Update TerminalConsoleAppender + JLine, use Jansi instead of JNA

[stephan-gh]

Fixes #1734. Fixes #1690 (?)

There were a few larger changes (new JLine/JNA version, changes in implementation of colors in TCA), so I am looking for some testing before I release the new version of TCA. @zachbr asked for a new release in Minecrell/TerminalConsoleAppender#7, so maybe you'd like to help with testing on Paper. :)

It looks fine to me in IDE+Terminal on Arch GNU/Linux, but I cannot test on macOS or Windows at the moment.

[zachbr]

The JNA change is breaking, this will need to be held until 1.14.4 if that is required.

[stephan-gh]

The JNA change is breaking, this will need to be held until 1.14.4 if that is required.

It seems so: jline/jline3@9d24f65

Edit: Urgh, this is going to cause me headaches for Forge...

[stephan-gh]

@zachbr Do you use JNA for anything except TCA on Paper?
Given that the new JNA version is going to cause me headaches for Forge, I've been thinking about using Jansi again. There is no functional difference for JLine; both are equally supported. And Jansi is actually much smaller (280 KB vs 1400 KB for JNA).

I think I stopped using Jansi back then because I felt it was old and somewhat messy. But it has been continuously updated together with JLine, so it's probably fine.

[astei]

@DemonWav has a branch for his experimental paperd daemon which uses JNA.

[astei]

I can test this new TCA release in Velocity if you want.

[stephan-gh]

@astei Sure. Any help with testing is appreciated!

[stephan-gh]

@DemonWav has a branch for his experimental paperd daemon which uses JNA.

Okay. Either way is fine really. I think I'll just make it clear in the TCA documentation that both can be used. Right now it strongly recommends JNA for no real reason.

[zachbr]

Looks good on macOS 10.15 and Windows 10 20H1

[DenWav]

I'm not doing anything fancy with JNA in feature/daemon that needs any one feature over another, as long as we don't downgrade.

It's also certainly possible for me to switch to a JNI approach. JNA doesn't require any other native code which is why I went with it, but we'd only need a small amount of C code to use JNI instead which could be a separate project where the prebuilt binary is simply downloaded at build time for Paper.

Another option that would be even better if I could get it to work is the JNI code could be part of the paperd binary itself, which could be used in daemon mode as well...

So don't worry about me, if removing JNA is the best way forward for this, go for it, using JNI won't be that hard (and will have a lot less overhead too).

[stephan-gh]

So don't worry about me, if removing JNA is the best way forward for this, go for it, using JNI won't be that hard (and will have a lot less overhead too).

I have decided that I will stop recommending one of them, so you will need to decide what to use in Paper. You could keep JNA now and switch to Jansi later once an alternative has been written for paperd.
Or whatever you like :) The functional difference is so minor that it does not matter which one is used.

[zachbr]

I wouldn't mind not using JNA because it would let us backport this change to older versions. I will have to try and find some time to test 1.12.2 and 1.13.2 with the jansi dependency instead.

[stephan-gh]

I've pushed a commit to TCA to stop depending on JNA explicitly. Therefore each project using TCA now has to add an explicit dependency on either jline-terminal-jansi or jline-terminal-jna.

I've changed this PR to depend on Jansi instead (and keep the JNA dependency as-is), so it is no longer breaking :)

[DenWav]

I'm testing out a new JNI implementation for paperd, should have it working today if all goes well. Should be cool to remove the JNA dep.

[DenWav]

paperd and the feature/daemon branch are now moved off JNA and onto JNI.

[stephan-gh]

Yay for conflicts! Now fixed :)
/me glances at #2324

[zachbr]

I've added a commit to this PR to remove the JNA runtime dependency.
Jansi version confirmed as expected on Arch Linux, macOS 10.15, and Windows Server 2016

[stephan-gh]

I've added a commit to this PR to remove JNA.
Jansi version confirmed as expected on Arch Linux, macOS 10.15, and Windows Server 2016

Great, thank you! I will release the new version of TCA soon and then we can merge this. 👍

[stephan-gh]

Updated TerminalConsoleAppender to the 1.2.0 release, ready for merge now!

@zachbr I squashed your commit into mine. I hope you don't mind :)

[zachbr]

Thanks for your continued support and assistance with TCA 👍

[HexedHero]

Since this update me and a lot of other owners have had this issue with their console:
https://cdn.hexedhero.com/u/uc2OBgyS7N.jpg

Using Pterodactyl 0.7.14
OpenJDK Runtime Environment (build 1.8.0_212-8u212-b01-1~deb9u1-b01)

Using -Dterminal.jline=false -Dterminal.ansi=true startup flags completely removed the >.... issue

[clrxbl]

@HexedHero Yeah I asked Minecrell about this earlier, I've made an issue for the Pterodactyl panel & the workaround PR with those arguments has been merged into parkervcp's eggs. (so this should be built in with the next daemon update) pelican-eggs/eggs#261

[stephan-gh]

I will need to reproduce this myself with some debugging code to be able to say more. The problem is likely a mixture of many things. JLine prints >.... in weird situations (see jline/jline3#411 for a similar problem that was already fixed), but Pterodactyl likely provides an incomplete console implementation if it is visible.

Unfortunately, Pterodactyl does not quite look like it can be installed in a few minutes....