Dispose SSH connection when disconnecting & fixed disconnection events handling

[sebjulliand]

Changes

Fixes #2578

This PR fixes:

• The SSH connection not being disposed after the user manually disconnects
• The handling of disconnection events:
  • When end happens, then a normal disconnection happened (i.e. the user manually disconnected)
  • When error happens it means the connection was dropped on the client or server side

The connection not being disposed was causing #2578.
A broken connection will now only show the reconnect dialog once, for the connection that actually dropped.

How to test this PR

1. Reproduce the steps from Opening connections then dropping vpn causes multiple pop ups #2578 (comment)
2. Shut down VPN, or Wifi, or pull you Ethernet cable
3. Wait for a timeout for happen
4. The reconnect dialog shoud show up just once, for the connection that just dropped

Checklist

• have tested my change

[github-actions]

👋 A new build is available for this PR based on 09fc3bf.

• Download here.
• Read more about how to test

[arco400]

@sebjulliand I have tested the vsix and it solved my issue. I also tested the steps described in #2578, and the connection popup only appeared once so also good.

Sidenote: While running on the vsix version, I got the message "Code for IBM i will not function correctly until the temporary library has been corrected in the settings." I looked at the settings but did not see any anomalies. The temp library setting is filled with ILEDITOR.

[sebjulliand]

Sidenote: While running on the vsix version, I got the message "Code for IBM i will not function correctly until the temporary library has been corrected in the settings." I looked at the settings but did not see any anomalies. The temp library setting is filled with ILEDITOR.

@arco400 thank you for your test and your feedback; much appreciated!
I alos had this message during my testing but thought it was a one of a kind issue. I'll try to dig it a bit.

[SanjulaGanepola]

@sebjulliand Glad you found the main issue with this. I was also running into this problem often!

I did notice one regression which we should address as I confirmed that it doesn't happen on the master branch.

While running on the vsix version, I got the message "Code for IBM i will not function correctly until the temporary library has been corrected in the settings."

As for this specific error, I wasn't running into this on my end so not sure what may have caused it for you both.

[sebjulliand]

@SanjulaGanepola It should be just fine now 🤞🏻

[SanjulaGanepola]

@sebjulliand After some more testing, I believe the issue I found is actually not due to the changes in this PR, but actually from the big Mapepire PR.

The steps to reproduce:

1. Set a breakpoint in IBMi.runSQL
2. Connect to your VPN
3. Connect to IBM i
4. Once it hits the breakpoint, disconnect from VPN
5. Continue your debug and notice the UI is just hung

The idea is to have the SQL job connected, then disconnect your VPN, and try to execute some SQL. Looks like when await query.execute is called it doesn't return or even throw an error. I gave this a try in mapepire-js, but it does throw an error there. So could the issue be in our extended class?

vscode-ibmi/src/api/components/mapepire/sqlJob.ts

Line 8 in 6030618

export class sshSqlJob extends SQLJob {

Side note: Can we pull in @types/ssh2 as a dev dependency?

[sebjulliand]

Side note: Can we pull in @types/ssh2 as a dev dependency?

I'm definitely adding it; it's been bugging me so far not having these types 😛

[github-actions]

👋 A new build is available for this PR based on 5303aac.

• Download here.
• Read more about how to test

[sebjulliand]

@SanjulaGanepola I think I fixed it all.
I reworked the disconnection workflow, so every resource that needs to be reset/disposed is so whenever the connection is closed, for any reason.

Dropping the VPN at any point in time should not be a problem anymore.

I also found what the issue was with Code for IBM i will not function correctly until the temporary library has been corrected in the settings.: the global counter used to record the position in the SQL job joblog was not reset after disconnecting. That came from the Mapepire jumbo PR 😅

To reproduce the bug, you simply need to connect, disconnect, connect again to the same system.
I fixed it by resetting that counter whenever we disconnect.

Hopefully it's all good now 🙏🏻