed the title
notebook/notebookapp.py
Outdated
| except ImportError as e: | ||
| self.terminals_enabled = False |
There was a problem hiding this comment.
Modifying this still seems like it's inviting confusion, because what .terminals_enabled means depends on the point at which you use it: before this, it means a config option, and after this it means whether the terminal machinery is used.
Can I suggest that we keep clearly separate terms for the three things we might want to know, e.g. terminals_available, terminals_enabled and terminals_in_use?
There was a problem hiding this comment.
+1, although I believe terminado_available as an indication of the package's presence is more clear than terminals_available if that's okay.
I also think it's important to issue the warning message when terminals are enabled but the terminado is not present even though we know that answer prior to init_terminals. I've retained that try-catch in case there might be other initialization failures down the road.
There was a problem hiding this comment.
No problem - I wasn't tied to those specific names, just keen to distinguish the concepts.
Now terminals_available on the webapp is the equivalent of terminals_in_use on the Jupyter app class. It's possibly worth aligning those, but I don't think it's a big deal either.
notebook/notebookapp.py
Outdated
| if term_mgr.terminals: | ||
| return # Terminals still running | ||
| if not self.terminals_in_use: | ||
| return |
There was a problem hiding this comment.
This doesn't look quite right. This method is to shut down the server after a certain time with no kernels/terminals. Terminals not in use means no terminals running, but this early return prevents it from ever shutting down.
There was a problem hiding this comment.
😄 Looks like we could do with another pass at naming, as you alluded to in another comment. I was coming to same conclusion about terminals_in_use yesterday (while doing yard work). I think a switch (alignment) to terminals_available might just be the best name:
a. terminado_available = Terminado package is installed
b. terminals_enabled = User has condoned the use of terminals (independent of terminado's presence)
c. terminals_available = User has condoned the use of terminals and terminado is installed
There was a problem hiding this comment.
To be clear, I think there's an actual logic bug here, regardless of which name we're using. ;-)
There was a problem hiding this comment.
You're absolutely correct. Thank you for spotting that.
There was a problem hiding this comment.
Thanks. I think this is in good shape now. :-)
|
Other than my latest comment about shutting down on a timeout, I think this looks good now. |
Because the change to add culling of terminals derived from terminado's terminal manager in a local class, it inadvertently broke Notebook server's functionality to tolerate a missing terminado package. These changes restore that behavior.