ssh: support dumb terminals in ssh client

[frazze-jobb]

Solves #7555

[github-actions]

CT Test Results

No tests were run for this PR. This is either because the build failed, or the PR is based on a branch without GH actions tests configured.

Results for commit d5a5199

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[u3s]

change of plans regarding tests. I've removed my commit because I needed it in #7499. I will merge to maint branch, hopefully before this PR. test for #7627 will be waiting on maint branch.

[attah]

What's the status on this? Anything more you need from me?
cc: @mhssler

[u3s]

for fix part, @frazze-jobb must comment.

for verification, a testcase is already merged to maint branch

otp/lib/ssh/test/ssh_connection_SUITE.erl

Line 768 in 67f87be

new_shell_dumb_term(Config) when is_list(Config) ->

[frazze-jobb]

Unfortunately I must do something more clever on the server side, which would require some effort, and currently I am too busy. Don't expect any news on this the coming two weeks. I will look at it after that.

[attah]

I'm hitting a very similar crash to what dialyzer found... so i think it is correct and not just a spec issue.

[frazze-jobb]

Thanks, I had missed to update two calls

[attah]

Unfortunately i'm still getting "\e[1022D\e[1B" in the data (and some other failures i have not had time to look at).
@u3s Did the test you mention pass? And did you have easy/ready access to our tests?

[frazze-jobb]

We've noticed the tests are not passing, and we are looking into it, but regarding "\e[1022D\e[1B" are your ssh client a dumb terminal? You can set environment variable TERM=dumb, and then it should act as if the client was running on a dumb terminal.

[u3s]

@u3s Did the test you mention pass?

No, it still fails - also in results for this PR here

And did you have easy/ready access to our tests?

Yes, sort of. AFAIK new_shell_dumb_term (failing above) mimicks your tests.

[attah]

My client is the OTP SSH client with {term, dumb}.

[frazze-jobb]

my code (and the tests) expects this type {term, string()}, so if you have a dumb atom there instead of a string, then it will not work.

[frazze-jobb]

Wops I see now that I was doing insert_chars requests instead of put_chars, thats why you saw \e[1022D\e[1B", not sure why I didnt see it before when I ran the test. But I saw them now this evening, so I have fixed it. Still failing a couple of tests though, Im working on it

[attah]

Nice! That indeed solves it for us.
We too have some remaining failures.

[frazze-jobb]

Alright I think I've fixed all the issues now

[attah]

All tests pass here too!