Import order causes warnings when working with epicscorelibs and softioc

[MJGaughran]

As epicscorelibs maintains its own libca, it provides the option to set the 'CATOOLS_LIBCA_PATH' before using cothread. This can be found here. This mechanism does not appear to be ideal.

For example, softioc uses this mechanism here. The problem comes when another program attempts to use both softioc and cothread itself. A simple program such as the following:

#! /usr/bin/env python3

import cothread.catools
import softioc

will result in the following error (and potentially problematic behaviour):

$ ./test_imports.py 
/path/to/pipenv/environment/lib/python3.7/site-packages/epicscorelibs/path/cothread.py:18: UserWarning: epicscorelibs.path.cothread must be imported before cothread.catools to have effect
  warnings.warn("epicscorelibs.path.cothread must be imported before cothread.catools to have effect")

There are two problems:

1. It is not immediately clear what is causing the error
2. If using standard tools such as isort, you need to manually force a particular order of imports to avoid this error. You need to isort:skip multiple lines of code

I am not sure why cothread does not use epicscorelibs by default. Would it be possible to implement this?

Alternatively, some kind of conditional import of epicscorelibs could be useful.

[AlexanderWells-diamond]

After discussion with @Araneidae , we believe this is actually an issue in epicscorelibs. I've raised an issue there and will update this one based on the outcome of that: https://github.com/mdavidsaver/epicscorelibs/issues/22

[mdavidsaver]

The message quoted above is a warning, not an error message. Does some misbehavior result? Or is this only a case of wanting to clear up a nuisance warning? (imo. a valid request, just not an "error")

[AlexanderWells-diamond]

I believe there is no technical issue caused by the warning for us. There is potential for an issue in strict tests that check for no warnings being emitted, but I don't offhand know of any of these. I believe this is mostly just getting rid of the nuisance (and, as stated in the linked epicscorelibs issue, an incorrect) warning

[Araneidae]

Think you're right, have edited the title accordingly. I think this was flagged because it was causing concerns about the possibility of something having gone wrong elsewhere.

[mdavidsaver]

Ok, that is clear. I see this as a "bootstrapping" issue between cothread and epicscorelibs, which needs to be straightened out somehow.