Add a cache for the last 10 offline lookups done by OpenInv#matchPlayer(String) to reduce the performance hit of repeat opens
Add config option to open self with no-arg command
MSHADE-366 was fixed in maven-shade-plugin 3.4.1. This is technically unnecessary either way as the only things currently being shaded are all our own modules.
Fix memory leak with loaded players' advancements
Fix missing transaction transfer for player inventories
Fix incorrect transaction transfer for ender chests
Minor code health
Fixes various permissions not being respected during login/logout with inventories already open. This will result in a performance hit for repeated open/closes - there will be significantly more disk I/O.
Closes lishid#171
Closes#56
Wish it was more clearly indicated that the action was failing anywhere but the actions tab. No notification even, kind of annoying to have actually made a PR to double-check and then still have it fail.
If online, owner is always a viewer. If offline, they are never a viewer. There's no reason to count them - keeps inventory permanently cached while online if opened, even if not in use.
Because there is no option to create a `remapped-mojang` artifact for 1.16.5 and the whole point of the scripts is to save me time, 1.16.5 support is being removed earlier than usual.
Also fixes issues with NMS-based shulker collision check.
Closes#36
Temporarily explicitly declare Spigot dependencies
I don't have the time to deal with whatever the difference between remote and local bash/maven versions is right now.
With 1.17 releasing soon, we'll get to see if this was the issue very shortly.
Worst case scenario the yaml is still messed up, but since the changelog is now the last element, unless it's so badly malformed that the entire action can't run the file should still upload even if the changelog does not.
With the recent uptick in requests for support by people running 1.12 who have recently updated past Java 9, I revisited the supported versions.
First, the "modern" versions: 1.15 is run by 2.9% of servers, 1.16.3 by 5%. Since all of these versions are supported to ease the transition of updating servers and servers have updated, there's no real reason to keep them around. It's a lot easier if you can update a plugin and just have it work on both versions rather than push all the plugin updates to master with the server update, but none of these versions have any reason for long-term support.
Regarding heavily outdated server software: As of the time of writing, 1.8 and 1.12 have market shares of 8.6% and 8.5% of servers respectively.
Regarding 1.8: 1.8 support is already a bit wonky - with the changes made to inventory names, it's not (easily) possible to bridge the gap, and future changes will make that more and more difficult. People use 1.8 because they disliked the 1.9 combat changes, but there are plugins that fully rework combat to how it used to be. I have yet to hear a compelling argument that cannot be resolved with plugins. In the interest of my own sanity (handling and backporting the inventory name change in particular was a real humdinger) I will no longer be backporting changes to 1.8.
Regarding 1.12: I am honestly not sure why people are not updating. I get it, 1.13 and the flattening was a rough transition, but pretty much every developer active at the time bridged the gap. If your server is heavily dependent on some software only available at that time, you can live with other software available at that time. At the time of 1.12's release, Java 9 was not released. Either update your server, downgrade Java, or fork OpenInv and backport Java 9 support yourself.
Adam