Lazy loading buggy until opening summery window

Describe the bug I’m a big fan of loading plugins lazy when they become necessary (e.g. I don’t need this plugin until I actually want to run a test or for writing an email not at all). But apparently this does not work. The plugin gets loaded and the commands etc are their. But if I run :Ultest or :UltestNearest, nothing happens. Silence. I added some debugging messages to the VimL stuff. And I can trace it down to this call where it switches to Python. In python I don’t anything to print. No clue what is happening here. If I execute :echo call('_ultest_run_all', ['path/to/file']) it always just prints 1. So the plugin is loaded and it basically does something. Just that nothing happens… When I open the summary window with :UltestSummary and then go to a test case and hit r to execute the case, it works. From now on I can also call :Ultest and :UltestNearest. If I turn of the lazy loading of the plugin and always source it at startup, it just works and I can call :Ultest right away.

Logs should be created using ULTEST_LOG_LEVEL=DEBUG ULTEST_LOG_FILE=vim-ultest.log nvim <test-file> and added here as well

11:36:06 | INFO | MainThread | logging.py:create_logger:71 | Logger created
11:36:06 | DEBUG | MainThread | __init__.py:__init__:40 | Handler created
11:36:06 | DEBUG | Thread-1 | jobs.py:_handle_coroutine:42 | Starting job with group run_all
11:36:06 | INFO | Thread-1 | __init__.py:run:101 | Running all tests in src/main.rs
11:36:06 | DEBUG | Thread-1 | jobs.py:_handle_coroutine:57 | Finished job with group run_all

To Reproduce

This assumes that vim-test is always available in path (simplification for the moment of testing). So this is the packer.nvim configuration:

lua packer.use({
      \   'rcarriga/vim-ultest',
      \   run = ':UpdateRemotePlugins',
      \   cmd = {'Ultest', 'UltestNearest', 'UltestSummary'},
      \ })

Expected behavior

The plugin works after it got loaded and you can execute :Ultest.

[rcarriga]

OK now should be working without any issue, just like normal loading.

[rcarriga]

I’ll leave this open and close it once I’ve implemented the forced update. I don’t think it’s fair to say that lazy loading works well until then