ZQuest Classic The Classic Adventure Game Engine
All Releases | 2.55.2 -> 2.55.3 -> 2.55.4
View on GitHub
2.55.3
Linux
Mac
Windows
64-bit
Windows
32-bit
Downloads:

View a summary of what's new in 2.55.

Bug Fixes

  • correct the version string for 2.55.x releases 387e6c05c
  • incorrect default right string margin ae1c890e7

Player

  • visual text bug in sound panel 32a169b2f
  • active shields interacting with light beams while inactive 7e52830f0
  • reset playing_field_offset at beginning of scroll 5e1f431eb  

    This prevents the draw offset from a quake from persisting during a scroll.

    See https://discord.com/channels/876899628556091432/1238226802867048529

Editor

  • "Used Tiles" completely failing to calculate 4066823d5

ZScript

  • TraceS parameter not compiling correctly 2761eb50b
  • handle engine subscreen in ScriptTypeToString 150aafa9f
  • dealloc owned objects for all combos in loadscr e9438011a  

    For some reason, only the combo at position 0 had its owned objects deallocated when loading a new screen. Now they all do.

Documentation

Editor

  • touch up attack/whimsical ring help text ('damage' -> 'power') 9879b835f

Refactors

Player

  • rename the "ZScript Debugger" to "ZC Console" 719738e74

ZScript

  • remove ZASM debugger b3b79f66e  

    This was already made inaccessible from the GUI in 1da2d02, but now it is removed entirely. Work on a real debugger will begin soon.

  • tweak Region-> and screen script internals in preparation for z3 e02113363  

    Region->OriginScreen is now of type screendata, and Region->OriginScreenIndex is added to return just the screen index.

    Similar change made to Region->GetScreenForComoboPos(int pos).

  • use faster register allocation in jit c6da3fe63  

    asmjit's register allocation was using a bin-packing algorithm. This turns out to be very slow for the assembly that our jit compiler emits, since the vast majority of variables exist for only a single basic block. Instead of bin-packing, the much faster and simpler linear scan register allocation algorithm is now used.

    For a very large 200k-instruction script, compilation went from 19.5s to 2.8s (a 85% decrease).

    With this improvement, the 20k-instruction limit for jit has been removed. One example of the impact of this is that yuurand.zplay, which the jit compiler was previously mostly disabled for since most of its scripts are huge, now has all its scripts compiled. This increased the average FPS for yuurand.zplay from ~320 to ~2000, while still reducing the total time spent compiling.

    Theoretically this change in register allocation may produce slower runtime code, but there was no drop in FPS realized in the maths.zs stress test.

Tests