looking for a few jline testers

On Friday, January 20, 2012 9:35:53 AM UTC-6, Paul Phillips wrote:

I pushed a new branch:
  https://github.com/scala/scala/tree/jline-1.0
I also built a distribution of it:
  https://github.com/downloads/scala/scala/scala-2.10-jline-1.0-ge8a7f6e.tgz
This eliminates our custom jline.  On my platform it works surprisingly well given that I did almost no tuning once it compiled.  But one thing I know for sure is that every platform is its own thing.  If there are any windows, linux, or whatever else people who can try this out and tell me if you see any shocking regressions (and don't expect it to be perfect just yet, but any behavioral difference from 2.9.1 is of interest) that would help a lot.

    On my Windows XP SP3 machine, I'm seeing regression regarding the issue described in SI-2547: in this new distribution, holding down the right and left arrow keys makes the cursor disappear, while that doesn't happen for me with 2.9.1.   Keith Hodel

On Fri, Jan 20, 2012 at 8:12 AM, kHodel <webinfo@khodel.info> wrote:

On my Windows XP SP3 machine, I'm seeing regression regarding the issue described in SI-2547: in this new distribution, holding down the right and left arrow keys makes the cursor disappear, while that doesn't happen for me with 2.9.1.

OK, that one kind of figures.  I was vaguely hoping that the accumulated windows tweaks would all turn out to have been bundled in 1.0.  I'll look into it.

On Friday, January 20, 2012 10:19:44 AM UTC-6, Paul Phillips wrote:

On Fri, Jan 20, 2012 at 8:12 AM, kHodel <web...@khodel.info> wrote:

On my Windows XP SP3 machine, I'm seeing regression regarding the issue described in SI-2547: in this new distribution, holding down the right and left arrow keys makes the cursor disappear, while that doesn't happen for me with 2.9.1.

OK, that one kind of figures.  I was vaguely hoping that the accumulated windows tweaks would all turn out to have been bundled in 1.0.  I'll look into it.

  I suspect this will be unsurprising, but for the record I've also noticed (still on the same Windows XP SP3 machine) regression of SI-976 (use of escape key after arrow usage crashes REPL).

I checked basic editing and moving around on linux and it works as expected

On Fri, Jan 20, 2012 at 16:35, Paul Phillips <paulp@improving.org> wrote:

I pushed a new branch:
  https://github.com/scala/scala/tree/jline-1.0
I also built a distribution of it:
  https://github.com/downloads/scala/scala/scala-2.10-jline-1.0-ge8a7f6e.tgz
This eliminates our custom jline.  On my platform it works surprisingly well given that I did almost no tuning once it compiled.  But one thing I know for sure is that every platform is its own thing.  If there are any windows, linux, or whatever else people who can try this out and tell me if you see any shocking regressions (and don't expect it to be perfect just yet, but any behavioral difference from 2.9.1 is of interest) that would help a lot.

On Fri, Jan 20, 2012 at 10:26 AM, kHodel <webinfo@khodel.info> wrote:

I suspect this will be unsurprising, but for the record I've also noticed (still on the same Windows XP SP3 machine) regression of SI-976 (use of escape key after arrow usage crashes REPL).

OK, if you can assist me a bit longer: now all the way at the other extreme, I have another snapshot of trunk, this time using jline as of five minutes ago.
  https://github.com/downloads/scala/scala/scala-2.10-jline-2.6-g0a42073.tgz
How do those regressions fare in that one?

 
On Friday, January 20, 2012 12:50:04 PM UTC-6, Paul Phillips wrote:

How do those regressions fare in that one?

  Mixed results. Arrowing left and right now works with the cursor showing properly, and the escape key isn't causing a crash as before, but it isn't quite working right either. When I press the escape key it doesn't clear the line, and then odd things will happen, depending on the next key press: two consecutive escapes triggers autocomplete, and using the left and right arrows immediately after escape causes "K" and "M" (respectively) to be inserted into the line. Tab also seems to fail to trigger autocomplete once I've used the arrow keys. Also, quite unfortunately, so far every time that I've tried to :quit, it has frozen rather than exiting normally. And when I did some typing and then deleted the text and pressed enter, it froze up on me then as well.

On Fri, Jan 20, 2012 at 12:04 PM, kHodel <webinfo@khodel.info> wrote:

Mixed results. Arrowing left and right now works with the cursor showing properly, and the escape key isn't causing a crash as before, but it isn't quite working right either. When I press the escape key it doesn't clear the line, and then odd things will happen, depending on the next key press: two consecutive escapes triggers autocomplete, and using the left and right arrows immediately after escape causes "K" and "M" (respectively) to be inserted into the line. Tab also seems to fail to trigger autocomplete once I've used the arrow keys. Also, quite unfortunately, so far every time that I've tried to :quit, it has frozen rather than exiting normally. And when I did some typing and then deleted the text and pressed enter, it froze up on me then as well.

