[fix] Improve reliability attempt 1 :)

[urban-1]

Hi,

I looked a bit on how errors are being handled and found out few things:

• Transport layer errors are handled within the individual host implementation (ftp or sftp) rather than in the module that initiates the connection. This makes sense if you think that different implementation might have to deal with different errors
• Emitter is used to notify the user of any errors during saving files
• The "blue dot" on the tab (save indicator) is misleading, since our plugin saves two times: (1) locally in the tmp folder and (2): uploading on the remote. As is currently the save indicator updates after a local save and does not change back if the remote upload fails (this PR fixes this)
• Error handling on save() was a bit wrong... there are two async operations that both must succeed for the save to be complete. This PR fixes this by introducing a parent/wrapper Promise that will only succeed if both local and remote save work as expected
• The upload method was never rejecting/failing the promise! and was ignoring errors from the SSH writeFile method. This is now being handled
• Error messages were a bit vague (user side, console was detailed), trying to fix this by appending the specific error message to the popup/emitter

This has been partly tested with a local connection and lock filesystem permissions (to trigger errors). It should be contributing towards a more stable plugin, issue (#17) but I think we should leave the issue open.

Let me know what you think about this PR and if you see space for quick improvements

[newinnovations]

Hi Andreas,

I've been testing your PR the last week. I have a server that is behaving strangely. There seems to be a timeout of sorts. Saving (and reconnecting) silently fails even with this PR. The only way to recover is to restart atom. I have not have the time to investigate, but would like to do that before committing this PR.

[newinnovations]

The problem I was experiencing was related to timeouts in firewalls / nat. I will add a keepalive to the ssh2 connect. The ftp module seems to have a keepalive by default.