vi-style paren blinking causes unwanted "echo" in emacs

[hvesalai]

When using sbt from emacs (through sbt-mode), the terminal is set not to echo. However, if the input line contains parenthesis, the line is partially repeated.

scala> "foo"
val res1: String = foo

scala> "foo)bar"
"foo)val res2: String = foo()

Note on the last line the additional "foo), which is not expected.

The behavior is possibly caused by the vi-style paren blinking: https://github.com/jline/jline3/blob/master/reader/src/main/java/org/jline/reader/impl/LineReaderImpl.java#L2167

The question is, why is this functionality enabled for dumb terminals in sbt? Is there something wrong in how sbt uses jline3 or is it a bug in jline3 proper? I am available to help with a bug fix as soon as I understand which module to change.

[hvesalai]

More specifically the erroneous output appears after calling redisplay() here:

https://github.com/jline/jline3/blob/jline-parent-3.18.0/reader/src/main/java/org/jline/reader/impl/LineReaderImpl.java#L2197

I wonder if that line should be removed since redisplay() is conditionally called also after applying the widgets here:

https://github.com/jline/jline3/blob/jline-parent-3.18.0/reader/src/main/java/org/jline/reader/impl/LineReaderImpl.java#L705

Or if the condition if (!dumb) { should be moved to be within redisplay so handling of dumb terminal would be localized to one place.

[hvesalai]

@gnodet I would like to make a pull request to fix this, but would want some feedback on the approach before that. Would you have a moment to advice.

[gnodet]

@gnodet I would like to make a pull request to fix this, but would want some feedback on the approach before that. Would you have a moment to advice.

With a dumb terminal, I'm even wondering if the read line loop which read bindings and evaluates widgets should be called at all. Given there's no line edition possible, I think the LineReader should simply read the full line and return it without any processing. Thoughts ?

[hvesalai]

That's true. However, history should be stored because sbt provides even dump terminals ways to access the history.

[hvesalai]

OTOH. Things such as history expansion would work equally on DUMP terminals (i.e. !!)

[hvesalai]

How about if we make a keymap for DUMP that only has a few things that will work in that environment?

[gnodet]

How about if we make a keymap for DUMP that only has a few things that will work in that environment?

Yes, sounds a good idea.

[hvesalai]

Actually we already have there the safe() keybindings. Couldn't we use just that for DUMP?

[hvesalai]

ping @gnodet. I would appreciate feedback on my fix (I have tested it on the repl and it works). Besides the master, I would also like it to be ported to the next release if that is not from master

[hvesalai]

This wasn't completed after all. The reason is that https://github.com/jline/jline3/blob/jline-parent-3.24.1/reader/src/main/java/org/jline/reader/impl/InputRC.java#L47 undoes what we implemented in LineReaderImpl.

The line in

[hvesalai]

@gnodet can you reopen the issue, it needs more fixing, see above comment