OK, thanks.  I've taken this as far as I can, but (for a change) I've published all the work.    https://github.com/scala/scala/tree/jline
All the scala tests pass, so all the aspiring scala-jline-improver has to do is identify and eliminate regressions against 2.9.1, the kind which are hard to test for (not impossible, but at least demanding test to write.) I can pitch in a couple more regressions: on OSX ctrl-O and ctrl-T work for next/prev word.  In 2.9.1 they do, that is, but no longer in the snapshot above.

(I'm the guy that was complaining about vi-mode on the other jline thread).
I was pretty excited about trying out this jline build because I saw that you built it on 2.6-SNAPSHOT, which does actually include the vi-mode stuff by Guillaume Nodet wrote in jline2. Sure enough, putting 'set editing-mode vi' in my ~/.inputrc did enable vi-mode...sort of.
It appears to be fairly broken. The main thing is that there appears to be no way to get into insert mode -- the repl stays in normal (command) mode all the time, and keys such as 'i' and 'A' don't appear to have any effect.
Here's a full list of the stuff I tried:
i, A - no effectl, h - work as expected for cursor movementj, k - work as expected for next/previous lines b, e, w - no effect$, 0 - work as expected for end/beginningd - no effect
-Caleb
On Fri, Jan 20, 2012 at 12:13 PM, Paul Phillips <paulp@improving.org> wrote:

On Fri, Jan 20, 2012 at 12:04 PM, kHodel <webinfo@khodel.info> wrote:

Mixed results. Arrowing left and right now works with the cursor showing properly, and the escape key isn't causing a crash as before, but it isn't quite working right either. When I press the escape key it doesn't clear the line, and then odd things will happen, depending on the next key press: two consecutive escapes triggers autocomplete, and using the left and right arrows immediately after escape causes "K" and "M" (respectively) to be inserted into the line. Tab also seems to fail to trigger autocomplete once I've used the arrow keys. Also, quite unfortunately, so far every time that I've tried to :quit, it has frozen rather than exiting normally. And when I did some typing and then deleted the text and pressed enter, it froze up on me then as well.

OK, thanks.  I've taken this as far as I can, but (for a change) I've published all the work.    https://github.com/scala/scala/tree/jline
All the scala tests pass, so all the aspiring scala-jline-improver has to do is identify and eliminate regressions against 2.9.1, the kind which are hard to test for (not impossible, but at least demanding test to write.) I can pitch in a couple more regressions: on OSX ctrl-O and ctrl-T work for next/prev word.  In 2.9.1 they do, that is, but no longer in the snapshot above.

On Fri, Jan 20, 2012 at 2:43 PM, Caleb Spare <cespare@gmail.com> wrote:

It appears to be fairly broken. The main thing is that there appears to be no way to get into insert mode -- the repl stays in normal (command) mode all the time, and keys such as 'i' and 'A' don't appear to have any effect.

Now is your chance to contribute.  I won't be able to take this any further.

On Fri, Jan 20, 2012 at 2:54 PM, Paul Phillips <paulp@improving.org> wrote:

On Fri, Jan 20, 2012 at 2:43 PM, Caleb Spare <cespare@gmail.com> wrote:

It appears to be fairly broken. The main thing is that there appears to be no way to get into insert mode -- the repl stays in normal (command) mode all the time, and keys such as 'i' and 'A' don't appear to have any effect.

Now is your chance to contribute.  I won't be able to take this any further.

You mean contribute by sending patches to jline2 and getting those guys to publish a new snapshot?

On Fri, Jan 20, 2012 at 3:25 PM, Caleb Spare <cespare@gmail.com> wrote:

You mean contribute by sending patches to jline2 and getting those guys to publish a new snapshot?

They didn't publish that one, I built it.  You don't have to send patches to jline2 (although that would be the best) you just have to write and publish the code.  If there's a repository somewhere which I can build and which gives us a better jline than we have, one not containing regressions against 2.9.1, then that's what we'll ship.  It's that regression condition (across multiple platforms) which tends consume one's time.

Oh, and I can't remember if I pointed at it explicitly, but here is the repo from which the 2.9.1 jline was built:

  https://github.com/paulp/jline2
It's also in the scala repo but it has a proper history in this one.  It includes an sbt project and other bonuses (like having no regressions against itself.)