Using FQN instead of adding import during ENC session

[genlu]

@tmat This feels like a better approach than listening to the debugger events, as completion provider is mostly stateless. What do you think?

FYI @dpoeschl This is not intended to be merged before snap today, just need to target some branch. Will decide and retarget if necessary to the proper branch later.

[jasonmalinowski]

The code doesn't look suspicious, but out of curiosity how much work is it to just allow imports to be added during an E&C session? I can imagine the difficulty comes from potentially rebinding other identifiers in the file, but it seems like we should be able to handle that or know if it's going to be a problem...

[tmat]

@jasonmalinowski It is indeed not impossible to implement and I would like to see us supporting it since it's the number one rude edit that customers run into.

I can imagine the difficulty comes from potentially rebinding other identifiers

Correct. We'd need to find out what methods are affected and recompile them, or report rude edits if we can't recompile a method that's affected.

[jasonmalinowski]

Is there a tracking item for this work we can point to this code (or stick a comment here) in hopes that if we fix it we remove this workaround?

[tmat]

#10043

[genlu]

FYI @jinujoseph @vatsalyaagrawal This needs QB approval

[genlu]

@jinujoseph It seems we have some issue with the integration test machines. I decided to merge this because:

1. 2 out of 4 legs of integration tests passed, before we start hitting this issue
2. This change can't affect integration tests, since import completion is not executed there.