Flaky behavior on trigger manipulation

[mighty1231]

Problem

This short plugin behaves as wrong way. Moreover, the project has flakiness. For each case euddraft compiles without any changes of project, the result map changes in many ways.

Expected behavior

• plugin: properly copy a trigger to another trigger.
• plugin: remove all the units of P1, and create battlecruisers on the center
• base map trigger: create a ghost and kill it for every frame

Observation

For each trial of euddraft compilation, the result map shows various behaviors.

1. No battlecruiser is observed. It generates EUD is not supported error after 2-3 seconds
2. No battlecruiser is observed. Trigger is stopped. The effect of killing ghost is not observed.
3. Battlecruisers are observed. However the ghost effect is not observed.

Files

issue.zip

It contains project, and three result map files with their own unique behaviors

Codes

project.edd

[main]
input: base.scx
output: result.scx

[test.py]

[freeze]
freeze: 0

test.py

from eudplib import *

trig_raw = Forward()
trig_inner = Forward()

@EUDFunc
def function():
    v = EUDVariable(1)
    if EUDIf()(v == 0):
        # Trigger 1 would not be executed
        trig_raw << RawTrigger(
            conditions = Always(),
            actions = [
                DisplayText("raw trigger"),
            ]
        )
        # Trigger 2 would not be executed
        trig_inner << RawTrigger(
            conditions = [
                Always(),
            ],
            actions = [
                CreateUnit(1, "Terran Marine", "Anywhere", P1)
            ]
        )
    EUDEndIf()

@EUDFunc
def main():
    function()
    f_repmovsd_epd(EPD(trig_inner)+2, EPD(trig_raw)+2, 2400//4)

    # Flaky behavior
    DoActions([RemoveUnit("(any unit)", P1),
               CreateUnit(100, "Terran Battlecruiser", "Anywhere", P1)])

def beforeTriggerExec():
    if EUDExecuteOnce()():
        main()
    EUDEndExecuteOnce()

[mighty1231]

@armoha reminded me that f_repmovsd_epd over trigger objects should be used carefully.

Although its GetDataSize returns 2408, it does not own all 2408 bytes. This is because of the compression on the allocation phase. It compresses trigger objects, allowing overlapping. RawTrigger is the only object that uses WriteSpace on allocation phase, and it determines overlap-able regions.