CRASH! [main/2026-09-06-1448] obj_formation_bar_Create_0:anonymous:82

[lyra-the-bot]

Your game just encountered a critical error! :(

System Details:

• Date-Time: 6-9-2026 20:28:28
• Game Version: main
• Build Date: 2026-09-06-1448
• Commit Hash: cf211c7

Save Details:

• Chapter Name: Fifth Interstellar
• Current Turn: 1
• Game Seed: 3933334266

Error Details:

Unable to find any instance for object index '-4' name ''

Stacktrace:

• obj_formation_bar_Create_0:anonymous:82
• obj_formation_bar_Draw_64:45

Reporter:

fredp

User Message:

N/A

[lyra-the-bot]

Investigation for #1521 - obj_formation_bar crash (instance '-4')

Crash signature

• Title: CRASH! [main/2026-09-06-1448] obj_formation_bar_Create_0:anonymous:82
• Error: Unable to find any instance for object index '-4' name '<undefined>'
• Stacktrace:
  • obj_formation_bar_Create_0:anonymous:82
  • obj_formation_bar_Draw_64:45
• Context: Turn 1, fresh chapter (Fifth Interstellar), so this is almost certainly the first visit to Formation Settings + first drag attempt.

-4 is noone in GML. This is the classic instance_nearest()-returned-noone dereference.

Code mapping

objects/obj_formation_bar/Draw_64.gml:44-46:

if (dragging == true) {
    drag_logic();
}

objects/obj_formation_bar/Create_0.gml:76-97, drag_logic():

drag_logic = function() {
    if (mouse_check_button(mb_left)) {
        x = mouse_consts[0] + rel_mousex;
        y = mouse_consts[1] + rel_mousey;
        obj_cursor.image_index = 3;
        col_target = instance_nearest(x, 224, obj_temp8);
        nearest_col = instance_nearest(col_target.x, col_target.y, obj_formation_bar);
        // ...

Line 82 (nearest_col = instance_nearest(col_target.x, ...)) is the reported crash line. It does not itself fail - it is the dereference of col_target.x on line 82 that fails when col_target == noone from line 81.

instance_nearest(x, y, obj) returns noone when zero instances of obj exist. There is no instance_exists() guard before either call, and no guard before line 84:

if (point_distance(col_target.x, col_target.y, nearest_col.x, nearest_col.y) > 2) {

Contrast with mouse_release() in the same file (lines 111, 130), which correctly uses instance_exists(mah_target) before touching mah_target.x / .col_parent.

Why can obj_temp8 be missing mid-drag?

obj_temp8 instances are the 10 column drop-zones, created in scripts/scr_ui_formation_bars/scr_ui_formation_bars.gml:178-182 (init_combat_bars / scr_ui_formation_bars()). At least three paths destroy them while dragging can still be true on a surviving obj_formation_bar:

1. Successful drop destroys its own drop-zones (most likely trigger).
   Create_0.gml:130-148 (mouse_release()):

   if (_in_drop_zone && (obj_controller.temp[te] + size <= 10)) {
       // ...
       with (obj_temp8) {
           instance_destroy();
       }
       with (obj_controller) {
           bar_fix = true;
       }
       exit;
   }

   This branch destroys all obj_temp8, sets bar_fix = true (which rebuilds everything on next obj_controller/Step_0.gml:22-25 via scr_ui_formation_bars()), and exits without clearing dragging, col_target, old_x/old_y. Next Draw_64 with LMB still held -> drag_logic() -> instance_nearest(..., obj_temp8) -> noone -> crash on col_target.x.

2. Destroy_0.gml has a global side effect.

   if (instance_exists(obj_temp8)) {
       with (obj_temp8) {
           instance_destroy();
       }
   }

   Destroying any one obj_formation_bar wipes all drop-zones for all remaining bars. During scr_ui_formation_bars() full rebuild or obj_controller/Step_0.gml:49-56 menu-exit cleanup this is redundant, and mid-drag it guarantees path (1).

