Bug in SSH-Autocomplete with non-standard ports

[Luzifer]

As a fish-shell user I want to use command autocompletion for the ssh-command even when I'm working a lot with server listening on non-standard ports.

Expected behavior:

• Typing ssh myser<tab>
• Getting ssh myserver.hoster.domain.tld
• OR getting ssh -p2022 myserver.hoster.domain.tld

Actual behavior:

• Typing ssh myser<tab>
• Getting ssh \[myserver.hoster.domain.tld\]:2022

Either the port should get added as -p 2022 or it should completely be ignored as most people working with a lot of those special cases already have specific rulesets in their ~/.ssh/config to handle those non-standard ports.

Reason

Standard-port-servers are added like this in ~/.ssh/known_hosts:

github.com,204.232.175.90 ssh-rsa [...]

Non-standard-port-servers are added like this:

[myserver.hoster.domain.tld]:2022,[10.0.0.1]:2022 ssh-rsa [...]

System

OSX 10.9.2

luzifer@knut-workstation01 ~> fish --version
fish, version 2.1.0

[ridiculousfish]

Probably share/functions/__fish_print_hostnames.fish should be taught about this second ssh style. Today it's:

sgrep -Eoh '^[^#@|, ]*' ~/.ssh/known_hosts{,2} ^/dev/null

[KamilaBorowska]

+1. I wanted to report this a long time ago (my server runs on port 98), but I guess I forgot. Also, consider it to be confirmed since introduction of ssh completion.

[ridiculousfish]

If anyone wants to fix up the print_hostnames regex to handle this case, jump in!

[Luzifer]

I proposed a fix for this in my pull request. On my system it does work.

@xfix PTAL whether it also works for you…
@ridiculousfish PTAL whether the solution is okay for you…

[Pyppe]

I've got fish 2.6.0 (on macOS/iTerm2), and for me the ssh autocomplete doesn't seem to suggest anything for non-standard ports. For example, I've got the following (anonymised) line in my .ssh/known_hosts:

[some.server.com]:10020,[1.2.3.4]:10020 ecdsa-sha2-nistp256 AAAAE2...

But typing ssh some<TAB> yields nothing. Am I missing something, or has this issue regressed?

[zanchey]

Yes, this regressed in fbd53f2. I'll open a new issue.