Difference in results.

[CMLL]

I've noticed sometimes that the completion results vary sometimes after a long time since a LS reload. I notice this from time to time, but this is the latest that I remember to took a shot of the issue to open an issue.

It occurs from time to time, but not pretty often, and I think is related to how long a code instance is open and how many files get opened and closed. Easiest way to clear it is by doing a window reload, and the the LS process will show the correct completions after the analysis is done.

I know is pretty vague information, but is a pretty hard bug to reproduce. I'll try to gather more info on this in the future.

I'm using 2 extra values to internal libraries in the python.autoComplete.extraPaths

[jakebailey]

That first screenshot looks like completion for the global scope, which may be a desync like in #1012 or #1009.

[MikhailArkhipov]

Desync was fixed some time ago.