3. Rebuild / menu-exit race.
   obj_controller/Step_0.gml:22-25 (bar_fix -> scr_ui_formation_bars()) and lines 49-56 (leave eMENU.FORMATIONS_SETTINGS -> destroy all bars + temp8) can run between Draw_64 frames while a drag is in progress.

All three converge on the same unguarded dereference.

Proposed fixes (in priority order)

Fix 1 (minimal, stops the crash): guard drag_logic() against noone.

drag_logic = function() {
    if (mouse_check_button(mb_left)) {
        if (!instance_exists(obj_temp8)) {
            nobar = true;
            col_target = noone;
            nearest_col = noone;
            return;
        }
        x = mouse_consts[0] + rel_mousex;
        y = mouse_consts[1] + rel_mousey;
        obj_cursor.image_index = 3;
        col_target = instance_nearest(x, 224, obj_temp8);
        if (!instance_exists(col_target)) {
            nobar = true;
            nearest_col = noone;
            return;
        }
        if (!instance_exists(obj_formation_bar)) {
            nearest_col = noone;
            nobar = true;
            return;
        }
        nearest_col = instance_nearest(col_target.x, col_target.y, obj_formation_bar);
        if (!instance_exists(nearest_col)) {
            nobar = true;
            return;
        }
        nobar = false;
        if (point_distance(col_target.x, col_target.y, nearest_col.x, nearest_col.y) > 2) {
            nobar = true;
        }
    } else {
        // existing button-released reset ...
    }
};

Also guard Draw_64.gml:44 call site if preferred: if (dragging == true && instance_exists(obj_temp8)).

Fix 2 (correct lifecycle): do not leave dragging == true after a successful drop.

In mouse_release() successful-drop branch, reset drag state before exit, mirroring the same-column early-out branch (lines 113-126):

obj_controller.temp[4800 + col_parent] -= size;
obj_controller.click = 1;
bat_for[@ obj_controller.formating] = mah_target.col_parent;
obj_cursor.dragging = 0;
obj_cursor.image_index = 0;
dragging = false;
col_target = 0;
nearest_col = 0;
nobar = false;
// then destroy temp8 / set bar_fix and exit

Alternatively, defer with (obj_temp8) instance_destroy() until after bar_fix rebuild (i.e. let scr_ui_formation_bars() do the cleanup), so drop-zones stay valid for the remainder of the drag.

Fix 3 (reduce blast radius): narrow Destroy_0.gml.

A single bar being destroyed should not wipe all 10 column markers for all other bars. Either delete the with (obj_temp8) block entirely (callers scr_ui_formation_bars() and obj_controller/Step_0.gml already destroy obj_temp8 explicitly), or only destroy it when the last bar is gone:

if (instance_number(obj_formation_bar) <= 1 && instance_exists(obj_temp8)) {
    with (obj_temp8) {
        instance_destroy();
    }
}

My recommendation: apply Fix 1 + Fix 2. Fix 1 alone stops this crash report; Fix 2 removes the one-frame window where every subsequent drag after a successful drop is running with no drop-zones. Fix 3 is optional hardening.

How to verify

1. Open Formation Settings on a new chapter, drag any bar (e.g. HQ) to another column, hold LMB one extra frame after drop - should no longer throw instance '-4'.
2. Repeat with column-full drop (snap-back path) and same-column drop.
3. Leave Formation Settings mid-drag (ESC / back) - no crash, no stuck obj_cursor.dragging == 1.
4. Feather / GoboCat clean; ASCII-only, snake_case, _ prefix for locals untouched.

Automated investigation only - no code changed yet. Happy to open a PR with Fix 1 + Fix 2 if maintainers agree.

[EttyKitty]

/oc Implement the proposed fixes 1 and 2. Fix 3 is allowed only if no side-effects verified via research beforehand. Proceed sequentially, verify each fix before proceeding to the next one. Commit each of them as separate commits. Follow the existing convention for commit names (under 50 symbols total, conventional commits type). Open a PR when done. Note that it closes this issue in the body.

[opencode-agent]

Created PR #1539

github run