Properly handle move feedback if mouse up happens outside window

[martin-fleck-at]

What it does

• Handle mouse down if we should be dragging as mouse up
• Reset feedback if selection has changed
  -- Ensure element does not have to be selected to revert move
  -- Add method for listeners to selection service

Fixes eclipse-glsp/glsp#1426

How to test

See eclipse-glsp/glsp#1426

Follow-ups

Changelog

• This PR should be mentioned in the changelog
• This PR introduces a breaking change (if yes, provide more details below for the changelog and the migration guide)

[ivy-lli]

Looks good.
Quickly tried the workflow example with this change -> seems good
Tried also our glsp project (via replace the @eclipse-glsp/client src and lib in the node_modules), but got an error (not sure if this is because of this change or something else?

[ivy-lli]

why rename? Still multiple elements right?

[martin-fleck-at]

Yeah, based on the wrong out-dated commit, good catch!

[martin-fleck-at]

Looks good. Quickly tried the workflow example with this change -> seems good Tried also our glsp project (via replace the @eclipse-glsp/client src and lib in the node_modules), but got an error (not sure if this is because of this change or something else?

Very interesting, do you already see that on the latest nightly? Because I don't think I changed anything regarding injection. As for the error itself, I wonder if the circular dependency is because of the ISelectionService in the file so moving that part to a separate file might help. I'll push an update to this PR soon which maybe you can try out. Otherwise we need to debug a little deeper.

[ivy-lli]

Seems to work with 2.3.0-next.389
However, I don't want to pretend that I did everything correctly with the lib/src replacement in the node_modules... Maybe I missed something here.

[tortmayr]

@ivy-lli I think this is indeed related to the lib/src replacement. I did a quick test and published the PR to a local npm registry (verdaccio). Consuming it from there seems to work without issues.

[tortmayr]

@martin-fleck-at There also seems to be an issue if I drag outside of the window during a move and then release the mouse.
Then the element gets not reseted and remains on the invalid position.
But I'm not sure if we can fix that easily.

[martin-fleck-at]

Based on the comment by Tobias, I just fixed the variable naming. I'd suggest we merge the PR if everyone is happy with it and then see if the consumption works and if not I can work on a fix since it is otherwise quite hard to test. What do you think?

EDIT: @tortmayr for your second comment: What should happen is that once you are back in the window, the element will follow the mouse cursor again and invalid clicks are correctly recognized.